Rumor: AMD’s 32-core chip will be called Ryzen Threadripper 2990X

Hong Kong hardware site HKEPC might have some fresh details on that 32-core Threadripper chip that you've all been salivating over. The site claims that the chip will be called the Ryzen Threadripper 2990X in an expansion of the current Socket TR4 CPU naming scheme.

Image: HKEPC

HKEPC doesn't specify the source for its information, but does say that it comes from a Taiwanese motherboard manufacturer. The site presents the purported CPU-Z screenshot above as well as a Windows 10 Task Manager screenshot showing all 64 threads. Curiously, only half of the threads are loaded in whatever workload the site is applying. 

The rumored 250-W TDP sounds high, but it's not bad for a 32-core CPU running at the speeds HKEPC purports to have seen. The site claims the chip will do 4 GHz single-core Precision Boost 2 speeds, 3.4 GHz minimum boost speeds across all four 12-nm Ryzen dies, and if all else fails, it'll hold a 3 GHz base clock. For reference, AMD's own 32-core, 64-thread Epyc 7601 has to hold to a 2.2-GHz base clock, a 2.7-GHz all-core boost speed, and a 3.2-GHz maximum boost clock to stay within a 180-W TDP.

Despite the CPU's power-thirsty nature at stock settings, HKEPC (or its source) already overclocked the Threadripper 2990X and claimed that it will do 4.12 GHz at 1.38 V. At that speed, the machine turned in a purported Cinebench R15 score of 6399.

HKEPC further claims that the 2990X will clock all the way up to 4.2 GHz using XFR, although this claim seems dubious. As implemented on AMD's second-generation Ryzen processors, Precision Boost 2 and XFR 2 work to raise all-core boost speeds, not peak single-core speeds. For example, the Ryzen 7 2700X can reach 4.35 GHz peak single-core speeds regardless of cooling—it's the chip's all-core speeds that are affected by the beefy heatsink needed to activate XFR. Second-generation Threadripper CPUs seem unlikely to use a different algorithm, and we would find it curious if the chip's peak boost speeds are actually lower than the Ryzen 7 2700X given AMD's past practice of binning dies for use in Threadripper CPUs.

Image: HKEPC

We'd normally expect a few cheaper versions to accompany the release of the new Threadripper, but HKEPC says that AMD might cancel those plans after Intel's bombastic 5-GHz 28-core demo at Computex and will instead simply launch the 32-core version. If that ends up being the case, perhaps Threadripper 2 will come in at a lower price than what we might otherwise expect. HKPEC unfortunately doesn't illuminate that particular data point for us. 

We do know that the upcoming chip will slot into existing X399-based Socket TR4 motherboards and that it's on its way at some point in the third quarter of this year. If folks like HKEPC are already overclocking them, we might be seeing these on store shelves sooner than later. Thanks to Videocardz for the tip.

Comments closed
    • ermo
    • 1 year ago

    An 8C/16T intel CPU @ 3.1GHz (the red graph in the tweet that chuckula posted) beating a R7 2700X @ 4.4 GHz in CB R15?

    That sounds fishy. The 10C/20T 7900X @ 4.0 GHz all core boost sure, since CB demonstrates nice scaling with more cores/threads.

    But I seriously doubt that the 8C/16T intel part @ 3.1 GHz has an IPC advantage of (2212/1964) * (4.4/3.1) =~ +60% over RyZen in CB like that screenshot seems to suggest. What is the rumoured all core boost of the 8C/16T part?

      • chuckula
      • 1 year ago

      It’s probably overclocked to around 5GHz, which is pretty standard for Coffee Lake.

        • ermo
        • 1 year ago

        So an HT IPC of (4.4/5.0) * (2212/1964) =~ 99.1% of the R7 2700X?

        An 8700K running at stock speeds (4.3GHz all boost clock) scored 1402 vs. the 4.4GHz R7 2700X which scored 1964. That translates to (4.4/4.3) * ((1402/6) / (1964/8)) =~ 97.4% IPC of RyZen?

