Get It Together, Intel: Core Ultra 9 285K CPU Review & Benchmarks vs. 7800X3D, 9950X, More

00:38:50
https://www.youtube.com/watch?v=XXLY8kEdR1c

Summary

TLDRThe video reviews the Intel Core Ultra 9 285K CPU, focusing on its power consumption, efficiency, and performance in gaming and production tasks. It highlights the challenges faced during testing, including issues with power connectors and BIOS settings. The 285K is compared to previous Intel CPUs and AMD alternatives, revealing that while it shows improvements in efficiency, it struggles against competitors in gaming performance. The video concludes that the 285K does not offer strong value for its price, especially compared to AMD's offerings and Intel's previous generation CPUs.

Takeaways

  • ⚡️ The Intel Core Ultra 9 285K is priced at $630.
  • 📉 It struggles against AMD's 7800 X3D in gaming performance.
  • 🔌 Power consumption issues were significant during testing.
  • 📊 The 285K shows improvements in efficiency over previous Intel CPUs.
  • 💻 It features a tile-based architecture with 8 P-cores and 16 E-cores.
  • 💰 The 285K does not provide strong value for its price.
  • 🔄 Alternatives include AMD's 7800 X3D and Intel's 13700K.
  • 🛠️ BIOS settings were not as Intel claimed, affecting performance.
  • 📈 The 285K has lower frame rates compared to the 14900K.
  • 🔮 The upcoming AMD 9800 X3D is a potential competitor.

Timeline

  • 00:00:00 - 00:05:00

    The review begins with an overview of the Intel Core Ultra 9 285k CPU, highlighting its power consumption and efficiency testing challenges due to its unique power connector design. The 285k is compared to previous Intel models, emphasizing its new naming convention and the competitive landscape against AMD processors.

  • 00:05:00 - 00:10:00

    The pricing landscape for the 285k is discussed, noting its pre-order price of $630, which is significantly higher than previous Intel models and AMD's offerings. The review emphasizes the importance of current market prices over launch MSRPs, indicating that the 285k is positioned against competitive AMD CPUs.

  • 00:10:00 - 00:15:00

    Intel's new architecture, referred to as Aake, is introduced, detailing its shift to a tile-based approach and the removal of hyperthreading. The 285k features a combination of performance cores and efficiency cores, requiring new motherboards that are incompatible with previous Intel sockets.

  • 00:15:00 - 00:20:00

    The review highlights several launch issues with the Aake platform, including problems with default BIOS settings and driver conflicts. Intel's failure to ensure proper functionality out of the box raises concerns about the readiness of the product for consumers.

  • 00:20:00 - 00:25:00

    Power consumption testing reveals that the 285k has higher power draw compared to the 14900k, particularly in the ATX 12vt rail. The review emphasizes the importance of isolating power consumption measurements to accurately assess efficiency and performance.

  • 00:25:00 - 00:30:00

    Efficiency metrics are presented, showing that the 285k has improved efficiency over the 14900k, but still lags behind AMD's offerings. The review provides detailed comparisons of performance per watt across various benchmarks, indicating that while the 285k shows improvements, it is not the most efficient CPU available.

  • 00:30:00 - 00:38:50

    The gaming performance of the 285k is evaluated against both Intel's previous models and AMD's latest CPUs. The review concludes that the 285k struggles to compete in gaming scenarios, with significant performance drops compared to its predecessors and AMD's offerings, ultimately questioning its value in the current market.

Show more

Mind Map

Video Q&A

  • What is the price of the Intel Core Ultra 9 285K?

    The Intel Core Ultra 9 285K is priced at $630, which is about $40 over its list price.

  • How does the 285K compare to the 14900K?

    The 285K has lower performance in gaming compared to the 14900K, despite improvements in efficiency.

  • What are the main issues faced during the review?

    Issues included problems with power consumption measurement and BIOS settings not being as Intel claimed.

  • How does the 285K perform in gaming?

    The 285K struggles against AMD's 7800 X3D and other Intel CPUs, showing lower frame rates.

  • Is the 285K efficient?

    Yes, the 285K shows improvements in efficiency, but it still lags behind AMD's offerings.

  • What is the architecture of the 285K?

    The 285K uses a tile-based approach with 8 P-cores and 16 E-cores.

  • What is the power consumption of the 285K?

    The 285K consumes about 162 Watts for ATX 12 volt and EPS 12 volt without GPU slot power.

  • What are the alternatives to the 285K?

    Alternatives include AMD's 7800 X3D and Intel's previous generation CPUs like the 13700K.

  • What is the conclusion of the review?

    The 285K does not provide strong value for its price, especially compared to AMD's CPUs.

  • What future CPUs are mentioned?

    The upcoming AMD 9800 X3D is mentioned as a potential competitor.

