Comments Locked

102 Comments

Back to Article

  • NextGen_Gamer - Friday, March 25, 2022 - link

    I would love for AnandTech to do a full Intel/AMD comparison, with not only the newly released Core i9 12900KS and Ryzen 7 5800X3D CPUs, but the (majority) of the rest of each product stack as well. There has been so many updates to both Windows 11 to improve performance on both sides, as well as numerous driver updates in particular for AMD to improve Win11 performance. It would be awesome to see how they compare now a fully up-to-date Win11 OS, drivers, BIOS. etc.
  • BushLin - Friday, March 25, 2022 - link

    ... With realistic RAM speed and timings!
  • DannyH246 - Friday, March 25, 2022 - link

    As long as AMD are ahead you can forget about seeing such a thing here. Rest assured though, the moment Intel are ahead there will be an Anandtech Mega Test of all demonstrating Intel's 'Leadership'.
  • WaltC - Friday, March 25, 2022 - link

    Yes, that does seem to be the case here, for some reason. AT really needs to broaden its experience portfolio, imo. They often seem stuck in that old mantra that "You're safe with Intel," etc...;)
  • Unashamed_unoriginal_username_x86 - Friday, March 25, 2022 - link

    You people might not have noticed, but A. The last relevant AMD launch was late 2020,
    B. Alder Lake was collocated with the launch of Windows 11 and DDR5 so they had extra coverage on that, and
    C. Ian Cutress has left, so Ryan and Gavin (part time IIRC) are left trying to cover everything (which they can't really)
  • at_clucks - Saturday, March 26, 2022 - link

    "C. Ian Cutress has left"

    Which actually works for the best as far as fairness is concerned since Ian was so absolutely terrified of making Intel look bad and ruining those "industry connections" he loved to brag about that he literally never even tried to put Intel in a bad light.

    His reviews were thus somewhat useless because I always had to wonder if Intel is truly good or if Ian just dismissed the "outlier" data points that could have made Intel look worse. Eventually I'd go verify against the reviews from outlets and in the past years 8 out of 10 times it was more the latter. And if I have to check with other outlets I might as well go straight to them.

    This kind of weasel behavior just poisons everything. He used the AT name and the doors it opened for his own gain and paid back in biased articles that didn't help readers and pushed them away from AT. Now I'd rather have one solid gold massive review per year then a string of fluff pieces. So fingers crossed whoever is left is willing to go ahead with quality and hopefully the leadership won't stand for another round of "cutressisms" or else pretty soon there will be nobody around to even point fingers at AT.
  • at_clucks - Saturday, March 26, 2022 - link

    And to back it up:

    Ian's article in which he was originally ecstatically announcing the 5GHz CPU despite the screenshots in his own article (https://images.anandtech.com/doci/12893/image591.j... saying 2.7GHz base frequency: https://www.anandtech.com/show/12893/intels-28core...

    Ian's article when when he "got a sneak peek" at the Intel setup and discovered the CPU actually was overclocked (gasp! every other outlet was clear to point this out from the first time) and that the secret sauce was a 1HP chiller (something else every other outlet has flagged as suspicious originally). Not a hint of outrage at being taken for a fool: https://www.anandtech.com/show/12907/we-got-a-snea...

    Ian's article when Intel finally came clean about the trickery, with a tame title like "some details are confirmed", Ian writes a long apology letter in Intel's name with his "outrage" being limited to saying the "announcement was not ideally communicated": https://www.anandtech.com/show/12932/intel-confirm...

    Ian, there's no amount of backsplanations that you can give now that your actions haven't made painfully clear in the past. You have bent over backwards to keep Intel happy to the dismay of your readers. Good luck with your industry connections, you worked far harder for those than for AT and your readers.
  • Nexing - Saturday, March 26, 2022 - link

    I am amazed at_clucks.
    Without getting into the issue, the heavy data based way your arguments are displayed, plus the care and tight focus on your use of adjectives, that are far from contentious in a sensitive matter that could become so easily divisive just by the way it is expressed... amazing indeed.

    This seems to be part of a trend about the way a few advanced channels that I follow tend to communicate. A tendency I've been expecting for decades and only recently I'd happen to find.
    Anandtech would very much benefit for such approach and execution, if someone from AT is attentive, do please take note.
    Are you available as tech writer? (do not expect your answer here, just trying to arch over). Do you write in some other subject, channel? Please share it here if you do.

    Now, getting at the actual conversation, if your take on this matter (over my head) is correct, this explains the rather AT linearity felt along the latter years that INTEL wasn't able to keep up tik tok and more so with their inability to bring 10nm timely into market.
    From my far view it has been rather obvious that every quarter of INTEL repeatedly posting positive news (over 10Kmillions of revenue) at shareholders meetings, sharply contrasted with the horrible news coming concurrently from their manufacturing side.

    Shareholders applause and the rather lack of critical tech voices (appart from Semiaccurate) could not silence the fact that their financial success has been mostly demographic and halo based (the later being what they are probably attempting to sustain with this release). Just passively watching Samsung and TSMC technological advances has been enough to show the underlying disparity at play.
  • at_clucks - Sunday, March 27, 2022 - link

    Look, don't focus on my adjectives as they're not meant to make my message impartial or anything. They mirror my feelings for the situation, the frustration that a once great tech news outlet was reduced to being a "we're here too" mouthpiece, the relative uselessness of articles even when they contain a lot of technically accurate info but presented in a severely biased way, especially since this was flagged time and again on AT by many readers. I'm frustrated with the writer who made it happen for personal gain, and with the leadership who allowed it for the short term gain of not losing an editor in exchange for the long term loss of credibility among the people who matter: readers. I was banned (by Ryan if I remember correctly) for flagging this in a language no harsher than what you see here today.

    Ian's articles show objectively that he has taken a *very* soft stance on anything bad from Intel, up to a ridiculous level where he just accepted being lied to, and tricked into lying to his readers... and the reaction wasn't to play down what he did ("they tricked me too, sorry" which would have been reasonably understandable) but to play down what Intel did to everybody which is outrageous and demeaning for him as a person and as a professional. This supports my (very) subjective opinion that he is a weasel and he abused his position of power - informing readers with the AT name behind him - for personal gain.