        In that light, the comparison and your 5GHz suggestion looks legitimate enough, though I’m a bit surprised to see that the Zen+ core has slightly better HT IPC than the Coffee Lake core…

          • chuckula
          • 1 year ago

          [quote<]In that light, the comparison and your 5GHz suggestion looks legitimate enough, though I'm a bit surprised to see that the Zen+ core has slightly better HT IPC than the Coffee Lake core... [/quote<] Your math assumes perfect clock scaling which is probably optimistic as you can see even in the numbers of the alleged Threadripper2 parts in the same graphic. But given that Cinebench relies entirely on the same SSE instructions that were first introduced in the Pentium IV and that RyZen is optimized entirely for SSE and only bothers to implement anything more advanced as an afterthought, it's reasonable to say that RyZen delivers similar performance in more than 10 year old SSE instructions as a Coffee Lake part that doesn't optimize for SSE.

            • freebird
            • 1 year ago

            Don’t worry Chuckula, since Intel is bringing back the Larrabee architect, maybe Intel will “jump ship” and bring back their original 64-bit server chip Itanium (Itantic, always loved that nick-name) to compete with AMD’s Eypc “Rome/Zen2” Server chip slated for 2019.

            Since Intel’s 10nm IceLake SP chip will be a no show for the 2019 fight,

            [url<]https://wccftech.com/amd-epyc-rome-zen-2-7nm-cpus-compete-favorably-against-intel-ice-lake-xeons/[/url<] Intel can come out with an 18-year Anniversary edition of Itanium running at 3.5+ GHz.

            • synthtel2
            • 1 year ago

            If SSE is that terrible, what is Intel doing still selling Coffee Lake CPUs that don’t even support AVX in the decoder?

            If it weren’t for *Intel’s* shenanigans, I’d be able to solely target AVX and make pretty good use of the SIMD abilities of whichever CPU my project happens to be running on (modulo AVX-512). Were I doing things only a little bit differently, multiple instruction set targets wouldn’t even be worthwhile, and the only one left would have to be SSE4.1 because of Intel.

            • ermo
            • 1 year ago

            Out of curiousity, do you have access to publicly available documentation that supports your assertion? It makes little sense to me that Zen would be optimized hard for SSE(2) yet Coffee Lake isn’t? The only related thing I can think of is that Zen is optimized for AVX as opposed to AVX2?

            Is it also not the case that the Zen architectural implementation of HT is slightly better than intel’s (for a suitable value of ‘better’)? Note that I’m not disputing that intel’s single-threaded IPC (and AVX2) performance is better.

            And lest you think I’m some sort of mindless AMD drone I’ll have you know that were it not for Meltdown et al., I’d be getting the 8C/16T Coffee Lake part over the Zen+ part as the intel part is clearly better if you do any sort of high-refresh rate gaming. And since I’m aiming to get a Vive Pro to support my flight and racing sim addiction…

            • synthtel2
            • 1 year ago

            Zen’s FPUs are half as wide as Coffee Lake’s, so it executes AVX/2 instructions at half rate (it can still see some slight benefits from AVX in the front-end). IMO, this is all as it should be; going crazy with SIMD in consumer CPUs seems like a pretty bad idea, but with that sort of setup, developers can target one instruction set and make optimal use of a variety of core designs (even the ones invested in that bad idea). Intel’s playing all that bass ackwards.

    • Unknown-Error
    • 1 year ago

    the hell are you going to do with all those threads?

      • auxy
      • 1 year ago

      … rip them? (;’ω’)

        • Unknown-Error
        • 1 year ago

        nice response.

        • RAGEPRO
        • 1 year ago

        I can’t believe AMD hasn’t used [url=https://www.youtube.com/watch?v=9EP-XruFccs<]this[/url<] in some bit of marketing for the Threadripper. I guess it's kinda political but I don't know who takes KMFDM seriously.

          • MileageMayVary
          • 1 year ago

          Maybe AMD knows that KMFDM sucks?

        • jihadjoe
        • 1 year ago

        Or knit them! AMD Threadknitter would make a good yarn.

    • anotherengineer
    • 1 year ago

    Not triple X, I am disappoint.