View more video summaries

Get instant access to free YouTube video summaries powered by AI!
Subtitles
en
Auto Scroll:
  • 00:00:00
    [Music]
  • 00:00:02
    reviewing the Intel Core Ultra 9 285k
  • 00:00:05
    CPU we're going to start with the name
  • 00:00:07
    actually we'll start with this take a
  • 00:00:09
    look at this giant spiky ATX 12vt line
  • 00:00:12
    for the 285k on this chart compared to
  • 00:00:14
    the 14900 K this is going to be a
  • 00:00:17
    problem for our power testing process
  • 00:00:19
    today and efficiency testing it's not as
  • 00:00:21
    simple as just measuring the EPS 12vt
  • 00:00:23
    cables because the differences from the
  • 00:00:25
    24 pin aren't just going to be the sort
  • 00:00:27
    of normal margin of error it's a lot
  • 00:00:29
    bigger than than that at least on Asus
  • 00:00:31
    we have to build this monstrosity to
  • 00:00:33
    isolate the power consumption of the CPU
  • 00:00:34
    for aake because Asus is pulling
  • 00:00:36
    significant power down the ATX 24 pin
  • 00:00:39
    connector this time and not just the EPS
  • 00:00:41
    12 volt cables for context if we didn't
  • 00:00:43
    do all this the 285k would look like
  • 00:00:46
    basically the most efficient CPU Intel
  • 00:00:48
    has ever made it's a big difference in
  • 00:00:50
    how the results get interpreted if it's
  • 00:00:52
    been a while since you've checked in you
  • 00:00:53
    can think of the 285k as a 15900 k the
  • 00:00:56
    245k as a 15600 K and the 26 5 as a 15
  • 00:01:01
    700k or something so they've changed the
  • 00:01:04
    naming a bit now Intel's big claim is
  • 00:01:06
    that the power consumption on these CPUs
  • 00:01:08
    is haved so we'll spend a lot of time
  • 00:01:10
    testing that we're also looking at the
  • 00:01:11
    usual gaming and production performance
  • 00:01:13
    production is a little more interesting
  • 00:01:14
    from a competitive standpoint because in
  • 00:01:16
    gaming Intel with the 285k is up against
  • 00:01:19
    not only the lowest prices that we've
  • 00:01:22
    seen for the 14 and 13 series but
  • 00:01:25
    relatively competitive prices for the
  • 00:01:27
    most part from AMD uh for the 7800 XD
  • 00:01:29
    and the 5700 x3d kind of asterisk there
  • 00:01:31
    because when the 7800 x3d sells out it
  • 00:01:33
    gets scalped for Crazy Prices and then
  • 00:01:35
    there's the 9800 x3d literally days away
  • 00:01:38
    at this point so it's going to be a
  • 00:01:39
    tough fight for Intel let's get into it
  • 00:01:41
    before that this video is brought to you
  • 00:01:43
    by thermal take and the tower 600 K the
  • 00:01:45
    tower 600 is a vertical case design with
  • 00:01:47
    thermal takes unique showcase
  • 00:01:49
    presentation which stands out further
  • 00:01:51
    with its separate chassis stand kit that
  • 00:01:53
    rotates the case for an angled showpiece
  • 00:01:55
    the tower 600 is heavily ventilated
  • 00:01:57
    around the sides has a ton of radiator
  • 00:01:59
    support including up to 420 mm Solutions
  • 00:02:02
    and offers two GPU mounting options for
  • 00:02:04
    display and cooling optimization it also
  • 00:02:06
    comes in colors not commonly seen in
  • 00:02:08
    cases learn more about the tower 600 at
  • 00:02:11
    the link in the description below we'll
  • 00:02:13
    start with a Roundup of the current
  • 00:02:14
    pricing landscape we put this table of
  • 00:02:16
    prices together a few days before launch
  • 00:02:18
    so the exact price is may be different
  • 00:02:19
    when this video goes live the 285k is
  • 00:02:22
    $630 from New Egg which is about $40
  • 00:02:25
    over the $590 list price on the slides
  • 00:02:29
    now it's possible that price comes down
  • 00:02:30
    a little bit before launch it's also
  • 00:02:32
    possible that some of the other prices
  • 00:02:34
    move up before launch however because
  • 00:02:37
    chronomancers haven't been a part of the
  • 00:02:39
    core rule set for D andd since Advanced
  • 00:02:41
    D andd version 2 as it's known today we
  • 00:02:44
    have no way to go into the future and
  • 00:02:46
    review that instead so we're going to
  • 00:02:47
    stay here and review the
  • 00:02:50
    present which will be the past when
  • 00:02:52
    you're watching you know what just never
  • 00:02:53
    mind the 285k is $630 with the pre-order
  • 00:02:56
    pricing or $160 more than the $14 900k
  • 00:03:00
    right now now we don't care about what
  • 00:03:01
    the launch MSRP was of those prior Parts
  • 00:03:04
    because it's irrelevant what they're
  • 00:03:05
    priced at today is what you would be
  • 00:03:07
    looking at if you were going to buy one
  • 00:03:09
    the 14 700k is $350 to $375 the 137 is
  • 00:03:13
    290 in some places 12 900k is down to
  • 00:03:16
    280 and the much weaker of course 12600
  • 00:03:18
    KF is now $160 which we only put here
  • 00:03:21
    because that's kind of crazy compared to
  • 00:03:23
    its launch buying into these dead
  • 00:03:24
    platforms isn't a great feeling though
  • 00:03:26
    so we can also look at amd's direct
  • 00:03:27
    Alternatives which include the $480 at
  • 00:03:30
    least right now 7800 x3d with a rumored
  • 00:03:32
    9800 x3d in a week so if that's true
  • 00:03:35
    we'll have more to say soon the 285k is
  • 00:03:38
    31% more money than the 7800x 3D at
  • 00:03:41
    least when it's priced properly we've
  • 00:03:43
    seen it go kind of crazy lately to like
  • 00:03:44
    600 and that doesn't make sense and on
  • 00:03:46
    the other side of the flank it's got the
  • 00:03:47
    9950 a and the 7950 a for production for
  • 00:03:50
    the absolute barebones Basics what you
  • 00:03:52
    need to know is that this code name is
  • 00:03:54
    aake it's the core Ultra 200 series
  • 00:03:55
    they've moved to a tile based approach
  • 00:03:58
    meaning that the Silicon if you were to
  • 00:03:59
    take the lid off is no longer a
  • 00:04:02
    monolithic piece of silicon also Intel
  • 00:04:05
    is working with tsmc on manufacturing
  • 00:04:07
    for this one so that's kind of unique
  • 00:04:08
    too Intel has gotten rid of
  • 00:04:10
    hyperthreading and it's instead moving
  • 00:04:12
    to just P cores and E cores with these
  • 00:04:14
    CPUs so the 285k has 8 p cores and 16 e
  • 00:04:18
    cores with 24 threads against the 14900
  • 00:04:21
    K's 32 total threads frequency is also
  • 00:04:23
    down now at 5.7 GHz advertised but
  • 00:04:26
    because the architectures are different
  • 00:04:27
    it's not that clean to just compare the
  • 00:04:29
    frequency numbers we'll review the 245k
  • 00:04:31
    next but that one's at 6p cores and AE
  • 00:04:34
    cores the new platform is LGA 1851 and
  • 00:04:37
    it requires new motherboards these will
  • 00:04:39
    not work in LGA 1700 socket motherboards
  • 00:04:42
    and LGA 1700 CPUs will not work in LGA
  • 00:04:45
    1851 boards there are also multiple
  • 00:04:47
    independent loading mechanism options
  • 00:04:49
    chosen by motherboard manufacturers
  • 00:04:51
    we'll have a separate video shortly with
  • 00:04:53
    laser scans and pressure maps of those
  • 00:04:54
    check back for that and more specs on
  • 00:04:57
    the basics can be found in our AOL Lake
  • 00:04:59
    announcement cover coverage that will
  • 00:05:00
    link below get into the review so first
  • 00:05:02
    off aake is really not ready for launch
  • 00:05:04
    they fumbled a number of things here
  • 00:05:06
    leading into launch and as an example
  • 00:05:08
    Intel told us this in a briefing quote
  • 00:05:10
    the big change is that every odm now
  • 00:05:13
    without fail every odm ships with this
  • 00:05:15
    or Apo in this conversation enabled in
  • 00:05:18
    the Bios the Camaro device is turned on
  • 00:05:21
    there will be a yellow Bane the driver
  • 00:05:22
    will install and APO will be enabled by
  • 00:05:25
    default that's the outof the-box
  • 00:05:27