    I won't pretend I'd be a good tech writer these days, I editorialize my writing too much to do a good job. I was a reasonably above average one 2 decades ago, maybe not even as good as Ian could have been if he had a spine and some honor. But I always put the value to the reader first, so the content and presentation were solid and the integrity unquestionable. I'm proud that some of my articles were the gold standard in OC communities on multiple continents for some years and that the people who taught Ian to OC probably learned a thing or two from me.

    This being said, while I will still visit AT once in a while I will certainly not recommend an article to anyone else. The fact that AT kept Ian in place for so long but let Andrei Frumusanu slip through their fingers was the last nail in the coffin. Andrei seemed to be the exact opposite of Ian when it came to making corporate enemies, repeatedly outing them for shenanigans. The whole thing just... rotted on the inside and in my book it's next to impossible to fix that.

    Trust and credibility takes eons to build and Ians to lose.
  • mode_13h - Monday, March 28, 2022 - link

    > Trust and credibility takes eons to build and Ians to lose.

    Heh.

    For my $0.02, I was critical of the way Ian pushed Intel to rebrand their manufacturing node. Who knows if his urging really had any effect, but I remember him taking some credit for at least raising the suggestion with folks at Intel. I felt that was definitely crossing a journalistic line.

    In the grand scheme of things, I do feel like Ian deserves more praise than criticism. On balance, the site is definitely worse for losing him.

    And I'd hope we can discuss the specifics of critiques without impugning the motives of the critic. I guess that's naive, but it's not out of reach, for most of us.
  • bji - Tuesday, March 29, 2022 - link

    at_clucks and Nexing read like the same person trying to confirm his own comments by writing in a fake response.
  • BushLin - Wednesday, March 30, 2022 - link

    You could be right, hard to tell on internet threads if you're not a site admin.
    My suspicions are aroused by the posts defending Ian without any balance, all appearing around the same time, several days after the article appeared but could just be coincidence.
    For what it's worth, I found the information in Ian's articles to be accurate and enjoyed discovering more about the way things actually operate but frequently questioned what wasn't said, tested or compared given the content was often lengthy yet what seemed like obvious points were missed... I can see why there are claims of bias but it's not like the content was of no value.
  • damianrobertjones - Monday, March 28, 2022 - link

    "backsplanations "

    Epic word is epic.
  • AshlayW - Tuesday, March 29, 2022 - link

    I don't think Ian is Intel bias and I am a pretty big AMD fan (or at least, was, pre Zen3). He comes across as a professional and has many articles showing AMD's leadership in the last few years, he just doesn't sugar coat AMD and kiss their boots like some people expect
  • Hulk - Sunday, March 27, 2022 - link

    You are of course entitled to your opinion but mine is diametrically opposed to your. I valued Ian's work at Anandtech and believe it to be informative, fair, and a good read.
  • at_clucks - Sunday, March 27, 2022 - link

    @Hulk, that's fine, of course. I sometimes value an IKEA hot dog even despite knowing it's the lowest grade meat product produced by the lowest bidder. As I said, Ian's articles are not necessarily inaccurate just heavily biased which is still bad. If you tend to lean with that bias though everything is peachy.

    But here's the thing, Ian's partiality towards Intel is indisputable. It's been shown above at the very least with his string of articles covering (pun intended) Intel's shenanigans to a degree that's disgusting. And it's been confirmed on this very page of comments when Ian himself insisted "AT engineers like JEDEC" and can't be bothered to change one setting that almost any user would go for to have their CPU run faster. But just recently he had no problem enabling an officially disabled/unsupported feature on an Intel CPU and benchmarking with it to show big bars.

    So I guess the point is, how does your claim of "fairness" reconcile with the stream of examples like I gave above? Is it that you consider that to still somehow fit the definition of "fair", or that you agree they're biased but "it's just those examples and no more" (something I'd expect you to have to repeat with every new example)?
  • DominionSeraph - Sunday, March 27, 2022 - link

    WE HAVE FOUND THE /g/ 24/7 ANTI-INTEL WOJAK POSTER!!!
    and his name is at_chucks

    Post your copypasta! Please! I miss it! You can't post it on 4chan because it will get you banned, but you can go ahead and post it here!
  • at_clucks - Monday, March 28, 2022 - link

    @DominionSeraph, I thought I made it pretty clear my qualm is with Ian's writing style and bias rather than Intel's products themselves. Reading comprehension be damned. Not that your writing was any better... :/
  • OreoCookie - Tuesday, March 29, 2022 - link

    I'm surprised by how personal some people take the Intel vs. AMD debate and Ian Cutress leaving. I haven't seen any evidence of favoritism nor does it make much sense for him to be “nice” to companies when reviewing products if he wants to become a technical consultant. All recent Intel reviews go into depth how much energy and power Intel CPUs have to spend to match or slightly surpass competitors in terms of performance — that's hardly a glowing recommendation. If you go beyond Intel and AMD, their coverage of e. g. Apple SoCs clearly points out to how and how much superior they are, sometimes in absolute terms, sometimes on a performance-per-watt basis.

    The only thing that seems apparent is the thinning out of talent. AT hasn't found anyone to take over Andrei's and Ian's duties. And even before them, I remember that Anand (the OG and namesake) did additional tests like touch screen responsiveness tests on iPhones that I was hoping would continue. I don't know who will replace them, but we shall see.
  • felixbrault - Tuesday, March 29, 2022 - link

    I had to login for this. How dare to you talk about Ian like this?!? What the hell is wrong with you. If there’s one thing Anandtech is not and never have been is biased. The simple reason why Intel wasn’t put under bad light since the Core2Duo era is simply because they had the best cpu in the industries (single thread performance which in 90% of use cast is what matter) except for 2 years before Alder Lake. This guy had opportunities handed to him because he worked his ass off for at least 10 years here and he’s one of the best tech writer in the industries.
    I’ve been coming to Anandtech.com every day for the last 20 years and I’ve never seen such comments from someone. You should be ashame of yourself.
  • mode_13h - Wednesday, March 30, 2022 - link

    > How dare to you talk about Ian like this?!?

    No one should be above reproach. Ian is a big boy. He doesn't need you to protect him.

    > I’ve never seen such comments from someone.

    Apparently you don't read all the news comments, though? I'm guessing not, because you sound as if you don't ordinarily login to the news comments.

    > You should be ashame of yourself.