    I also thought Chuck was going to reverse psychology troll the AMD fanboys with this
    [url<]https://www.youtube.com/watch?v=ozcEel1rNKM[/url<] Oh well, double X for me today 🙂

      • albundy
      • 1 year ago

      i think you went to the wrong site looking for triple x.

    • uni-mitation
    • 1 year ago

    I want to thank chuckula for all of his hard work! The check is just being cleared by accounting. Included is a bonus for exceeding all performance metrics plus stock options!

    Rick MoarCoars
    AMD PR Executive for TechReport

      • chuckula
      • 1 year ago

      You’re welcome.

      I’d also like a bonus for being the only person to post and discuss the un-censored version of that photo to this thread and getting downvoted for it by the same AMD fans who disingenuously pretend to actually care about these products.

        • uni-mitation
        • 1 year ago

        Duly noted. This will obviously delay the check in order to add a couple of more zeros. Your loyalty and work ethic to the firm is unquestioned. We expect that you will continue to exceed all performance metrics for the rest of the year.

        As for those “AMD fans”, management has been notified and appropriate action will be taken.

        Rick MoarCoars
        AMD PR Executive for TechReport

    • chuckula
    • 1 year ago

    You want leaks?

    I’ve got your leaks right here: [url<]https://twitter.com/KOMACHI_ENSAKA/status/1009414134977269761[/url<] But just remember, that orange score doesn't count because it clearly required a chiller unit to overclock it. And to give AMD equal-time: [url<]http://www.guru3d.com/articles-pages/amd-ryzen-7-2700x-review,9.html[/url<] [Funny how the AMD fan-squad doesn't want to see the full picture when they were slobbering over the censored version in the original story. Tell ya what, if an 8 core failure from Intel based on a core design from 2015 can beat a RyZen 2700X 2018 miracle chip in a single Cinebench [praised be its holy name] run then how about you buy me that pathetic 8 core Intel part since it will be so slow I won't be able to browse the web!]

      • ShowsOn
      • 1 year ago

      [ Tell ya what, if an 8 core failure from Intel based on a core design from 2015 can beat a RyZen 2700X 2018 miracle chip in a single Cinebench]
      How much do you think that “8 core failure from Intel” (your words) is going to cost? The 7820X is on sale for around $470 (The list price is $590) but the boost clock is only 4.3 GHz. The 8 core failure will probably need another 500 MHz boost and a higher base, which will send the price over $500, maybe even close to $600.

      So you’re comparing the 2700X to an “8 core failure” that is going to cost 50+% more.

      • auxy
      • 1 year ago

      What even is the joke here? This post is dumb. We already knew 6C CFL-S was similar to 8C Ryzen in MT stuff. This just shows that an 8C CFL-S performs about like we would expect it to, and nothing else. What’s the joke? Your AMD vs. Intel trolling is amusing when there’s a joke but this is just trying way too hard and not entertaining, interesting, or clever.

      • Zizy
      • 1 year ago

      Considering i7 8700k is only at 1400 points in your linked review, I have hard time believing 33% extra cores at lower clocks manage 2200 points (57% higher).

      Though I wouldn’t be surprised if this supposedly 3.1GHz chip is running at 5GHz with suitable cooling. Wouldn’t be the first time, eh 😀

    • Srsly_Bro
    • 1 year ago

    I saw on a different site saying that this is fake. Look at the”tools” on the cpuz screenshot. 1950x gets around 3k to 3.2k on R15. A 4.12 over clock should do much better than 6400.

    This is suspect.

      • chuckula
      • 1 year ago

      [quote<]A 4.12 over clock should do much better than 6400.[/quote<] True. In fact, it might even be over 9000.

        • Srsly_Bro
        • 1 year ago

        I hadn’t considered that. Thanks, bro.

        • cheddarlump
        • 1 year ago

        Lol, nobody has a sense of humor anymore. I see what you did there..

      • auxy
      • 1 year ago

      The “Tools” on the CPU-Z screenshot looks different because the system is in Chinese locale.

        • jihadjoe
        • 1 year ago

        So why is it the only thing that’s rendered in a different font? Also the Threadripper logo doesn’t match other CPU-Z Threadripper results.