    experience everybody is going to have
  • 00:05:28
    you'd want to make what folks are going
  • 00:05:30
    to see talking about the testing this is
  • 00:05:32
    going to be something that's going to be
  • 00:05:34
    probably just as important as the
  • 00:05:35
    hardware improvements all your odms the
  • 00:05:37
    board vendors it's all enabled by
  • 00:05:39
    default and should autoinstall 13th gen
  • 00:05:41
    onward now this is important because we
  • 00:05:43
    don't test with APO so we had to
  • 00:05:45
    consider it for this review Intel
  • 00:05:46
    presented first-party Benchmark slides
  • 00:05:48
    with APO enabled and emphasized with us
  • 00:05:51
    that Not only would it be on by default
  • 00:05:53
    on z890 but that it would also be on by
  • 00:05:55
    default with z790 neither was true we
  • 00:05:58
    discovered this issue and brought it to
  • 00:06:00
    Intel only because we weren't sure if it
  • 00:06:02
    was something we did wrong and it wasn't
  • 00:06:05
    the company seemed totally unaware of
  • 00:06:06
    this failing and after some back and
  • 00:06:08
    forth it realized eventually that it had
  • 00:06:10
    simply screwed up as a result of our
  • 00:06:12
    findings Intel is moving to publish a
  • 00:06:13
    list of motherboards that support APO
  • 00:06:15
    out of the box on by default and bios
  • 00:06:17
    going forward but our concern is that
  • 00:06:20
    clearly things are not fully ready here
  • 00:06:23
    if the statement was without fail a
  • 00:06:26
    feature would be on by default and then
  • 00:06:29
    it's not and we're the ones who find out
  • 00:06:31
    about it uh and they respond to that so
  • 00:06:34
    that's just we're not really supposed to
  • 00:06:36
    be beta testing it that's not a great
  • 00:06:37
    sign the dynamic tuning also no longer
  • 00:06:39
    existed on our board despite Intel
  • 00:06:41
    telling us that that's what needs to be
  • 00:06:43
    enabled the option we found has been
  • 00:06:45
    renamed to Innovation platform framework
  • 00:06:48
    and it was off by default which meant
  • 00:06:49
    APO was off for the 200 series with the
  • 00:06:52
    latest bios and micro code for z790 and
  • 00:06:55
    a brand new windows install we also
  • 00:06:57
    found APO was off so from method ology
  • 00:07:00
    perspective we consider APO a
  • 00:07:02
    non-default feature in the state the
  • 00:07:04
    product is in now we review what we
  • 00:07:06
    receive we don't review Promises of the
  • 00:07:08
    future C earlier reference of
  • 00:07:10
    chronomancer 1995 TSR publication I
  • 00:07:13
    believe it was version 1.2 but also
  • 00:07:16
    seriously we don't review the future
  • 00:07:17
    promises that it doesn't make any sense
  • 00:07:19
    so uh even with it on though which we
  • 00:07:22
    did test it was ultimately a relevant in
  • 00:07:23
    the test Suite didn't really matter much
  • 00:07:25
    it's more of an indicator it was one of
  • 00:07:26
    many unforced errors from Intel which
  • 00:07:28
    included blue screens of death on
  • 00:07:29
    windows installed due to driver conflict
  • 00:07:31
    with Nvidia devices when the igp is
  • 00:07:32
    disabled this issue didn't exist on
  • 00:07:34
    prior architectures seems to be a
  • 00:07:35
    combination of the new igp claiming pcie
  • 00:07:38
    resources and of nvidia's driver code
  • 00:07:40
    not having an error Handler for zero PCI
  • 00:07:42
    resource allocation you can bypass it by
  • 00:07:44
    disabling the igp but for a lot of
  • 00:07:46
    people this is going to be frustrating
  • 00:07:48
    the platform also has issues with easy
  • 00:07:50
    anti-che Which Intel will soon be
  • 00:07:52
    publishing a statement about we'll leave
  • 00:07:53
    that to Wendell's coverage but basically
  • 00:07:55
    if you disable security features you're
  • 00:07:57
    able to work around the issue we just
  • 00:07:58
    wouldn't recommend it Intel sent a
  • 00:08:00
    statement out about power profiles where
  • 00:08:01
    balanced power on 24 H2 results in
  • 00:08:04
    uncharacteristically bad performance we
  • 00:08:06
    test with high performance except we're
  • 00:08:08
    specifically required to use balanced in
  • 00:08:10
    some x3d situations so this didn't
  • 00:08:12
    affect us but it is another one of many
  • 00:08:14
    of Intel's launch problems in tell even
  • 00:08:16
    got its own specr on on this slide where
  • 00:08:18
    it said the chipset can support up to 32
  • 00:08:20
    USB 3.2 devices but actually it's
  • 00:08:23
    supposed to be 10 this is something that
  • 00:08:25
    Taiwanese publication unico's Hardware
  • 00:08:27
    spotted previously as we said a Lakes
  • 00:08:29
    really just not ready for launch the
  • 00:08:30
    numbers are what's really important here
  • 00:08:32
    including efficiency because that's what
  • 00:08:33
    Intel makes big claims about so uh to
  • 00:08:35
    keep it short on at least the z890
  • 00:08:37
    motherboard we have the 24 pin power
  • 00:08:39
    connector is now far more relevant than
  • 00:08:41
    it has been in the past so he's been a
  • 00:08:42
    little relevant but typically that kind
  • 00:08:44
    of those numbers come out in the noise
  • 00:08:46
    of data and now the 24 pin is doing
  • 00:08:48
    disproportionate work as likely a board
  • 00:08:51
    to board then not AOL Lake as an
  • 00:08:52
    architecture as a whole uh but we need
  • 00:08:55
    to look at that so let's get into that
  • 00:08:56
    part proving our work this plot shows
  • 00:08:58
    the 14900 case power draw on a known
  • 00:09:00
    workload spread across multiple rails
  • 00:09:02
    you can see that eps1 and eps2 power is
  • 00:09:04
    similar both around 70 to 80 watts in
  • 00:09:06
    this lighter weight gaming workload the
  • 00:09:08
    24 pin connector has several voltages
  • 00:09:10
    including 3.35 5vt standby and 12vt for
  • 00:09:13
    the things that are shown here the ATX
  • 00:09:15
    12vt rail includes things like fans and
  • 00:09:17
    if we hadn't isolated it slot power
  • 00:09:19
    among other miscellaneous controllers on
  • 00:09:21
    motherboards pcie slot poers had around
  • 00:09:23
    3035 Watts which is why it's important
  • 00:09:25
    to isolate it if capturing 24 pin power
  • 00:09:27
    here's why we isolate slot power this
  • 00:09:28
    benchmark was for seven zip on the 1400k
  • 00:09:30
    we'll come back to the 285k the GPU does
  • 00:09:33
    literally nothing in this test except
  • 00:09:34
    for spit out display it doesn't engage
  • 00:09:35
    in the test however occasionally in any
  • 00:09:38
    Tas we measure seemingly random spikes
  • 00:09:39
    to slot power it could be for a number
  • 00:09:41
    of unpredictable reasons such as Windows
  • 00:09:43
    background tasks or nvidia's drivers
  • 00:09:45
    doing something in the background like
  • 00:09:47
    Telemetry here's the comparison we've
  • 00:09:49
    all wanted in szip ignoring EPS 12vt we
  • 00:09:51
    noticed that the ATX 12vt power was
  • 00:09:53
    exceptionally high on the 285k and the
  • 00:09:55
    z890 hero combination compared to the
  • 00:09:57
    14900 K and z79 hero it's at about 50
  • 00:10:01
    Watts here whereas the 14900 K had ATX
  • 00:10:03
    12vt down around 30 Watts ATX 3vt power
  • 00:10:06
    is comparable on both and within two
  • 00:10:09
    watts of each other ATX 5 volt isn't
  • 00:10:11
    easy to call a simple average it
  • 00:10:13
    fluctuates a lot the range though is 14
  • 00:10:15
    to 30 Watts on the 285k and Asus
  • 00:10:17
    combination and similar on the 14900 K
  • 00:10:20
    with Peaks about 3 to 5 Watts lower than
  • 00:10:23
    on the 285k we don't know if Asus is
  • 00:10:25
    using any of that power to feed things
  • 00:10:27
    like maybe Regulators that go to the CPU
  • 00:10:29
    or something but if any of it is it's
  • 00:10:31
    not much A lot of it is driving other
  • 00:10:33
    components like IO the ATX 12vt line is