    LOL. Everyone is as entitled to their opinions as you are. The best way to counter a claim you disagree with is to cite counter-evidence. Making it personal just antagonizes the recipient and the exchange will likely spiral downward, quickly thereafter.
  • Mike Bruzzone - Thursday, April 7, 2022 - link

    The system of mutually dependent access and allocation typically emphasizes and even exaggerates Intel on production volume value to sales distribution and media sales preview channels, on the shear weight of intel production volume and its financial trickle down potential.

    The system typically underplays and can ignore alternate x86 competitors that can include AMD primarily on AMD slimmer volumes and values it's advantageous to have a lot of options and value types to trade. AMD has more value types than they've ever had currently.

    Ultimately it's all about exchange. Whether Intel or alternate players that have also been restrained and limited, by Intel associate channels primarily, on their limited values to trade.

    Piss off Intel, or piss off AMD, and you lose your access and allocation. It works both ways.

    mb
  • biigD - Sunday, March 27, 2022 - link

    It's pretty clear to me that Ian let his enjoyment of feeling 'close' to Intel influence his reviews/interviews. But c'mon, we're not a bunch of rubes. I'm here because I enjoy the deep dives into hardware, and I'm more than capable of filtering the BS from the technical facts and benchmarks. Ian wrote huge, detailed reviews and provided interesting interviews. Did he take it easy on Intel or inject less than objective language from time to time? Yep. Did any of that have any bearing on how I chose to proceed with my next build or how well it might fit my workflow? Not one bit.
  • at_clucks - Sunday, March 27, 2022 - link

    @biigD, your explanation has some merit but let's be real, "the articles are for experts who can discern bias from fact" is one of the most feeble attempts to cover lack of journalistic integrity. Moreover, that "experts would have figured out the bias" would be some very shameful victim blaming, putting it on the "rubes" who foolishly trusted an AT tech writer instead of being experts in the first place.

    The fact of the matter is that Ian applied this bias mainly towards Intel so it's not just his writing stile "to be filtered out", and there's no "only for experts" tag on any of the articles. Ad they are actually open to the whole world which means countless people were mislead by them.
  • biigD - Sunday, March 27, 2022 - link

    Fair enough. Remember that I'm not looking at this through the lens of a journalist, so while I don't condone Ian's bias, it doesn't generate such a visceral response either.
  • WaltC - Monday, March 28, 2022 - link

    Looks like Ian Cutress is sharing the byline for the article. Is that not true, you're saying?
  • goatfajitas - Monday, April 11, 2022 - link

    Yup, Anandtech really stopped being that great unbiased source for tech reviews before Anand even left.
  • Unashamed_unoriginal_username_x86 - Friday, March 25, 2022 - link

    Anandtech is biased towards Intel? Prove it, shill
  • Oxford Guy - Friday, March 25, 2022 - link

    Testing Zen 1 and Zen 2 with JEDEC RAM.

    Only JEDEC.
  • Ian Cutress - Friday, March 25, 2022 - link

    Every CPU tested at JEDEC, Intel or AMD. What's your point? AT has engineers that love the JEDEC data.
  • Alexvrb - Friday, March 25, 2022 - link

    Showing performance at both default and "typical gamer/enthusiast XMP speeds" is always nice, especially for an enthusiast site. With that being said, the larger performance deficit at low RAM speeds was AMD's fault, and I don't think testing at JEDEC shows deliberate bias.
  • at_clucks - Sunday, March 27, 2022 - link

    It's true that it's AMD's fault for poor performance in default settings. But a reviewer worth their salt should very clearly point out that it can be mitigated with relative ease. Will it take more effort? Yes. But it will also paint a complete picture. Half a picture is worse than no picture because it leaves you assuming that's the whole truth. It's bias and it's what Ian has been doing for years, ever since he started rubbing elbows with some important people and his integrity dissolved.

    And let's not forget that Ian had no quarrels writing articles about the performance of overclocked Intel CPUs. He reviews CPUs using premium water cooling systems, AT is peppered with articles where the voltage is hiked in the red, OC limit is tested, far more advanced BIOS settings are tweaked on motherboards chock full of OC options but changing memory timings via XMP profiles to show that some bad results can me successfully mitigated is too much because AT engineers love JEDEC? That's a BS excuse from someone trying to cover for their massive failures as a professional reviewer.

    Ian spent his final years at AT polishing his relationships in the industry with a neverending string of interviews, and towing Intel's line wherever possible.

    His AMD reviews are... maybe objectively accurate but bland and monotonous. No matter how many times an AMD CPU tops the charts there are no superlatives, just an enumeration "the AMD processors have been doing well" or "the 5700G is +28% over the previous generation [...] we're seeing good yearly improvements" (yes, a ~30% improvement is that bland when it's AMD).

    Intel reviews are sprinkled with amazement, superlatives, specific mentions where the Intel CPU does better than the AMD, "improvements are extremely massive, and represent a major jump in performance, something which undoubtedly lead to larger IPC gains" and "showcases also some very large gains in some of the workloads, +33%" (yes, a ~30% improvement in some esoteric workloads is *that* exciting when it's Intel).

    Keep in mind, these may all be objectively true but if you're making it far more obvious when Intel is better than AMD then the other way around, if you're letting your personal excitement show more for one than for the other, then that's literally the definition of bias. He repeatedly and with intention painted a biased picture with his articles and reviews and this can't be hand-waved away with "we like standard settings".

    There are many ways of deceiving and some don't actually use bad data, just bad framing. A biased review is a worthless review. Ian is the Donald Trump of tech writing (with every possible negative connotation).
  • at_clucks - Sunday, March 27, 2022 - link

    That's strange to say Ian after repeatedly acting in the exact opposite way if it favored Intel.

    Why did you not use some sort of standard box cooler for your i9-12900 review? Why did you bother to "dig into the story" to enable AVX-512 support which was officially disabled (officially fused off) by Intel, and why did you even run benchmarks with it?

    If you can spare the effort to benchmark such an officially unsupported setup for Intel why was using XPM profiles a step too far for AMD? Just so you can conclude that: "Overall though, it’s no denying that Intel is now in the thick of it, ***or if I were to argue***, the market leader."?