          • auxy
          • 1 year ago

          The Threadripper logo probably doesn’t match because it’s a pre-release CPU. CPU-Z can recognize the string “Threadripper” but doesn’t have it set up with the official logo (yet).

          As far as “why the only thing rendered in a different font” goes, it’s pretty common when you are using Asian locale. I use Japanese locale and it does the same thing. It’s probably something to do with the Tools drop-down menu using some function that is locale-aware while the other buttons are not. That font is MS Mincho which is the default font that includes support for Asian characters in Windows. It’s like Arial except for Asians. I can show you a dozen other [url=https://i.imgur.com/OWHDpZe.png<]screenshots[/url<] of the same thing, that's just how it is.

      • dragontamer5788
      • 1 year ago

      [quote<]A 4.12 over clock should do much better than 6400. [/quote<] I find it suspect for the opposite reason. EPYC 7601 x2 (64-cores / 128-threads) only gets ~6109. See: [url<]https://www.reddit.com/r/Amd/comments/8qzkwf/amd_epyc_7601_64_core_cinebench_benchmark_6109_cb/[/url<] EPYC 7601 x2 has an all-core turbo of 2.7GHz. So sure, its downclocked a bit, but 64-cores @ 2.7GHz should perform better than 32-cores at 4.1GHz.

        • chuckula
        • 1 year ago

        Given the level of noise in Cinebench scores that starts to happen at this level I wouldn’t be surprised if you could get everything from about 6000 points up to close to 6500 points just by re-running the test and selecting random results.

        • Srsly_Bro
        • 1 year ago

        The 1950x does 3k stock and 3.4k overclocked with 3200 cl14.

        We know ryzen scales well with RAM speed and the epyc were definitely not running RAM that speed. I would have to do more research to see how much of a difference that makes.

        [url<]http://www.guru3d.com/index.php?ct=articles&action=file&id=33417[/url<] Also check the maximum threads for cinebench. I don't think it's 128.

    • kuttan
    • 1 year ago

    The Zen 3xxx built on 7nm process is rumored to feature 6 Cores/CCX meaning 48 cores TR CPUs incoming.

      • chuckula
      • 1 year ago

      So there’s no reason to buy a 32 core part? What exactly is the point of your post?

        • freebird
        • 1 year ago

        He doesn’t need to have a “point” to post… just like most of your posts… ;D

      • ptsant
      • 1 year ago

      Well, that’s clearly too much for 99% of users. Great for the server market, but I’d rather have a few % IPC and GHz instead for desktop uses.

        • Goty
        • 1 year ago

        First generation TR was “too much” for 99% of users, but that’s never been what it’s about. (On a side note, how well does Windows 10 run on 640K of RAM? I’ve always been curious.)

        Also, why do you assume there are to be no benefits to IPC or clockspeed for 7nm Zen?

          • chuckula
          • 1 year ago

          [quote<]First generation TR was "too much" for 99% of users, [/quote<] YOU'RE WELCOME, -- Scott Wasson & SSK [the silent founder]

          • auxy
          • 1 year ago

          [quote<]Also, why do you assume there are to be no benefits to IPC or clockspeed for 7nm Zen? [/quote<]He didn't. Why do you put words in his mouth? ('ω') I am shaming you for your aggressively black-or-white thinking. I hope you feel ashamed.

            • Goty
            • 1 year ago

            Did they not?

            [quote<]Great for the server market, but I'd rather have a few % IPC and GHz instead for desktop uses.[/quote<] Assuming 7nm Zen goes to 6 cores/CCX, when they say they'd [i<]rather[/i<] have IPC and clockspeed increases (their words, not mine), that means [i<]instead of[/i<], not [i<]in addition to[/i<]. If this is not the case, why then make the comment at all?

            • chuckula
            • 1 year ago

            I’m assuming there’s IPC increases but what evidence has AMD shown that its presumably 12 core first-generation 7nm parts that are supposed to be available in early 2019 are outclocking its 8 core late 2018 parts?