  • 00:10:35
    clearly important though as not
  • 00:10:37
    factoring in that 20 to 30 watt
  • 00:10:39
    difference in this test would mean
  • 00:10:41
    representing the 285k as artificially
  • 00:10:43
    efficient because the power is sort of
  • 00:10:45
    hidden in a cable where it's not
  • 00:10:47
    typically meaningfully high and where we
  • 00:10:49
    wouldn't normally look asus's reasoning
  • 00:10:51
    for this is theoretically better power
  • 00:10:53
    regulation we don't know if any other
  • 00:10:54
    board vendors are doing this finally as
  • 00:10:56
    an example of AMD here's a 7800 XD and
  • 00:10:58
    285 K the 285k and 14900 K ATX 12 volt
  • 00:11:01
    remained from last time the 7800 XD was
  • 00:11:04
    pulling about 12 Watts ATX 12 volt fixed
  • 00:11:06
    during this test and didn't seem to
  • 00:11:08
    fluctuate based on load at all it
  • 00:11:10
    appears fairly isolated typically this
  • 00:11:12
    gap of 10 watts comes out in the wash
  • 00:11:14
    when you're talking about a 280 wat
  • 00:11:16
    14900 K versus an 80 wat 7800 x3d so
  • 00:11:19
    it's just the usual part of the margins
  • 00:11:20
    it's never really been relevant before
  • 00:11:22
    in this scenario however the difference
  • 00:11:24
    against the 285k really starts to show
  • 00:11:26
    here's atx3 volt the 7800 x2d platform
  • 00:11:29
    is marginally higher here we're using an
  • 00:11:31
    Asus board for this also this is part of
  • 00:11:34
    why 5vt normally comes out in the wash
  • 00:11:36
    some rails are a few Watts higher some
  • 00:11:38
    are a few Watts lower there's always
  • 00:11:39
    going to be error ATX 5 volt is about
  • 00:11:41
    the same between them all there's
  • 00:11:42
    spikier behavior on the 285k from the io
  • 00:11:45
    but that's about it all this is done
  • 00:11:46
    with Hardware we also can't trust
  • 00:11:48
    various software readings because they
  • 00:11:49
    can be manipulated heavily by software
  • 00:11:52
    or by the motherboards years ago there
  • 00:11:53
    was an issue where AMD motherboards were
  • 00:11:55
    Under reporting some of the uh power
  • 00:11:58
    line items and in a way that allowed
  • 00:11:59
    boards to pull more or less power while
  • 00:12:01
    hiding it in software depending on what
  • 00:12:03
    they were trying to manipulate we also
  • 00:12:05
    spoke with d Bower about his experiences
  • 00:12:07
    and he noticed that CPU package power is
  • 00:12:09
    calculated based on seemingly VI so if
  • 00:12:12
    anything's wrong with the voltage or
  • 00:12:14
    tweaked it can manifest in erroneous
  • 00:12:16
    software reading so that's why we use
  • 00:12:18
    this monster that we built to do all of
  • 00:12:20
    this we're fairly confident in the
  • 00:12:21
    results but we have to cave out that
  • 00:12:23
    this is a very opaque platform
  • 00:12:25
    pre-launch right now and it's not fully
  • 00:12:27
    clear how all the power is being w it
  • 00:12:29
    for the most part should be proportional
  • 00:12:31
    to the vrm split anyway of 18 to4 of the
  • 00:12:33
    22 phases total so we were able to use
  • 00:12:36
    that as a guide post as well now finally
  • 00:12:38
    you might be asking who measures the
  • 00:12:40
    measurers the ansers us we we measure
  • 00:12:43
    the measurers so working with Elmore we
  • 00:12:46
    set up multiple current monitoring
  • 00:12:48
    devices to test them against each other
  • 00:12:50
    so we used a PMD a pmd2 a bench lab and
  • 00:12:52
    then not from Elmore we used a
  • 00:12:54
    straightforward fluke Uh current clamp
  • 00:12:57
    and using all of that we were able to
  • 00:12:59
    evaluate the current determine which
  • 00:13:01
    devices were producing the most reliable
  • 00:13:03
    results and then proceed with those for
  • 00:13:05
    testing so let's get into the efficiency
  • 00:13:07
    sevens of decompression efficiency is
  • 00:13:08
    measured in Ms per watt or millions of
  • 00:13:10
    instructions per jeel if you simplify it
  • 00:13:12
    in this test the 285k ended up in the
  • 00:13:14
    lower third of the chart inels 14900 K
  • 00:13:16
    pulled 273 Watts for this Benchmark with
  • 00:13:18
    a 285k now at 162 Watts for ATX 12 volt
  • 00:13:22
    and EPS 12vt without the slot power for
  • 00:13:25
    the GPU or 175 Watts if we look at the 8
  • 00:13:29
    ATX 12vt and the ATX 5vt with EPS 12vt
  • 00:13:32
    we think the 162 watt number is more
  • 00:13:34
    accurate but we want to transparently
  • 00:13:36
    present both since the power split is
  • 00:13:38
    still not fully clear because
  • 00:13:40
    performance is lower and the formula
  • 00:13:41
    relies on both the efficiency increase
  • 00:13:44
    isn't as impressive as the power drop is
  • 00:13:46
    the 285k runs with an efficiency of 1101
  • 00:13:48
    Ms per watt decompression with the
  • 00:13:51
    reading that includes 5 volts for that
  • 00:13:52
    one uh which results in an improvement
  • 00:13:55
    of 29% over the 14900 K the reading
  • 00:13:58
    without 5 volts is 40% improved over the
  • 00:14:01
    14900 k for efficiency which we think
  • 00:14:03
    again is the more accurate of the two
  • 00:14:05
    whether 30% or 40% is more precise the
  • 00:14:08
    point is that it's a big uplift AMD
  • 00:14:10
    still dominates here holding the entire
  • 00:14:12
    top third of the chart though the 7950 X
  • 00:14:15
    in eco mode sets a seemingly Untouchable
  • 00:14:17
    score at 1936 mips per watt and then the
  • 00:14:20
    7800 x3d which is much slower than the
  • 00:14:23
    285k and 7zip decompression scoring is
  • 00:14:26
    more efficient thanks to its 70 wat
  • 00:14:28
    power drop here the 7even zip
  • 00:14:29
    compression result power consumption is
  • 00:14:31
    measured across compression
  • 00:14:32
    decompression of the same test Suite so
  • 00:14:33
    those figures are unchanged the 285k
  • 00:14:36
    scored 1051 mips per watt with ATX 12
  • 00:14:38
    volt and EPS 12 volt which is an
  • 00:14:40
    improvement over the 149k 672 mips per
  • 00:14:43
    watt of 56% this uplift versus
  • 00:14:46
    decompression comes from a higher
  • 00:14:48
    performance in seven zip compression
  • 00:14:50
    scoring uh in terms of relative scaling
  • 00:14:53
    versus where it lands in decompression
  • 00:14:55
    we'll talk about that later AMD 7800 x3d
  • 00:14:57
    is at the top here again it's not the
  • 00:14:59
    best performer but it 70 wat power is
  • 00:15:01
    impressive and that allows it to remain
  • 00:15:02
    the true most efficient CPU we've tested
  • 00:15:04
    recently in a desktop balers Gate 3
  • 00:15:06
    could get a lot more complicated we have
  • 00:15:08
    however isolated out the PCI slot Power
  • 00:15:11
    by using an interposer to power the
  • 00:15:12
    video card instead the entire x3d lineup
  • 00:15:15
    continues its domination in this game
  • 00:15:18
    and most others the high frame rate and
  • 00:15:20
    low power land the 7800 x3d at 2.3 FPS
  • 00:15:23
    per watt as it's pulling only 55 Watts
  • 00:15:25
    during this workload for gaming the 5700
  • 00:15:27
    x2d is about the same with the 800 x2d
  • 00:15:29
    slightly less efficient Intel's new 285k
  • 00:15:32
    isn't remotely close to these numbers
  • 00:15:34
    the 285k with atx1 12vt and EPS 12vt ran
  • 00:15:37
    at 1.1 FPS per watt so the 7800 XD is
  • 00:15:40
    creating over two times as many frames
  • 00:15:42
    per watt uh of power here and even with
  • 00:15:45
    some overage for some sort of fudge
  • 00:15:48
    factor on a couple Watts here and there
  • 00:15:49
    that might not be filtered or accounted
  • 00:15:51
    for the 7800 x3d is still just it's
  • 00:15:54
    basically unapproachable at this point
  • 00:15:55
    the 14900 K ran at 0.7 FPS per watt here
  • 00:15:59
    looking at just Intel the 1.1 result of
  • 00:16:01
    the 285k has at least improved
  • 00:16:04
    substantially over the 14900 K with an