    You took your job at AT as an opportunity to argue Intel's case again and again, forgoing integrity. Even after slinking away you still feel compelled to come back with some feeble attempts to justify your failures. You are an embarrassment to AT and journalism in general. Careful not to break that hand while handwaving off the mounting arguments of your bias.
  • BushLin - Sunday, March 27, 2022 - link

    Why is there love for JEDEC data at AT or anywhere? Years after the release of a memory standard, even lame Dell/HP/Lenovo spreadsheet fodder PCs aren't using JEDEC spec as it's not saving any money to have lower spec RAM. Testing at JEDEC is a waste of testing time as the results are useless.
  • at_clucks - Sunday, March 27, 2022 - link

    AT doesn't love JEDEC but it makes it easy to claim "standardization". Although the only value is less effort for the writer while actually diminishing the value of the write-up. But the real value for some of the writers is that they can control the conclusion much easier and introduce bias by design.

    Using JEDEC timings was a hit for AMD but it's fair because it's standardized, which every reader should be told about. But every reader should also be told that XMP profiles mitigate that. This "standardization" of course didn't apply when Ian did bother to dig into a motherboard's BIOS to enable the AVX-512 officially disabled ("fused off") by Intel, and then benchmark this thoroughly non-standard setup to show some massive bars for Intel CPUs.

    In reality what Ian loved doing was to mislead the readers not by presenting outright false data but by cherry picking data and adjusting the way it was expressed, by his "tone", his choice of superlatives or banalities depending on the company being targeted. Google "lying with data", it's a form of art.
  • Flying Aardvark - Thursday, March 31, 2022 - link

    Yes, because stability is not guaranteed outside of JEDEC. All I care about as a reader is how it's warrantied and guaranteed to run. I guarantee if going outside of spec made AMD look bad, you wouldn't be such a fan. You just want to help them out. I had many Ryzen rigs and if we count stability with various RAM kits, Intel mops the floor with AMD.

    What your precious AMD deserved were JEDEC AND stability comparison articles. Thrashing your sub-rate platform.
  • 29a - Tuesday, March 29, 2022 - link

    Not finishing AMD reviews.
  • aagello24 - Sunday, April 10, 2022 - link

    funny you say that. cause tomshardware does the SAME thing!!
  • Makaveli - Friday, March 25, 2022 - link

    Yes no Jedec cas 22 nonsense. At atleast DDR4 3200 Cas 16 which is standard these days.
  • Oxford Guy - Friday, March 25, 2022 - link

    That was only partially a self-inflicted wound on AMD’s part. There is an entire sector of enthusiasts involved in using XMP profiles — which is why they’re standard for RAM sticks and motherboards.

    This site routinely used unsafe voltages and unstable results in its CPU overclocking articles and then moved to claiming that anything requiring a user to enter BIOS (i.e. switching on XMP) is so esoteric and demanding on terms of knowledge and skill that covering anything involving that isn’t worthwhile.

    Despite that, motherboard articles talk about overclocking favorably (more difficult than entering BIOS and switching on XMP) and the site just posted an article featuring Intel’s overclocking ambassador (without a word, as usual, about how Intel dropped its overclocking warrantee — unless I missed that bit which I very greatly doubt). I’ve not seen a warning in motherboard articles showing off fancy VRMs that, since things like entering BIOS are too esoteric, such VRMs are irrelevant (so long as one doesn’t buy a junk board that can’t handle its stock support list). I also didn’t see this site dismiss Intel’s overclocking promoter with the line about how no one can be expected to enter BIOS so overclocking is too complicated to be part of our coverage.

    The platform where using XMP was tricky was Zen 1. That’s basically it. My relative’s Zen 1 chip required a change from 3200 CL16 to 2933. After that switch to the other XMP profile, it worked and has for years. He likely could have tuned the settings to achieve 3200 CL16 but 2933 was a good-enough quick solution. That’s Zen 1 — by far the touchiest platform for XMP in affordable RAM. Even novices, though, gained significant performance easily via XMP due to the infinity fabric speed linkage. In short, even for a novice it was worth entering BIOS twice for XMP. I advised against trying to overclock the CPU with the stock cooler. Yet, CPU overclocking has long been a staple of this site.
  • Khanan - Sunday, March 27, 2022 - link

    Using Jedec Ram speeds is a artform, it’s hard to find these terrible ram sticks that use the lowest speeds and anyone who buys 3200 RAM should know how and want to run it on their proper speeds. There’s no excuse to use anything lower than 3200 these days other than if the processor simply can’t do it.
  • ikjadoon - Friday, March 25, 2022 - link

    Would absolutely love this. Alder Lake has had a few microcode + BIOS updates, as well.

    However, I do think waiting on the next Spectre patch against Intel mostly, but also AMD would be helpful. I don't think Microsoft has released it, though the Linux kernel has.
  • Silver5urfer - Saturday, March 26, 2022 - link

    Why would anyone want to run that Windows 11 disaster ? It has horrendous downgrades from Explorer.exe itself to Shell32 of core Windows OS. And they made stupid changes like VBS on the CPU side of things. Plus more extra nonsense for AMD parts. Moreover that OS is designed like a Tablet OS rather than a PC centric one.

    But yeah the name checks out lol, 12900K for gaming is absolutely stupid. A 12700K with PL1 and PL2 unlocked with a modest OS would get every single game and a 3090 / 6900XT run around any game. This CPU is not for gaming and 5800XD is a garbage processor. Cannot OC, and priced just because of gaming. A 5900X is the best AM4 processor. For gaming and MT workloads both. With reduced price it's even no-brainer.
  • Oxford Guy - Saturday, March 26, 2022 - link

    'Why would anyone want to run that Windows 11 disaster?'

    Anyone wanting/needing to use Windows (unless air gapped) will have no choice. That's the way 'upgrades' work.
  • Kangal - Sunday, March 27, 2022 - link

    You're right, it's not ideal to think about those edge-case processors. When looking at the Mainstream PCs, I think a good way of looking at progression roughly is:
    2010 Core i7-920 (4c/8t), 8GB-1600, HD 6970
    2012 Core i7-2600k (4c/8t), 8GB-2133, GTX 680
    2015 Core i7-6700k (4c/8t), 16GB-3000, GTX 980
    2017 Core i7-8700k (6c/12t), 16GB-3000, GTX 1080
    2018 Core i9-9900k (8c/16t), 32GB-3600, GTX 1080Ti
    2020 AMD r9-3900x (12c/24t), 32GB-3600, RTX 2070-Super
    2022 AMD r9-5950x (16c/32t), 64GB-4000, RTX 3090Ti
  • Flying Aardvark - Thursday, March 31, 2022 - link