            People tend to forget that even the first generation of Ryzen launched well after GloFo had larger 14nm parts on the market (the glorious RX-480 that’s a bigger hunk of silicon than RyZen).

            As it stands today RyZen still doesn’t clock as well as 32/28nm Piledriver-esque parts. It’s a vastly better product but it doesn’t clock as well.

            There’s the usual assumption that Dennard scaling is only dead for Intel and that everybody else is hitting 5GHz with their first 7nm products. We’ll see if AMD has a 12-core part with substantially better IPC than Skylake and faster clocks out in the first half of 2019.

            • Goty
            • 1 year ago

            The fact that AMD’s 14/12nm parts don’t clock particularly well actually makes me believe more strongly that the 7nm parts will clock relatively better; you know, regression to the mean and all.

            • chuckula
            • 1 year ago

            I like your optimism!

            The fact that Intel’s 10nm parts are delayed means that Icelake is going to be AWESOME!

            This is a fun game!

            • Brother Michigan
            • 1 year ago

            Hey look! More trolling posts from Chuckula!

            • Goty
            • 1 year ago

            Notice that I said “relatively better.” With the chance for process and architectural improvements, it’s more reasonable to assume things will do better than the previous generation than worse since that’s the typical trajectory of things. I don’t imply 7nm Zen is going to be some miracle, just that it’s likely to be better in one aspect where Zen on 12LPP/14HP seems to be lacking.

            • freebird
            • 1 year ago

            “As it stands today RyZen still doesn’t clock as well as 32/28nm Piledriver-esque parts. It’s a vastly better product but it doesn’t clock as well.. ”

            Yeah, that logic is right up there with comparing the with Intel’s CoreX generations taking years to catch up with the Pentium 4EE 3.73Ghz

            Both of which are straw man arguments since the underlying CPU architecture were completely different. The longer instruction pipeline CPUs (Piledriver/Prescott) allowed higher clock speeds, while the shorter instruction pipelined CPUs (Ryzen/CoreX) were lower clock speed, but vastly more efficient.

            Not to mention that the 32nm SOI process used for Piledriver is also vastly different than 14/12nm FinFet bulk process used for Ryzen.

            That being said, your point of
            ” We’ll see if AMD has a 12-core part with substantially better IPC than Skylake and faster clocks out in the first half of 2019.”

            Is still valid especially concerning the time frame of 1H2019. I fully expect AMD to pump as many Zen2 dies into Epyc processors rather than Consumer desktops. Not knowing the yield rates/maturity of 7nm production may even limit the parts to higher selling SKUs (Epyc) until yields are up to snuff.

            Who knows, maybe AMD will even carry two masks one with six Execution units per CCX and another keeping 4 per CCX or maybe even just no dual CCX but a monolithic 8 core that they tie together with an infinity fabric 2.

            Speculation is just that…

            We’ll all see how this “game” plays out in mid-2019 with both AMD Zen2 7nm & Intel Krappy-Lake 14++++/10nm. Execution is the name of this game and we’ll see how both execute over the next year or two.

            • auxy
            • 1 year ago

            I can’t believe I’m having to explain this most basic of linguistic logic.

            For reference, YOU said: [quote<]why do you assume there are to be no benefits to IPC or clockspeed for 7nm Zen?[/quote<] in response to [quote<]I'd rather have a few % IPC and GHz instead [of a bunch of extra cores] for desktop uses.[/quote<] I hope you understand that "I'd rather have Y" in response to "there is X" does not mean or even imply the absence of Y. You could even say it in the explicit presence of Y. (example: we sit down to eat. there is mac & cheese, which I know you like. I say "there's mac & cheese." You see the mashed potatoes and say "I'd rather have the mash taters.") All it means is that when offered X, they'd rather have Y. That's all it means. This whole "why do you assume" thing is fabricated from whole cloth on your part. Do you understand that? I need to know that you understand that. You made an assumption apropos of [i<]nothing[/i<]. You understand this, right? It's something you see a lot on the internet these days and it's disgusting. You make me sick. Even if you're not intentionally being obtuse ("trolling," to misuse the word as people do these days) you're still making absolutely zero attempt to give the benefit of the doubt or even think about what you're saying! Instead, you're going on the attack and attempting to force ptsant into the viewpoint opposite your own so you can feel justified in your assault. And don't start making a reply with backpedaling and finger-pointing. Have some damn personal accountability. You screwed up and I'm shaming you for it. So feel ashamed and apologize, you absolute unit. And most importantly, [b<]DO BETTER.[/b<]

            • uni-mitation
            • 1 year ago

            Shamed just for this? I better not leave that toilet seat up! Bad dog, very bad dog!

