Mindfactory CPU sales show Coffee Lake couldn’t be stopped in June

One way or another, retail processor sales numbers from German retailer Mindfactory make their way onto the internet courtesy of the charting work of reddit user ingebor. Those figures give us a peek into the shape of processor sales in the enthusiast and DIY market. ingebor has compiled those figures for the month of June, and their work shows fascinating trends over the past year and in the past couple of months.

Source: ingebor, via reddit

You'll want to click through to these charts using the source links below each image. Going by the raw number of CPUs sold at Mindfactory in June, Intel's Core i7-8700K continues to be a juggernaut. We can also see that the Ryzen 7 2700X is outselling the non-X Ryzen 7 2700 by a many-to-one ratio, although that ratio doesn't hold when we consider the Ryzen 5 2600X versus the Ryzen 5 2600. The six-core second-gen Ryzen parts appear to sell in roughly equal proportions at Mindfactory. Many first-generation Ryzen CPUs appear to still be moving off store shelves, too, perhaps thanks to discounts and price cuts. Overall, 45% of the CPUs Mindfactory sold in June were made by the red team and 55% by the blue team, down from near-parity in May.

Source: ingebor, via reddit

Looking at Mindfactory's revenues by CPU manufacturer, it's clear that AMD has had to chase an aggressive value-player strategy in the wake of the Coffee Lake launch. Even as raw CPU sales have returned to something approaching parity in the wake of strong i7-8700K sales from broad Coffee Lake availability, the contribution of AMD CPUs to Mindfactory's CPU revenues has shrunken considerably with time. The lucre pulled in from i7-8700K sales alone last month would make for a goodly portion of the revenue contribution of AMD's entire product stack. Compare the fact that AMD accounted for 45% of CPU sales in raw numbers but just 38% of revenues last month versus a much closer proportion of both sales and revenues a year ago.

Source: ingebor, via reddit

ingebor also breaks down CPU unit sales by generation over the past four months. These graphs show a surprisingly strong showing in June for the Summit Ridge silicon that powers first-generation Ryzen CPUs, even in the wake of the superseding Pinnacle Ridge parts. A similar breakdown for revenue shows that Pinnacle Ridge parts are pulling in a rather more lopsided 51% of all AMD CPU revenue versus 30% for Summit Ridge CPUs in the past month. Meanwhile, Coffee Lake CPUs account for an overwhelming majority of unit sales and revenue alike from fans of the blue team.

That massive difference perhaps goes to show just how important the introduction of the Coffee Lake family proved in spurring enthusiasts to upgrade their Intel systems, while AMD buyers can still get relatively similar performance from first- and second-generation Ryzen parts with some judicious overclocking.

Overall, it's clear from Mindfactory's numbers that the i7-8700K was both a necessary response to AMD's Ryzen onslaught and an enduring smash hit among enthusiasts. Even the introduction of Pinnacle Ridge Ryzen CPUs hasn't done much to slow the most-caffeinated Coffee Lake part's roll. It's worth remembering that these numbers are a look at activity from one retailer in one region only, but it's still fascinating to see how the battle between a newly-competitive AMD and a playing-catch-up Intel has run its course over the past year. We can't wait to see the developments this competitive pressure has yet to yield from both companies.

Comments closed
    • Zizy
    • 1 year ago

    1800X vs 7700k had Intel slaughtered in most MT tasks, while AMD was somewhat behind in games and lacked iGPU.
    2700X vs 8700k has Intel somewhat behind in MT tasks and AMD somewhat behind in games and lacking iGPU.
    Plus the 8400 is a good cheap gaming CPU, 8600k is great for gaming and 8700 is saved by its iGPU and TDP, making it a compelling compute chip (AMD requires dGPU).