    Not everyone is worried about saving the dollars that you are. AMD Poverty Gaming may be for you, but others are running 3090s and will run 4090s after that without batting an eye.
  • Oxford Guy - Saturday, April 2, 2022 - link

    I agree. I was fortunate to be born with two extra kidneys so I was able to sell the surplus and pay a scalper — no, sorry, an OEM — for a GPU Nvidia wouldn’t sell directly to me for a fair price.
  • Karandar_ - Tuesday, April 5, 2022 - link

    +1 to this idea. A lot of work I know, but I would love to see how they all compete given mature drivers and the recent security mitigations. I think the current numbers would be different enough to offer readers great value.
  • mode_13h - Tuesday, April 5, 2022 - link

    Well, I think that's the idea behind:

    https://www.anandtech.com/Bench/

    ...except, they seem to be a couple years out of date.
  • Bluecobra - Friday, March 25, 2022 - link

    Buyer beware: this CPU will need some serious cooling/tinkering. I have i9-12900K and at default settings I immediately hit 100C in Cinebench despite pairing it with a brand new Corsair H150i 360mm liquid cooler (with the correct LGA 1700 hardware). It's pretty frustrating as I bought this CPU so I would have something fast so I didn't have to muck around with.

    That being said, Cinebench is an artificial workload and I don't do anything CPU intensive enough to cause it to hit those extreme temps.
  • ikjadoon - Friday, March 25, 2022 - link

    Intel is always at the worst part of the performance-power curve. The lines are essentially straight, even straight from Intel themselves:

    https://cdn.mos.cms.futurecdn.net/yNDvYxQzhRKyKbfp...

    Intel hasn't focused on a wide, low-power architecture and, boy, does it show. CPUs were never really meant to boost past 4 GHz; everything else is just mountains of wattage for a minor performance bump.
  • bwj - Friday, March 25, 2022 - link

    It sounds to me like you should check the way your cooler is mounted. I'm using a Noctua air cooler and it takes creative thinking for me to get core temps above 80C, and it takes about a full minute to get them there.
  • Alistair - Friday, March 25, 2022 - link

    I thought Corsair got in trouble as their coolers don't actually mount right with the LGA 1700 adapter. You need a different cooler. I don't hit 100 degrees with a Noctua U12A for example.
  • Kvaern1 - Sunday, March 27, 2022 - link

    I think your issue may be that the default settings are NOT Intel's defaults but the motherboard manufactures crazily overpowered and slightly overclocked defaults.

    Case in point the default settings on my Asus board with a 12700KF/Noctua NH-D15 will easily make it hit 100c @ 4700Mhz (Default Asus OC) in Handbrake. Using my own powerprofile it runs 20c cooler @ 4900Mhz.
  • Hulk - Friday, March 25, 2022 - link

    Same base clocks and higher TDP. Interesting. I guess the higher frequency part has greater leakage.
  • PeachNCream - Friday, March 25, 2022 - link

    Typo alert - "The major PC parts retailer listed the uannounced Intel chip for sale..."

    Should be "unannounced"?
  • Ryan Smith - Friday, March 25, 2022 - link

    Thanks!
  • ikjadoon - Friday, March 25, 2022 - link

    >Neither Intel nor AMD have access to each other’s chips right now, so a direct comparison using both sets of data is likely to be inconclusive right now.

    IMO, AMD could probably just overclock a nicely-binned i9-12900K.

    I can't quite tell for even the released ADL CPUs: what is the all-P-core boost? Is that also 5.5 GHz here? No, right?
  • DannyH246 - Friday, March 25, 2022 - link

    Todays major news article - Intel's newest CPU BRIEFLY listed on NewEgg.
    Tomorrows major news article - Intel CEO breaks wind.
  • at_clucks - Monday, March 28, 2022 - link

    It gave away the top of the line Intel CPU and specs. I think we can get past how it happened (Newegg listing).
  • bwj - Friday, March 25, 2022 - link

    It would take a miracle for this thing to actually hit that speed, right? You'd need to have only 1 core active and the CPU stone cold.

    My 12700K rarely records hitting 5GHz even when I hit it with a sudden single-threaded workload.
  • Bluecobra - Friday, March 25, 2022 - link

    I think that Task Manager is inaccurate when it comes to these CPUs, have you tried HWMonitor? I have an i9-12900K running at stock settings and I see cores #4 and #6 jumping to 5.2Ghz all the time with normal use in Windows 10.
  • Alistair - Friday, March 25, 2022 - link

    I have the same chip. It is called a manual OC. The only thing that would actually matter is if you can run it at 5.4ghz/5.7ghz easily.
  • Oxford Guy - Friday, March 25, 2022 - link

    Intel no longer offers an overclocking warranty.

    (We’ll ignore that the company has the nerve, though, to trot out one of its staff as being the ambassador of Intel overclocking.)
  • coburn_c - Friday, March 25, 2022 - link

    Wow! Such Amazing! Get the 12700k
  • Oxford Guy - Friday, March 25, 2022 - link

    ‘Given the extreme clockspeeds, this is going to be a ‘thin-bin’ part, which means that Intel is going to need to do extra binning to bring these processors to market in sufficient quantities with the characteristics determined by the bin.’

    The overclocker known as ‘The Stilt’ claimed that AMD’s 9590 CPU was the opposite of what people commonly believe binning means. He said parts needing less voltage aren’t necessarily better. Instead, they are higher leakage which means less power-efficient (less voltage but higher current). They hit the thermal limit faster than lower-leakage chips needing higher voltage for the same clock.

    Most consumers believe that a chip that requires less voltage is a better-quality more-efficient bin but he said it’s the opposite — that AMD had to downgrade the AM3+ specification to accommodate the desperate exploitative (my words) 9000-series release. He said AMD would have had to send the chips to the crusher.

    He also said that with conventional cooling (air and non-chilled water), a lower-leakage rather than lower-voltage part is better for performance. When trying to break records via nitrogen, then higher-leakage chips are preferred. The exception to the general rule is the very low-leakage design — where the very low leakage impairs its ability to reach high clocks even with conventional cooling.

    What he claimed is the complete opposite of everything I have seen written about in consumer/enthusiast journalism. Is he correct?

    If so, this part may be a bin that’s worse than the others in terms of efficiency. The higher leakage, though, can help (with lots of cooling) guarantee frequencies at stock beyond the typical range.