            And I am like, “I like the abuse. What do you feel about double-dipping?”

            “Don’t you have no shame, you disgusting dog?”

            uni-mitation

            • auxy
            • 1 year ago

            [url=https://i.imgur.com/gnfCLj7.png<](´・ω・`)[/url<] [sub<]this is a link[/sub<]

            • uni-mitation
            • 1 year ago

            How disgusting? The devil is in the details.

            uni-mitation

            • Goty
            • 1 year ago

            Wow, I must have really hit a nerve. Maybe tone down the personal attacks a bit and I might take what you say more seriously.

            • auxy
            • 1 year ago

            If you take “you made a mistake, own up to it” as a personal attack then it really says a lot about your character. I hope you reflect on this instead of making another post attempting to deflect your error back onto me.

            • Goty
            • 1 year ago

            I didn’t deflect. He literally used the word “instead” in his own post, but your irrational hate and desire to call me names is obviously blinding you to that fact. So, have a good day and please make sure to not go [i<]too[/i<] far off the deep end.

            • auxy
            • 1 year ago

            You’re still deflecting, and also crying about “personal attacks” to try and paint yourself as the victim. You aren’t the victim here. You are the aggressor. You aggressively misrepresented what someone else said and I’m calling you out on it.

            I already pointed out to you and explained that “instead” does not give you license to assume what you did and you’ve completely failed to acknowledge it. I’ve got you dead to rights, so either own up and be a mature adult or shut up and prove to everyone present that you don’t have the integrity to admit when you made a mistake.

            • Goty
            • 1 year ago

            I didn’t make a mistake, but hey, believe whatever you want. Go on committing the same acts you’re accusing me of, it’s not making you look any better.

            • Srsly_Bro
            • 1 year ago

            Let me see if I can help, bro.

            I say, “I would like to do more research before we enact strict emissions regulations.”

            You respond, “wtf you don’t care about the planet. I have kids. They could die. So the polar bears have already died. You’re a monster. You have no soul.”

            Your logical jumps result in those type of reactions.

            • Goty
            • 1 year ago

            Notice the difference in vocabulary between your post and the one in question. That is [i<]literally[/i<] the point here. Maybe ptsant actually meant "before" rather than "rather" (see what I did there?), but that's not what was written and that's the only thing we have to go on. Also, I find it funny that the people doing all of the attacking in this thread are the ones accusing me of attacking ptsant when all I did was ask a(n entirely reasonable) question. There is a polite way in which each of you could have responded to my post, but you decided to troll instead, and I think that says a lot.

            • Srsly_Bro
            • 1 year ago

            I illustrated the point to help you understand your logical Superman-like leaps.

            • Srsly_Bro
            • 1 year ago

            My name is srsly_bro, and I approve this post.

            Thank you, auxy.

        • VincentHanna
        • 1 year ago

        “640Kb of ram ought to be enough for anybody.”

          • Srsly_Bro
          • 1 year ago

          Said millions of other dorks on forums trying to sound edgy..

          Let it go, bro.

            • chuckula
            • 1 year ago

            Bill Gates, while a dork, wasn’t one of them BTW!

            • Srsly_Bro
            • 1 year ago

            Agreed.

            I was teasing him for a desperate attempt to be funny or edgy.

        • blastdoor
        • 1 year ago

        I guess it depends on how many of those cores are working, right?

        If just 1% yield 6 working cores per CCX then AMD will be looking pretty smart 😉

    • Chrispy_
    • 1 year ago

    I’m more curious about the result that has been censored.

    Seems like it could be an R7 2800X maybe, for which they are under NDA still?