  • 00:16:06
    uplift of 57% if we include the 5vt
  • 00:16:09
    power since we're not certain if any of
  • 00:16:11
    it is being used for the CPU or how much
  • 00:16:13
    that would be a 29% uplift instead
  • 00:16:15
    Starfield is a heavier workload from
  • 00:16:17
    gaming in this one the 7800 X is again
  • 00:16:19
    the most efficient its result was 2.1
  • 00:16:21
    FPS per watt holding a large lead over
  • 00:16:23
    the closest CPUs that are up behind it
  • 00:16:25
    the next closest would be the 5700 x3d
  • 00:16:28
    also at about 71 Watts the 5800 X 9700x
  • 00:16:31
    and 7700 Nona follow this and we
  • 00:16:33
    eventually hit the 3700x and then the
  • 00:16:35
    285k with the ATX 12 volt and EPS 12 Vol
  • 00:16:38
    measurements and without the pcie slot
  • 00:16:40
    we end up at about 147 Watts for a 1.0
  • 00:16:43
    FPS per watt rating with some of the
  • 00:16:45
    usual rounding in there the 14900 K ran
  • 00:16:48
    at 0.7 FPS per watt here so Intel has
  • 00:16:50
    improved by 43% it's Ultra 9 is now
  • 00:16:53
    where the 14600 K was previously
  • 00:16:56
    measuring 5 volts as well had it at 72
  • 00:16:59
    Watts or 0.8 FPS per watt that would
  • 00:17:02
    instead be an improvement of 14% as
  • 00:17:04
    before we believe it's somewhere in the
  • 00:17:06
    middle here but probably closer to that
  • 00:17:08
    1.0 figure the next game is Stellaris
  • 00:17:11
    where we measured simulation time in
  • 00:17:12
    seconds and Power in order to make this
  • 00:17:15
    as easy to follow as possible we've done
  • 00:17:17
    some simple arithmetic to convert this
  • 00:17:18
    into simulations per watt hour meaning
  • 00:17:21
    that a bigger number is better despite
  • 00:17:24
    the base metric result being lower is
  • 00:17:26
    better that's because you want to
  • 00:17:27
    simulate more things
  • 00:17:29
    with a watt hour the 7800 x2d has an
  • 00:17:32
    impressive score here at 2.7 simulations
  • 00:17:34
    per wat hour the next closest is down at
  • 00:17:36
    2.4 with the 5700 x2d the 285k ran 1.5
  • 00:17:40
    simulations per watt hour that means for
  • 00:17:43
    each watt hour the 7800 x2d is able to
  • 00:17:45
    complete 1.2 more simulations or about
  • 00:17:47
    an 80% increase the 14900 K completed
  • 00:17:50
    0.9 simulations per wat hour that means
  • 00:17:52
    the 285 K's best entry is an improvement
  • 00:17:55
    of 67% in efficiency remember most of
  • 00:17:57
    Intel's presented numbers were only for
  • 00:18:00
    power reductions not necessarily for
  • 00:18:02
    efficiency the 285k simulation
  • 00:18:05
    performance is better than the 14900 K
  • 00:18:06
    is in this Benchmark so it's combination
  • 00:18:08
    of higher performance and lower power
  • 00:18:10
    which was reduced by a total of 46.6
  • 00:18:13
    Watts from the 14900 Cas that falls
  • 00:18:15
    basically once you look at Intel doing
  • 00:18:17
    total system power sort of exactly
  • 00:18:19
    within their geoman expectation uh this
  • 00:18:21
    yields the huge increase that we're
  • 00:18:23
    seeing in efficiency though finally
  • 00:18:25
    Final Fantasy 14 for efficiency in this
  • 00:18:27
    one the 7800 X 3D LED with an impressive
  • 00:18:29
    42 and2 Watts power draw while spewing
  • 00:18:31
    hundreds of frames per second allowing
  • 00:18:33
    it a result of 8.3 FPS per watt that's
  • 00:18:36
    great the 5700 X keeps it second place
  • 00:18:39
    spot at 7.7 the 14900 K ran at 3.1 FPS
  • 00:18:42
    per watt so the 285k at it sort best is
  • 00:18:44
    32% improved by these numbers this is
  • 00:18:47
    one where the 285k was regressive in
  • 00:18:49
    performance against the 14900 K and by
  • 00:18:52
    sort of a lot we'll come back to that in
  • 00:18:54
    fps results but because the FPS is down
  • 00:18:57
    for this one the result is reduced
  • 00:18:59
    efficiency gains despite also being
  • 00:19:02
    reduced power next we'll validate the
  • 00:19:03
    frequency Behavior to establish how the
  • 00:19:05
    CPUs behave out of the box this chart
  • 00:19:07
    plots the highest single core frequency
  • 00:19:08
    per interval during a single threaded
  • 00:19:10
    cinebench workload with a 285k the max
  • 00:19:13
    single core frequency is 5700 MZ during
  • 00:19:15
    this test with frequent drops to 5400
  • 00:19:17
    MHz these drops are abnormal on Intel
  • 00:19:20
    during this testing the 14900 K with the
  • 00:19:22
    latest micro code held 6,000 MHz these
  • 00:19:24
    are different architectures so it's not
  • 00:19:26
    as simple as stating that higher is
  • 00:19:28
    better but it will help illustrate where
  • 00:19:30
    some of the performance losses are
  • 00:19:31
    coming from later the original launch
  • 00:19:33
    micro code for the 14900 K had it at the
  • 00:19:35
    same frequency this chart looks at PE
  • 00:19:37
    core frequency averages during blender
  • 00:19:39
    the 285k maintains an average PE core
  • 00:19:41
    frequency of 5400 MHz in this testing
  • 00:19:43
    it's also relatively flat that has it
  • 00:19:46
    higher than the 2024 Asus test entry
  • 00:19:48
    which is what we use for our review
  • 00:19:49
    comparison today at around 5100 to 5220
  • 00:19:51
    MHz the 2023 launch entry was higher at
  • 00:19:54
    5280 to 5,400 MHz on average the micro
  • 00:19:57
    code changes may have affected this
  • 00:19:58
    since launch though that's the point of
  • 00:20:00
    doing these the MSI zx11d micro code
  • 00:20:03
    entry had the CPU at around 5,000 MHz
  • 00:20:05
    from the same window as the Asus 2023
  • 00:20:07
    entry we never Ed that board in any of
  • 00:20:09
    our reviews process so this is new data
  • 00:20:11
    uh with old micro code and the most up
  • 00:20:14
    today frequency entry sits between the
  • 00:20:16
    two flanks of the early 14900 K tests
  • 00:20:19
    finally this chart shows all core
  • 00:20:21
    averages in the same test the 285k
  • 00:20:23
    averaged 4867 MHz when factured in the
  • 00:20:25
    ecores with the modern 14900 K just
  • 00:20:28
    below 45 500 MHz and flanked by the two
  • 00:20:30
    older entries now we're getting into
  • 00:20:31
    gaming benchmarks we have a lot of 2023
  • 00:20:33
    and 2024 games in this test we with a
  • 00:20:35
    couple of Main Stays from the past
  • 00:20:37
    Dragons Dogma 2 is up now this is one of
  • 00:20:39
    the 2024 titles we added to our suite
  • 00:20:41
    last time it's had updates that affect
  • 00:20:43
    performance significantly since our last
  • 00:20:44
    round a benchmark so this is all new
  • 00:20:46
    data like everything else here APO
  • 00:20:48
    doesn't do anything in this Benchmark
  • 00:20:49
    and isn't supported we tested it with on
  • 00:20:52
    and off for the 4900 K and 285k and its
  • 00:20:55
    performance was the same so we removed
  • 00:20:56
    the Redundant entries here because they
  • 00:20:58
    they have no impact we didn't want to
  • 00:20:59
    take the chart space the 285k landed at
  • 00:21:02
    about 104 FPS average with lows at 64
  • 00:21:05
    and 51 this allows the AMD 7800 x3d at
  • 00:21:08
    $480 to lead the $590 285k by 6.1% with
  • 00:21:14
    the 14900 K leading at 5% over the 285k
  • 00:21:18
    the lows between these three CPUs are
  • 00:21:20
    functionally identical so there's no
  • 00:21:21
    explicit Advantage there the 14700 K and
  • 00:21:24
    13700 K also lead the 285k here with the
  • 00:21:28
    14 00 K far cheaper at $350 not even
  • 00:21:32
    close the 5700 x3d and 5600 x3d are
  • 00:21:35
    effectively tied with the 5700 x3d at
  • 00:21:38
    $230 right now or sometimes down closer
  • 00:21:40
    to 200 depending on where you buy it an
  • 00:21:42
    f124 which for hopefully obvious reasons
  • 00:21:45
    is a 2024 title we end up with this set
  • 00:21:48
    of results when fully retested in 24 H2
  • 00:21:51
    and with the new micro code in the GSA
  • 00:21:52
    for everything the 7800 XD sets a