    This site’s coverage of Fermi (GTX 480) said Nvidia (claimed) higher-leakage transistors were used in parts of the design to increase performance at the cost of higher temps/power. That appears to the exactly the logic ‘The Stilt’ presented. So, again... why does everyone (outside of that Fermi coverage) equate lower-voltage chip bins with higher clocks (and efficiency/quality)?

    If both pieces of logic are correct, it’s more a matter of chasing inefficiency to obtain diminished returns (where nitrogen cooling is the most obvious end game).
  • Sunrise089 - Friday, March 25, 2022 - link

    Interesting comment. This isn’t my wheelhouse to reply with any authority, but seems like an interesting question to have raised.
  • The Von Matrices - Tuesday, March 29, 2022 - link

    It would all depend on the limiting factor in the chip's design. If clock speed cannot be raised regardless of the voltage, then leakier transistors would make sense. But if the limit is power, with the chip (and VRM) overheating before voltage becomes the issue, then you would want less leaky transistors to achieve a higher clock speed.
  • croc - Friday, March 25, 2022 - link

    Please wake me when Intel releases an HEDT part with at least 36, preferably 40, pcie lanes from the CPU. The W parts are getting old, rare, and thus higher-than-kite priced. Not to mention low clocks speeds. AND for the AMD fanbois... No new threadripper part for almost two years now. Pro parts don't count for me, as they are currently unavailable to anyone not Lenovo. And, if they are ever released to the public at large, they will be as rare as hens' teeth. AMD is still not performing at the 7mm node well enough to service its EPYC backlog, and these parts come from those wafers.

    HEDT segment is getting the shit end of the stick these days.
  • Khanan - Sunday, March 27, 2022 - link

    AMD isn’t building processors, TSMC is. So if you wanna blame anyone for not enough options that would be TSMC as they have a limited amount of production or wafers. And AMD chose to build the more expensive Pro line with those. That said, Threadripper Pro will probably leak to the market in a few months or be officially buyable, so with the right money anyone can buy it for a self made WS. Upside is, those have octa channel and more lanes.
  • mode_13h - Monday, March 28, 2022 - link

    > The W parts are getting old, rare, and thus higher-than-kite priced.

    Are you talking about the Cascade Lake parts, or did you know they released an Ice Lake refresh?

    https://www.anandtech.com/show/16822/intel-launche...

    Of course, being Ice Lake means it won't have great single-thread performance.

    > And, if they are ever released to the public at large, they will be as rare as hens' teeth.

    If you look at how TR Pro 3k was handled, Lenovo only had exclusivity for a few months. Then, they were released to the public.

    > HEDT segment is getting the shit end of the stick these days.

    Wait until Sapphire Rapids' refresh of Xeon W. That should also spur on AMD to give TR more love, due to pressure by their OEMs, if nothing else.
  • sonicmerlin - Saturday, March 26, 2022 - link

    What is a P core and E core? Is that power and efficiency? Is that official jargon these days for desktop CPUs?
  • sonicmerlin - Saturday, March 26, 2022 - link

    Oh never mind I see it on the Intel ad. Too bad I can’t delete old comments.
  • Khanan - Sunday, March 27, 2022 - link

    It’s what a company does if they are unable to fit 16 full performance cores on a processor without it sucking 500W of power and being too big to produce. It’s funny how AMD doesn’t need it because their cores simply scale perfectly from low to high. Big.little isn’t the way to go for a PC desktop processor unless you simply can’t do it the proper way. It’s not bad for mobile but that’s it.
  • mode_13h - Monday, March 28, 2022 - link

    > It’s what a company does if they are unable to fit 16 full performance cores
    > on a processor without it sucking 500W of power and being too big to produce.

    That's a different spin on it, but essentially yes. It's a more area-efficient and power-efficient way to scale performance for highly-threaded workloads. And it obviously has downsides and obvious pitfalls. That's not to say it's a bad idea, but it's also not a pure win.
  • TheinsanegamerN - Monday, March 28, 2022 - link

    > It's a more area-efficient and power-efficient way to scale performance for highly-threaded workloads
    Now THAT's a new spin on it. And based on intel's power numbers, "power efficient" is still wrong, intel cant even begin to touch AMD's perf/watt numbers for zen 3. zen 4 is gonna destroy them.
  • mode_13h - Monday, March 28, 2022 - link

    > Now THAT's a new spin on it.

    No, it's actually not. It's what many have been saying since before the launch of Alder Lake.

    > And based on intel's power numbers, "power efficient" is still wrong,

    It's not wrong. The added E-cores deliver more performance than if they'd used P-cores within the same power envelope or area. The data is quite clear from the original i9-12900K review. Go have another look.

    > power efficient" is still wrong, intel cant even begin to touch AMD's perf/watt numbers

    Note that I didn't say Alder Lake as a "power efficient" CPU. Just that adding E-cores was more efficient than adding yet more P-cores.
  • Khanan - Monday, March 28, 2022 - link

    Adding more P cores wasn’t even a option. As I already explained, it would’ve made the CPU both too big and too inefficient to produce. So their only other option would’ve been to go with 10 cores as that is exactly what you can do instead of using 8 E cores, as 2 Extra P cores use up the same room and suck about as much power as 8 E cores probably would do. Intel then, back to 10 cores it would’ve been, and only progress on the IPC and nothing else. With these E cores they can at least claim to have the same amount of cores, standard desktop it is. But everyone can see that it only has 24 threads. Will be funny with Zen 4, when its 48 threads vs just 32 at Intel, this isn’t a battle Intel can win, not even with E cores.

    2008 called and wants their monolithic trash back.
  • mode_13h - Wednesday, March 30, 2022 - link

    > Adding more P cores wasn’t even a option.

    It sort of was, as you go on to explain.

    > So their only other option would’ve been to go with 10 cores

    Yes. And in terms of perf/W, I think it's clear their 8 + 8 configuration is better than 10 + 0 would've been. That's my take on why they did it - a cheap, if not problem-free, way to improve performance (both per-$ and per-W) of highly-threaded workloads.

    I think we can agree that it's an interesting experiment. I don't mind seeing Intel try this. There are surely things to be learned, and it will certainly motivate improvements in Big.Little support, among popular x86 operating systems.

    > 2008 called and wants their monolithic trash back.