      • chuckula
      • 1 year ago

      That’s the 28 core Intel part!
      Intel has been bribing the leakers to hide its results.

        • kuttan
        • 1 year ago

        Yes Intel has been bribing the leakers to hide its 8 core CPU result beating its own 10 core CPU!

          • chuckula
          • 1 year ago

          You paid shill.

          It’s a well established fact that Intel’s failed so-called “14nm” process is incapable of producing chips with 8 cores.

            • kuttan
            • 1 year ago

            “Gold Subscription” just to BS comments ? Despite all your intense 24×7 paid AMD hate posts, AMD is actually growing! Go outside and have some girlfriends 😀

            • Shobai
            • 1 year ago

            I won’t try to speak for chuckula but man, I found one to be almost too much to handle…

            • uni-mitation
            • 1 year ago

            But what about those that don’t have one at least to handle? BTW, do you put a muzzle on yours?

            uni-mitation

    • chuckula
    • 1 year ago

    Not bad for a wraith-cooled early sample that hits 4.2 GHz at 1.05V AMD!

    Some of you paid Intel shills might question why the Cinebench score of a 4.2GHz Ripper2 is [i<]purportedly[/i<] lower than the 4 GHz Ripper 2. Well I've got news for you: It's not. It's just that Ripper2 is so fast that Cinebench's score maxes out and wraps around giving the totally fake impression that the 4.2GHz overclock is slower than 4GHz when it's totally faster.

      • Pville_Piper
      • 1 year ago

      “Some of you paid Intel shills might question” I would but they haven’t sent my check yet…

      • Beahmont
      • 1 year ago

      Yeah… those scores don’t make any sense at all.

      Even the Intel Scores are questionable. There wasn’t much IPC improvement between Ivy Bridge and Haswell, but there was some IPC improvement. A 1 GHz faster Haswell processor at 4.4 GHz should at least scale linearly compared to a 3.4 GHz Ivy Bridge processor, if not better than linearly. But it doesn’t according to this screenshot. A 29% increase in clock speed (4.4 GHz / 3.4 GHz) on the Haswell chip only produces a 24% increase in score (822 / 662).

      I mean isn’t that the whole point of using Cinebench in the first place to have a test that is known to scale more or less linearly with clock speed, IPC improvements, and cores? But here we have a 4.4 GHz Haswell chip turning in a performance that should be less than just over clocking a 3.4 GHz Ivy Bridge chip to 4.4 GHz.

        • chuckula
        • 1 year ago

        Nice to see that at least one person on this site doesn’t reflexively down the koolaid and actually thinks about what the numbers mean.

        Bonus points for looking at the fonts on the buttons of the CPU-Z screenshot.

        • Goty
        • 1 year ago

        ServeTheHome (https://www.servethehome.com/cinebench-r15-is-now-a-broken-as-a-benchmark-and-11-5k-surpassed/) has an article about the variance experienced in Cinebench runs which becomes exacerbated with higher performance parts (basically, it boils down to variance in initialization having an outsized effect on overall scores as performance increases and that initialization time becomes a significant portion of the overall render time.)

        Either this or something like thermal throttling (not exactly unexpected with 32 cores at 4.2 GHz unless you’re Intel and throw a water chiller in the mix) could easily explain that result.

        *EDIT* I realize this post is a little scatterbrained. The first part applies to variation in all results, the last just to the supposed 4.2 GHz TR score being lower than the others.

          • Beahmont
          • 1 year ago

          Okay, but if it’s self overclocking via various features (XFR2, Precision Boost 2, etc) and it’s losing performance because it’s hitting it’s thermal limits compared to it’s lower clocked brethren, doesn’t that mean there’s an issue with the self overclocking features?

            • Goty
            • 1 year ago

            I think PB2 is disabled when manually overclocking, though I might be wrong about that.

      • kuttan
      • 1 year ago

      Your comment now made your employer Intel even more awful 😛

      • Kretschmer
      • 1 year ago

      Damn it! I’ve been paying Intel for 15 years now; whoops.

Pin It on Pinterest

Share This