  • 00:21:54
    sealing pretty high and establishes a
  • 00:21:56
    28% lead over the more expensive and
  • 00:21:59
    higher power consuming 285k at 438 fps
  • 00:22:03
    to 344 FPS average the 149k is also
  • 00:22:07
    advantaged This Time by 12% with a 385
  • 00:22:10
    FPS result the lows for the 285k were
  • 00:22:12
    186 FPS 1% which is a worse entry than
  • 00:22:15
    the 14900 K's 250 result the frame time
  • 00:22:18
    pacing appears better in step with the
  • 00:22:19
    average on both the 14900 K and the 7800
  • 00:22:22
    x3d looking now at the APO results
  • 00:22:25
    toggling APO appears to do something in
  • 00:22:28
    this game the 285k gained 1.3% with APO
  • 00:22:31
    on so that's very exciting we can hardly
  • 00:22:36
    contain ourselves I'm trying to keep it
  • 00:22:38
    in right now you'll surely notice the
  • 00:22:40
    extra 3 to four frames per second on top
  • 00:22:42
    of the other 343 before them and we're
  • 00:22:46
    glad that we spent half a day
  • 00:22:47
    troubleshooting Intel's fumbled
  • 00:22:48
    pre-launch settings to gain those frames
  • 00:22:50
    back because whatever would we have done
  • 00:22:53
    without the extra
  • 00:22:54
    0.369 milliseconds reduction in frame
  • 00:22:57
    time if had not troubl shot that for
  • 00:23:00
    half a day so that was phenomenally
  • 00:23:02
    exciting the 14900 K also gained about
  • 00:23:04
    the same amount it is irrelevant on both
  • 00:23:07
    using ddr5 8600 on the 285k with Gear 2
  • 00:23:10
    and blasted vdm boosted the 359 FPS
  • 00:23:13
    average result we see here that's a 4.5%
  • 00:23:16
    likeforlike lead over the 344 result
  • 00:23:18
    this is almost enough to get the 285k
  • 00:23:21
    tied with the two generation old 13 700k
  • 00:23:23
    from 2022 but not quite it does however
  • 00:23:25
    tie it with amd's 2020 architecture
  • 00:23:27
    found in the cash boo boosted 5700 x3d
  • 00:23:30
    with its lower power 1440p results are
  • 00:23:32
    mostly uninteresting the top has
  • 00:23:34
    truncated heavily because the increase
  • 00:23:35
    in resolution even still the new Ultra 9
  • 00:23:37
    ends up in the middle it's fitting that
  • 00:23:39
    the core Ultra 9 is in the middle of a
  • 00:23:42
    cluster here Final Fantasy 14 Dawn
  • 00:23:43
    Trails up now this is another 2024
  • 00:23:45
    addition to our testing the 7800 x3d
  • 00:23:47
    again establishes the ceiling it leads
  • 00:23:49
    the 285k by 31% in this game at 353 fps
  • 00:23:53
    to 270 the 14900 K is 310 FPS average
  • 00:23:56
    establishes a 15% lead over the 285k
  • 00:23:59
    marking one of the largest declines in
  • 00:24:01
    Intel's performance in our gaming Suite
  • 00:24:03
    the 5800 x2d and 5600 x2d are also at
  • 00:24:05
    the top here showing that da tril
  • 00:24:07
    generally likes cash at least on these
  • 00:24:09
    CPUs the 5700x 3D is a little bit lower
  • 00:24:12
    down since its frequency is 300 MHz
  • 00:24:15
    below that of the 5600 x3d so these
  • 00:24:17
    results make sense but it shows that da
  • 00:24:19
    tril is also sensitive to the frequency
  • 00:24:22
    the 285k ends up below the 14700 K and
  • 00:24:25
    13700 K with APO doing nothing on the
  • 00:24:28
    285k it's within error Final Fantasy 14
  • 00:24:32
    is officially supported by APO though
  • 00:24:34
    and it's on the games list and we do see
  • 00:24:37
    a slight uplift on the 14900 K of 1%
  • 00:24:40
    although that's basically error
  • 00:24:42
    switching to DDR 58600 on the 285k
  • 00:24:45
    provided an uplift of 4% with it still
  • 00:24:47
    below the 13700 K after that of course
  • 00:24:50
    the same memory treatment would lift the
  • 00:24:52
    others up also it just be an arms race
  • 00:24:54
    at 1440p its limited ceiling is
  • 00:24:56
    shuffling the stack as a result of the
  • 00:24:58
    frame rate bouncing off of other
  • 00:24:59
    limitations we saw a 1.9% uplift from
  • 00:25:02
    APO on the 14900 K the 285k again did
  • 00:25:06
    not benefit from APO with both entries
  • 00:25:08
    at 255 FPS average plus or minus one FPS
  • 00:25:11
    Ballers Gate 3 is up now a 2023 title
  • 00:25:14
    that we added to our permanent test
  • 00:25:15
    Suite this year we test an act three in
  • 00:25:17
    a densely populated city for a heavy CPU
  • 00:25:20
    load the 7800 x3d establishes a 26% lead
  • 00:25:23
    over the 285k at 126 fps to 100 FPS
  • 00:25:27
    average the 149 K leads the 285k by 42%
  • 00:25:31
    at about 105 FPS average the 5800 x3d 57
  • 00:25:34
    x3d and 56 x3d also lead the 285k as do
  • 00:25:38
    the 13900 K 13700 K and 14700 K toggling
  • 00:25:42
    APO did nothing here and was within
  • 00:25:44
    variance for both the 14900 K and 285k
  • 00:25:47
    swapping to DDR 58600 memory boosted the
  • 00:25:50
    285k by 7.6% allowing it to pass
  • 00:25:53
    everything except the x3d CPUs this also
  • 00:25:55
    perfectly aligns with why x3d does so
  • 00:25:58
    well here of course shoving better
  • 00:26:00
    memory into the 14900 K or 9000 Series
  • 00:26:02
    would also boost them fairly
  • 00:26:03
    proportionally Stellaris is up next
  • 00:26:05
    first up APO has no impact in this game
  • 00:26:07
    at least not in the simulation time the
  • 00:26:08
    285k required 32 and 1 half seconds for
  • 00:26:11
    the simulation finally putting it ahead
  • 00:26:13
    of the 49k the 285k requires 1 to 2%
  • 00:26:17
    less time than the 147 and 14900 K CPUs
  • 00:26:20
    the 7800 XD outranks the 285k again and
  • 00:26:24
    requires about 3.7% less time for the
  • 00:26:27
    work the 97 X Remains the fastest here
  • 00:26:29
    which is consistent with our last round
  • 00:26:31
    of data and boosting memory helps in
  • 00:26:33
    this game so the 285k benefited from a
  • 00:26:35
    4.3% reduction in time to complete the
  • 00:26:38
    simulation Rainbow Six Siege is up next
  • 00:26:39
    in this one the 7800 x2d leads by a much
  • 00:26:41
    smaller amount at only 6.6% for 622
  • 00:26:44
    versus 584 FPS average when APO is in
  • 00:26:47
    its default off position that this board
  • 00:26:49
    shipped in the 14900 K was tied enabling
  • 00:26:52
    APO did nothing on the 285k the result
  • 00:26:54
    was within error when considering we're
  • 00:26:56
    at almost 600 FPS and anyway enabling it
  • 00:26:59
    boosted the 14900 K by 1.7% and this is
  • 00:27:03
    with Intel's shitty Microsoft Store app
  • 00:27:05
    confirming that it's enabled for both
  • 00:27:07
    CPUs this boost is reduced for the 1400k
  • 00:27:10
    versus apo's launch because Rainbow 6
  • 00:27:13
    has updated the code that was causing
  • 00:27:15
    problems going DDR 58600 increase
  • 00:27:17
    performance over the 584 result by 2.6%
  • 00:27:20
    the 285k is at least better than the
  • 00:27:22
    14700 K in this one with an uplift of 2%
  • 00:27:26
    looking at the stock entries it's also
  • 00:27:27
    about 1.8% ahead of the 7700 non-x which
  • 00:27:31
    has been $280 to $290 when you can still
  • 00:27:33
    find it as a general note this game has
  • 00:27:35
    some serious 0.1% low consistency issues
  • 00:27:38
    actually we noticed that the 12 to 14
  • 00:27:41
    series CPUs have significantly higher
  • 00:27:43
    0.1% lows than amd's options and then
  • 00:27:46
    the 285k alike so it's easier to look at
  • 00:27:48
    a frame time plot here it is because
  • 00:27:50
    Intel noted that Rainbow 6 had
  • 00:27:52
    specifically tuned to reduce Reliance on
  • 00:27:54
    APO and because of the low advantage of
  • 00:27:56
    Alder Lake onward we're assuming that
  • 00:27:58
    Rainbow Six Siege has specific and
  • 00:27:59
    manual tuning by the Developers for
  • 00:28:01
    Alder like architecture and its more
  • 00:28:03
    recent derivatives this Frame time plot
  • 00:28:04
    shows the 149k is overall highly
  • 00:28:06
    consistent frame to frame with most of
  • 00:28:08
    the intervals deviating at most by 2