    Not sure what you mean by this.
  • The Von Matrices - Tuesday, March 29, 2022 - link

    Thank you. I don't understand why this is not stated more often; the reviews call them "efficiency" cores but that efficiency is in die area more than power. There are no workloads (excluding those that are artificially limited by software licensing) that can make full use of 8 cores but also won't scale to an unlimited amount of threads. So you might as well have a few high performance cores for the single-threaded workloads and then cram as many area-efficient cores as possible in the remaining die area to speed up the multithreaded workloads as much as possible.
  • mode_13h - Wednesday, March 30, 2022 - link

    Also, it's not as if operating systems are completely naive to scheduling on hybrid CPUs, if you count Hyper-Threading (aka SMT) as such. If modern OS schedulers can already manage to load up one thread/core before doubling-up, then I don't really see why it's so hard to insert a step in the middle (as Intel suggests).

    I've heard some concerns that Big.Little designs will unfairly victimize workloads running on the Little cores, but this is similar to problems that OS schedulers are already designed to solve. Threads within the same priority group should get roughly equal execution time, and a good scheduler merely has to weight that execution time according to the power of the core, to ensure fairness. Worst case, that might entail moving a job from a Little core to give it some time on a Big core, but the cost of such migrations are insignificant if they're done infrequently.
  • GreenReaper - Saturday, March 26, 2022 - link

    The product might not be a reaction, but the timing of its release might be. If Anandtech can prepare a story, Intel can prepare a release. At the same time, not many who could prove it either way.
  • Khanan - Sunday, March 27, 2022 - link

    The fastest desktop professor? Not really. Despite auto-overclocking itself to 250W it’s still slower than the 5950X. And then, imagine, the 5950X can be overclocked too! Suffice to say the 5950X easily destroys this half breed trash that doesn’t even have 32 threads and a mish mash of desktop and mobile processors. And that without using over 200W of power. Hopefully Zen 4 will destroy anything Intel release in 2022, we can’t have this corrupt company up there again, and inefficient hardware isn’t the way to go either.
  • TheWanginator - Sunday, March 27, 2022 - link

    Am I the only one who sees it as totally hilarious that after alll those years of AMD getting dinged for power efficiency, its now the competition going in the totally opposite direction. Just for the sake of the win
  • Khanan - Monday, March 28, 2022 - link

    AMD was always more power efficient than Intel if you forget FX for a second, though. We can forget FX as a misstep that won’t happen again.

    Since Intel is in the low tech role they’re just pushing the clocks and that’s inefficient. Their cores are too big and fat and monolithic is yesteryears news. Maybe their manufacturing is garbage too, who knows. A friend suggested that “intel 7” is to blame for the inefficiency, but somehow I doubt that. Maybe part of it at best.
  • blppt - Monday, March 28, 2022 - link

    Remember that Intel once made Prescott.
  • The Von Matrices - Tuesday, March 29, 2022 - link

    This comment has aged as well as "Can it run Crysis?". Much like any low-end GPU can run Crysis today, Prescott in its fastest model only had a 115W TDP - about as much power as today's mid-range CPUs at full load. It's only remembered as hot because of its competition's higher efficiency and the jump in power compared to its predecessor.
  • Khanan - Tuesday, March 29, 2022 - link

    What are you on about? Efficiency is relative, an 115W was a lot back then. If your CPUs aren’t any faster and still suck way more energy they are being seen as powerhogs. P4 in general was a mixed bag, I wouldn’t say it was a great architecture, it barely delivered sufficient performance with high costs and inefficiency. Has a reason why they abandoned it and went back to the original Pentium design.
  • blppt - Tuesday, March 29, 2022 - link

    Due to the much larger process node back in the Prescott days, that relative pittance of a TDP resulted in a blast furnace.
  • mode_13h - Wednesday, March 30, 2022 - link

    > that relative pittance of a TDP resulted in a blast furnace.

    Huh? TDP is in Watts. Now, leaving aside the wiggle-room in how it's actually defined, a chip putting out 115 W is generating the same heat no matter what process node it's made on.
  • Shmee - Sunday, March 27, 2022 - link

    I agree that performance testing should be done with better quality RAM, there is no point in using lower speeds when there is a capable XMP setting available. Also, reviews should include OC details on RAM and the scaling in performance. It is too often that reviewers don't cover tweaking well enough.

    That said, I do have a lot of respect for Ian and I know that AT really needs more quality reviewers.
  • Mike Bruzzone - Monday, March 28, 2022 - link

    Alder 12900KS volume, total production? We can look at Coffee Refresh, 9900KS total unit volume over 9th gen full run is 0.16% and 9900K is 41.03% and all of i9 9900_ is 42.95%. Pursuant Alder, 89% of channel available continues to be i9/i7 and all else is simply fall out from sort.

    Mike Bruzzone, Camp Marketing
  • ThirteenthDominion - Wednesday, March 30, 2022 - link

    9900KS total unit volume over 9th gen full run is 0.16% and 9900K is 41.03% and all of i9 9900_ is 42.95%

    That is interesting Mike, I didn't think the previous KS was such a low volume product. How many 9900KS and 9900_ units were produced?

    Also, there's the Xeon E-2288G that was siphoned from the same pile of chips. Do you know how many of those were made?
  • mode_13h - Wednesday, March 30, 2022 - link

    I came really close to buying a Xeon E-2288G. That was when I'd first come to accept that I'd been priced out of the true workstation (i.e. Xeon W and TR) market.
  • Mike Bruzzone - Wednesday, April 6, 2022 - link

    @ThirteenthDominion,

    Good question "how many 9900KS and 9900_ units were produced". I have the base data to determine the answer but that can take several days so I'll give you my quick take.

    Traditionally Intel desktop runs are 60 M to 90 M units annually. But at Rocket S it's down around 28M units, for the reason no one wanted Rocket as an Intel place holder and there was at the time plenty of Comet S surplus in the channel. Also moving into Alder Lake S made prior gens moot unless sticking with Comet stability at a clearance sale price on Windows 10.

    Right now, I believe Intel desktop S is produced in half runs; 30 M Rocket then + 30 M Alder = 1 run + 30 M Raptor = 1.5 runs (however the whole Alder/Raptor run) and where in this example Alder Ramp volume mirrors Raptor run down volume think isosceles triangle. Solely my thesis currently I've seen this Intel production behavior in the past associated with cost optimization.

    The way the volume answer is determined takes Intel volume statements associated with specific product categories, Gelsinger's second half 2021 Tiger mobile volume statements for example. Compares that stated 'Intel known' to an unknow on channel supply volume, to come up with a proximate unit volume. The exercise compares to ebay by generation supply data that are offers for sale as a proxy; testing through peak, through run down, over some set amount of time in weeks that can look at the ramp or through any annual period, basically, more than one time period comparison as check.

    So I've been comparing ebay offer data from the known Tiger 2021 mobile supply volume against the unknown Coffee Refresh volume in this example for four hours and realize this is a week's long puzzle to address.

    I can also turn to Intel supply signal cipher data 2005 through 2012, where Intel and associates gave quarterly 'by category' supply volume' in data bits, relying on a disintegrated story problem recomposed gave the volume answer and that's among the longer exercises. Again, compares a stated known product generation volume in this example on Intel supply signal against unknowns on channel supply data.

    What can I do here quickly?

    On Gelsinger Tiger mobile volume on channel data comparison what is 9th gen? Gelsinger said 100 M Tiger mobile sold in 2021 and gave that in three chirps at 50 M through q2, 70 M through q3, + 30 M or 100 M total through q4 and my result is 100 M Tiger mobile confirms 52 M Vermeer in 2021.

    After which the Tiger Mobile v Coffee Refresh presents an anomaly. The anomaly is that Coffee Refresh ebay offer data presents a huge volume much large than Comet and I need to go back and compare it with Kaby, Skylake and Haswell.

    Canalys says 75 M desktop in 2019 and 52 M in 2020 and 60 M is 2021 eyeballing the available bar charts.

    IDC says 65 M to 74M desktop in 2019 and 50 M in 2020

    Coffee Refresh desktop 9th enters the channel on 6.30.18. Begins with Pentium and Celeron that is unusual and top bin i9/i7 does not really show up in the channel until October 2018.

    Comet desktop 10th enters the channel on 5.16.20. Shows 9th generation is produced at least for 18 months and 9th does show two production peaks on 9.21.19 and again on 11.1.20 that is higher and a sustained block of product. I'd say 9th is easily 140 M units so 9900K = 57,419,964 that is a one Intel traditional desktop run and the single threaded is for the enterprise security vulnerability 1HT only enterprise Office buyer so 60 M 9900K makes since for enthusiasts and then 9900KS = 2,239,141 units. 3 M units is not unusual for an Extreme Edition run.

    9th might be 160 M units but that would be high, however, Intel does produce certain generations for longer than thought, longer than addressed by the commercial analysts on 1st tier OEM input that hides volumes and its apparent over 30 years especially pursuant Xeon total generation volumes.

    Specific Coffee Refresh 9th desktop you can then apply the channel percent grade SKU split against my 140 M quick take of the commercial analyst volume that I know does not represent and falls well short Coffee Refresh total production.

    9900K = 41.03%
    9900KS = 0.16%
    All i9 = 42.95%
    All i7 = 15.2%
    All i5 = 28.59%
    All i3 = 6.95%
    Pentium = 2,8%
    Celeron = 3.48%

    But that standard method, comparing against commercial analyst annual desktop statement, in my opinion right now, won't give the actual answer where Coffee Refresh is a huge volume run produced well beyond the PC OEMs Canalys and IDC get their by vendor and total volumes from, for any one year. Coffee Refresh is at least two Intel full desktop runs and the second of two peaks ramps parallel Comet ramp into Comet's volume peak.

    My best quick take on four hours of data examination.

    mb

    .
  • Mike Bruzzone - Wednesday, April 6, 2022 - link

    @ThirteenDominion, I'll get back on more data, still working on this Coffee refresh quandary and I'll add this as a to do for an update in my current SA desktop slide set that is not finished yet. I'm now thinking its less than 140 M but more than 80 M . . . I need to do the whole exercise. The main issue with many Intel generation's is that once commercial analysts stop recording for any one generation in its so called production year; 12 to 18 months, Intel still keeps making the prior generation parallel the next generation and those units are not being recorded by the commercial analysts querying first tier OEMS because their primary production has moved on, however, the channel data continues to capture the parallel generation's production. 2286G is pulled from Comet and these 'Xeon E' runs are meager in relation their Core desktop volumes. mb
  • Mike Bruzzone - Wednesday, April 6, 2022 - link

    ThirteenDominion . . . bingo!

    I've figured it out just took the rest of the day, it's as I suspected originally, 9th gen Coffee Refresh is 128,655,096 to 137,556,418 units of production and I need to get back to GPU Today that got pushed to tomorrow. I'll have this desktop assessment posted on my SA blog spot showing unit volume for Haswell through Alder by week, June 2021 through2021 shortly. mb
  • Mike Bruzzone - Wednesday, April 6, 2022 - link

    Whoops that's June 2012 through February 2022. 9900KS = 2,058,482 units and 9900K = 52,787,186 and for either product volume range high +7%. mb
  • blppt - Monday, March 28, 2022 - link

    I'd love to get one of these, but apparently people with the regular 12900K are having trouble keeping these cool under load (non AVX), even with top line AIO water cooling.

    Can only imagine the nightmare this one is going to be.
  • Mike Bruzzone - Sunday, April 10, 2022 - link

    ThirteenDominion inquired how many 9900KS were produced in unit volume for a potential 12900KS volume comparative. 9900KS is 0.016% of 9th full run volume.

    This was my last best result on channel supply data assessment. 9900KS = 2,058,482 units and 9900K = 52,787,186 and for either product volume range high +7%.

    My current thesis is 9th Coffee Refresh full run volume is likely range 128,655,096 to 137,556,418 units of production. My final full run volume is 118,349,737 units, plus end run contract completion (tie) sales reward? That gets to my thesis range high?

    Find three new slides #16 - 18 showing Intel desktop full runs supply by week not including Xeon HEDT, Haswell through Alder current ship with AMD Ryzen desktop over lay. Production volume, market share and channel supply share are stated.

    https://seekingalpha.com/instablog/5030701-mike-br...

    Mike Bruzzone, Camp Marketing
  • aagello24 - Sunday, April 10, 2022 - link

    this intel core thing is old. remember AMD did it with the apus. just intel made theirs with new tech. i smell a law suite.
  • Isaac McLean - Monday, May 2, 2022 - link

    now I have an i7-12700K, but I want to put this processor.

Log in

Don't have an account? Sign up now