  • 00:28:10
    milliseconds there are a few spikes but
  • 00:28:12
    only one notable Spike to 6 milliseconds
  • 00:28:14
    which is a 4 millisecond deviation
  • 00:28:16
    unless it wouldn't be noticed by the
  • 00:28:18
    vast majority of users adding the 285k
  • 00:28:20
    we see several spikes to 10 to 11
  • 00:28:21
    milliseconds by themselves these frame
  • 00:28:23
    times aren't bad UH 60 FPS would be
  • 00:28:25
    16.67 milliseconds so this is isn't a
  • 00:28:28
    huge stall but it's a change from rafter
  • 00:28:30
    Lake and it's objectively worse in
  • 00:28:32
    Starfield another 2023 game the 7800 X
  • 00:28:34
    maintains a 2% advantage over the 285k
  • 00:28:37
    at
  • 00:28:38
    145.47 5 FPS average normally we
  • 00:28:41
    wouldn't read the decimal points but
  • 00:28:43
    here they are significant to showing any
  • 00:28:45
    kind of percent scale we're bound
  • 00:28:47
    elsewhere here and that elsewhere is
  • 00:28:49
    obvious and pretty cool actually the
  • 00:28:51
    ddr5 8600 result jumps to the top of the
  • 00:28:53
    chart illustrating that at least some of
  • 00:28:55
    our limit here is memory as a reminder
  • 00:28:57
    applying the memory tuning treatment to
  • 00:28:59
    everything would boost all numbers so
  • 00:29:00
    looking at Baseline will give you an
  • 00:29:01
    idea for sort of their head room plus
  • 00:29:03
    the Baseline the 149 her K actually
  • 00:29:05
    falls behind for once it's at a slight
  • 00:29:07
    loss to the 285k we think this
  • 00:29:09
    corresponds with the cash advantages
  • 00:29:11
    based on the memory performance that we
  • 00:29:12
    just saw where it's really using uh cash
  • 00:29:15
    and memory APO once again does nothing
  • 00:29:18
    between the two main 285k entries Total
  • 00:29:21
    War Warhammer 3 is up now the 7800 x2d
  • 00:29:23
    led the 285k by 7.2% with the 14900 K
  • 00:29:26
    leading by 2.8%
  • 00:29:28
    total Warhammer 3 is on the APO list so
  • 00:29:31
    it is supported on the 14900 K we saw an
  • 00:29:33
    uplift of about 1.2% if that's
  • 00:29:36
    considered supported then great on the
  • 00:29:39
    285k we measured an improvement of 0.6%
  • 00:29:43
    which is actually just within the
  • 00:29:44
    run-to-run variant lows in this game
  • 00:29:46
    have been historically awful on Intel's
  • 00:29:49
    I9 CPUs which we previously proved is
  • 00:29:51
    due to scheduling challenges with the
  • 00:29:52
    thread count the 285k seems to at least
  • 00:29:55
    somewhat improve here possibly by
  • 00:29:57
    reduction in threads uh this is
  • 00:29:59
    something that we've seen historically
  • 00:30:00
    we looked at this back in Alder Lake
  • 00:30:02
    we're getting to production benchmarks
  • 00:30:03
    now run through these pretty fast so
  • 00:30:05
    blenders up first we've updated it it's
  • 00:30:06
    not comparable to the previous results
  • 00:30:08
    the 285k is our second highest performer
  • 00:30:10
    on the charts for desktop class CPUs the
  • 00:30:12
    950x benefits from a render time
  • 00:30:14
    reduction of 5.6% against the 285k while
  • 00:30:17
    the 285k reduces render time from the
  • 00:30:19
    4900 K by an actually meaningful 16.5%
  • 00:30:23
    so finally some uplift these types of
  • 00:30:26
    tests are also wear AMD 7800x and 5800
  • 00:30:28
    x2d show their limits as eight core CPUs
  • 00:30:31
    both falling far down the charts that's
  • 00:30:33
    a known quantity at this point if you're
  • 00:30:35
    only gaming they make the most sense if
  • 00:30:37
    you mix in a good amount of heavily
  • 00:30:39
    threaded software for work it may make
  • 00:30:41
    more sense to buy something else on the
  • 00:30:42
    list in file comprar with 7even zip the
  • 00:30:44
    285k completed 170,000 mips and roughly
  • 00:30:47
    tied with the i7 14700 K the 14900 K
  • 00:30:51
    holds an advantage of 8% the 950x and
  • 00:30:53
    750x outperform the 285k significantly
  • 00:30:56
    with the 9900x just behind memory seem
  • 00:30:59
    to really help in this test pushing the
  • 00:31:00
    285k up by gargantuan 18.7% which is a
  • 00:31:04
    huge uplift decompression doesn't
  • 00:31:06
    benefit as much posting a 4% Improvement
  • 00:31:08
    for the memory the stack also shuffles
  • 00:31:10
    with the 9900x and 7900x both surpassing
  • 00:31:13
    the 285k here the memory subsystem
  • 00:31:15
    benefits don't translate as much as they
  • 00:31:17
    did in compression both stock and with
  • 00:31:20
    better Ram the 14900 K leads the 285k by
  • 00:31:23
    a staggering 21% in this Benchmark
  • 00:31:25
    benefiting from its extra threads the
  • 00:31:27
    9950 DX establishes an impressive ceing
  • 00:31:29
    at 42% ahead of the 285k next is our
  • 00:31:32
    chromium code compile test for this test
  • 00:31:34
    we double the ram capacity on some CPUs
  • 00:31:37
    when they page out or otherwise fail to
  • 00:31:39
    complete the compile this only occurs
  • 00:31:41
    with the highest performing few CPUs on
  • 00:31:43
    the chart this test has been updated
  • 00:31:45
    since the last round so load on CPUs has
  • 00:31:47
    slightly changed in the situation of the
  • 00:31:48
    950x there's one disclaimer here we were
  • 00:31:51
    not able to get it stable at DDR 56000
  • 00:31:53
    with 64 GB when running our usual
  • 00:31:56
    tighter timings we experiment Ed briefly
  • 00:31:58
    with loosening the timings uh raising
  • 00:32:00
    the voltages and reducing the infinity
  • 00:32:02
    fabric none of that really worked and so
  • 00:32:05
    we ended up keeping everything the same
  • 00:32:07
    but pulling the frequency down to DDR
  • 00:32:09
    55600 to accommodate 64 GB we had to
  • 00:32:11
    call out there for now we'll look into
  • 00:32:13
    that more later so that makes the 950x
  • 00:32:15
    an imperfect comparison with that
  • 00:32:18
    limitation disclosed due to the
  • 00:32:19
    stability issues the 950x and 285k are
  • 00:32:22
    roughly equal that may change if we can
  • 00:32:25
    rerun the 950x and find stability for
  • 00:32:27
    direct comparisons the 285k leads the
  • 00:32:30
    14900 K by only about a 3 minute time
  • 00:32:32
    reduction or a total compile time
  • 00:32:34
    requirement reduction of 3.8% as usual
  • 00:32:37
    the8 core x3d parts are lower down the
  • 00:32:39
    stack like the 78 and 58 x3ds now we're
  • 00:32:42
    testing the Adobe software with the
  • 00:32:43
    Puget Suite kind of like how the Adobe
  • 00:32:45
    software tests my patients every single
  • 00:32:48
    day by it existing Premiere has the 285k
  • 00:32:51
    as the new chart topper at 11,336 points
  • 00:32:55
    that's how this whole review should have
  • 00:32:56
    been especially with the power reduction
  • 00:32:58
    and unfortunately it was not the lead
  • 00:33:00
    over the 14900 K is small at 2.5% uplift
  • 00:33:03
    in aggregate the 9950 X is amd's top CPU
  • 00:33:06
    here at 10,914 points the 285k therefore
  • 00:33:10
    has a 3.9% lead in Adobe Photoshop AMD
  • 00:33:12
    has a clean sweep of the entire top half
  • 00:33:14
    of the chart those of you who've been
  • 00:33:16
    watching us for at least 5 years will
  • 00:33:17
    remember an era where it was actually
  • 00:33:19
    the opposite the first Intel entry
  • 00:33:20
    doesn't appear until below the 7600 X
  • 00:33:23
    and that's the 14900 K the 950x leads
  • 00:33:25
    the 285k by 22% here and the 14900 K
  • 00:33:28
    leads it by three and a half%
  • 00:33:30
    simplifying all this we had some more
  • 00:33:31
    data too for spec and stuff like that
  • 00:33:33
    but it kind of all comes out the same
  • 00:33:35
    way which is the 285k does not have a
  • 00:33:37
    particularly Strong place in the market
  • 00:33:40
    with its competitors right now including
  • 00:33:42
    Intel's own competition which is lower
  • 00:33:44
    in price than it used to be the 7800 x3d
  • 00:33:46
    is far superior in both gaming and
  • 00:33:48
    efficiency the 285 hardly makes any
  • 00:33:50
    sense even against the 5700 x2d and
  • 00:33:52
    amd's 9800 x2d is days away at this
  • 00:33:54
    point for non-gaming the CPU is more
  • 00:33:56
    interesting but we still don't think
  • 00:33:57
    it's broadly justifiable at the current
  • 00:33:59
    CPU and platform costs even Intel's
  • 00:34:01
    prior gen parts are far cheaper right
  • 00:34:03
    now like the 13700 K and 14700 K which
  • 00:34:06
    theoretically have been fixed with micro
  • 00:34:07
    Cod if you were to argue that the
  • 00:34:09
    inefficiency of the prior gen Parts
  • 00:34:11
    counters some of the Savings in the form
  • 00:34:13
    of lower power being lower cost because
  • 00:34:16
    of lower power it's just not a strong
  • 00:34:18
    argument the 14700 K is $350 now so it's
  • 00:34:21
    $28 cheaper than a 285k if your power
  • 00:34:24
    cost 10 cents per kilowatt hour like
  • 00:34:25
    hours then you need to play the highest
  • 00:34:27
    CP you load games for 8 hours a day for
  • 00:34:29
    an entire year in order to gain $19 of
  • 00:34:31
    value in power bill reductions for the
  • 00:34:34
    285k against the 14700 k at 15 cents the
  • 00:34:37
    shut in gaming levels would give the
  • 00:34:38
    285k a savings of $30 over one year of
  • 00:34:42
    literally playing Starfield or cyberp
  • 00:34:43
    punk as a full-time job at the peak of
  • 00:34:45
    degeneracy that we all strive to achieve
  • 00:34:47
    you'd have to play games at this pace
  • 00:34:49
    for 10 years straight to have spent more
  • 00:34:50
    on power than the cost Savings of buying
  • 00:34:52
    the cheaper CPU it's the Delta not the
  • 00:34:54
    total cost maybe throw in a cheaper
  • 00:34:56
    cooler as well since the power's down
  • 00:34:59
    call it 8 years even at 30 cents per
  • 00:35:00
    kilowatt hour the $280 price savings of
  • 00:35:02
    the 14700 K would require 5 years to
  • 00:35:04
    wipe out the power bill while playing
  • 00:35:06
    cyberp punk8 hours a day for 5 years or
  • 00:35:08
    in other words
  • 00:35:09
    14560 hours of cyberpunk let's just say
  • 00:35:12
    you pay 50 cents per kilowatt hour in
  • 00:35:14
    one standard three-year upgrade cycle
  • 00:35:16
    you can save the amount of money and
  • 00:35:18
    power you'd save buying the 14700 K
  • 00:35:20
    which we still don't recommend for
  • 00:35:22
    gaming by the way we're just using it as
  • 00:35:24
    an example that'd be a savings of $93
  • 00:35:27
    per year so at 37.7 minus 25.1 or you go
  • 00:35:31
    with the 7800 x2d and get higher frame
  • 00:35:33
    rates and still a power cost reduction
  • 00:35:35
    even at these high electricity prices of
  • 00:35:37
    $114 per year at these crazy gaming
  • 00:35:40
    hours and if we ignore all the cost and
  • 00:35:42
    we say it's purely about stewardship for
  • 00:35:45
    the environment then the 7800 x2d still
  • 00:35:47
    makes way more sense because it's just
  • 00:35:49
    lower power so it is good that Intel has
  • 00:35:52
    improved over its prior
  • 00:35:53
    CPUs but this truly is a building block
  • 00:35:57
    type of of situation there needs to be
  • 00:35:59
    more uh before it's truly competitive in
  • 00:36:02
    that department enough where you just
  • 00:36:03
    point to the one thing and you say yeah
  • 00:36:05
    that's the reason you buyas it's not
  • 00:36:06
    there yet the reduction power
  • 00:36:07
    consumption is important though
  • 00:36:08
    objectively Intel is more efficient than
  • 00:36:10
    it used to be it's just that the
  • 00:36:11
    regressive performance in games if you
  • 00:36:13
    look at some of these is hard to deal
  • 00:36:15
    with and it does kind of eat some of
  • 00:36:16
    that efficiency gain away production
  • 00:36:19
    performance is also mostly improved
  • 00:36:21
    considering the lower thread count so
  • 00:36:22
    that actually is great and some of the
  • 00:36:24
    performance gains for the thread
  • 00:36:25
    reduction is is impressive here like in
  • 00:36:27
    lender the problem is that at $630 there
  • 00:36:30
    are very few cases where the 285k makes
  • 00:36:32
    sense right now the conclusions are
  • 00:36:34
    going to depend on the types of
  • 00:36:35
    reviewers you watch we focus on the
  • 00:36:37
    value that means the money is always a
  • 00:36:40
    factor in the equation for the most part
  • 00:36:41
    of whether a product is worth it or not
  • 00:36:44
    and we can't get away from the price on
  • 00:36:46
    the
  • 00:36:46
    285k uh there really are not many good
  • 00:36:50
    cases you could make for it in
  • 00:36:52
    competition with the higher-end gaming
  • 00:36:55
    parts from AMD or even Intel's prior
  • 00:36:58
    generation Parts again even considering
  • 00:37:01
    all the efficiency stuff the the cost
  • 00:37:04
    argument is just not there for the ultra
  • 00:37:06
    200 series right now so uh that means
  • 00:37:08
    for us we need to see a strong value
  • 00:37:10
    argument to recommend aart we do not see
  • 00:37:12
    that currently with the 285k we'll look
  • 00:37:14
    at the 245k next and see if that's any
  • 00:37:15
    different it is easier to hit higher
  • 00:37:17
    memory clocks than before so that was
  • 00:37:18
    kind of cool we might do some additional
  • 00:37:19
    memory tuning on that and the benefits
  • 00:37:22
    in tes like seven zip show up in those
  • 00:37:24
    scenarios of course you can tune
  • 00:37:26
    everything else too but we just just
  • 00:37:27
    think they are better CPUs for the price
  • 00:37:29
    right now make it really simple for
  • 00:37:30
    everyone for gaming the 7800 XP makes
  • 00:37:33
    the most sense now when it's $600 or
  • 00:37:34
    it's stupid prices because you're
  • 00:37:36
    getting scalped don't buy it just wait
  • 00:37:38
    especially with the 9800 x2d around the
  • 00:37:40
    corner we don't know what that'll be
  • 00:37:41
    priced yet we'll see uh rumors are all
  • 00:37:43
    over the place for it but it just x3d
  • 00:37:45
    makes more sense basically if you're on
  • 00:37:46
    am4 the 5700 x3d is great if you're on
  • 00:37:49
    say 3,000 or 2,000 or something uh you
  • 00:37:51
    save a lot of money that way you don't
  • 00:37:52
    get all the fun of building a new
  • 00:37:53
    machine but you save a lot and then for
  • 00:37:56
    Intel competition the prior generation
  • 00:37:58
    part uh if all the issues have fully
  • 00:38:02
    been resolved and you know we tested the
  • 00:38:03
    micro code it does look better it's just
  • 00:38:05
    that we're not really going to know for
  • 00:38:07
    a while uh because that's a that's a
  • 00:38:10
    delayed issue but anyway those parts are
  • 00:38:12
    a lot cheaper now too so even those kind
  • 00:38:14
    of make more sense in most scenarios
  • 00:38:16
    there are places that
  • 00:38:17
    285k is
  • 00:38:19
    impressive but they are really not
  • 00:38:22
    enough to justify buying it so that's
  • 00:38:23
    going to be it for our review of this
  • 00:38:24
    one thanks for watching the power
  • 00:38:26
    testing stuff was really fun and made
  • 00:38:28
    the process interesting for us and then
  • 00:38:29
    we've got 245k and 265k coming up and
  • 00:38:32
    subscribe for more go to store. Gamers
  • 00:38:34
    access.net if you want to support our
  • 00:38:36
    in-depth testing we're basically
  • 00:38:38
    majority funded by the audience at this
  • 00:38:39
    point which is great because it reduces
  • 00:38:41
    our Reliance on ads as well and you can
  • 00:38:43
    head over to the store to help us out or
  • 00:38:44
    patreon.com gam Nexus to throw a few
  • 00:38:47
    bucks a month our way subscribe for more
  • 00:38:49
    we'll see you all next time
Tags
  • Intel
  • Core Ultra 9
  • 285K
  • CPU Review
  • Gaming Performance
  • Power Consumption
  • Efficiency
  • AMD
  • Performance Comparison
  • Tech Review