    So, good performance of the coffee lake is understandable. I really wonder why would people buy i7 7700k though, which still shows as a reasonable portion of Intel pie. 311 eur for it, vs 340 eur for the 8700k and 229 eur for the 8600k… there is no reason to get the old i7 even ignoring AMD parts. (excluding very niche scenarios of “want to upgrade just the CPU”, which would be generally better served by selling the chip with board and buying both new)

      • Kretschmer
      • 1 year ago

      I don’t get it either. I’m happy with my 7700K but would definitely pay 10% more for 2 more cores.

        • K-L-Waster
        • 1 year ago

        Upgrades from Skylake i3 or i5 into existing mobo perhaps? Either that or really good discount deals?

    • ronch
    • 1 year ago

    Let’s not forget how Intel compilers intentionally make non-Intel processors
    [url=http://www.agner.org/optimize/blog/read.php?i=49<]look less compelling [/url<]. Yeah I know it's a different topic but I think it's worth recalling.

      • Redocbew
      • 1 year ago

      The reality is that if there’s a problem with your software because you used the wrong toolchain to build it, then that’s on you. There are alternatives, and performing some song and dance about the way it should be is just going to get you labled as someone hopelessly out of touch who doesn’t understand their role in the business. Even if you did manage to find a sympathetic ear what would you expect them to do about it?

      I suppose it’s a good thing that we still have idealists in the world who get bent out of shape about these things. It’s just not an argument which is all that useful towards building a finished commercial product.

        • ronch
        • 1 year ago

        Did you read what was discussed by Agner there?

          • Redocbew
          • 1 year ago

          Did you? The last post in that thread states that in the latest version of ICC Agner Fog isn’t sure exactly how the CPU dispatcher is selected. Meaning you don’t really have a case, because this could all be for nothing.

          This story is nearly a decade old, and if true that alone should convince you that nobody really cares. I’ll ask you again: even if you found someone who did what would you expect them to do about it besides using a different toolchain? This is a useless point, and if it’s true then you’re doing the same thing Intel is by attempting to skew the facts to fit your own agenda.

            • ronch
            • 1 year ago

            Nobody cares because most folks just choose to ignore it but that doesn’t vindicate Intel. And even if it’s a decade old do you really think it’s beyond Intel to continue doing what they were caught doing and fined?? I have no agenda, just pointing out facts that someone found out and are likely still going on to this day. Intel wants to win even if it means engaging in shady practices, and if you choose to support a company like that then it says quite a bit about your principles. Carry on.

            • Redocbew
            • 1 year ago

            [quote<]I have no agenda[/quote<] Seriously bro? [quote<]Intel wants to win even if it means engaging in shady practices[/quote<] Since this is the Internet what I say next might sound strange, but I have only a limited amount of outrage to spend on any particular thing. I choose to spend it elsewhere, because the shenanigans we see from the companies we often talk about here pale in comparison to many others. If GCC or LLVM didn't exist, then perhaps I might consider this a bigger deal(if it's still a thing), but they do, so whatever.

      • Kretschmer
      • 1 year ago

      You should redirect your partisan energy to politics or religion, where it would be more welcome.

        • ronch
        • 1 year ago

        So you hate it when someone tells you that a company engages in different shady practices to hurt a much smaller competitor? Is Intel your favorite company?

      • K-L-Waster
      • 1 year ago

      It’s funny how with the AMD Defense Force if AMD sells well it’s because they’re awesome but if anyone else sells well it’s because they’re cheating.

        • ronch
        • 1 year ago

        Isn’t Intel really cheating though? See how those agencies went after Intel as well. You think they’re no longer doing it now?

      • Krogoth
      • 1 year ago

      It is hardly a secret.

      Intel’s marketing has been known to use its clout to make Intel platforms look “better” in popular synthetic benchmarks for almost 20-25 years now.

      It rarely translates into real-world stuff but the marketing damage is done. It creates a distorted image that Intel is the “superior” brand name for the masses.

        • Krogoth
        • 1 year ago

        I guess the IDF can’t allow the cold, hard truth to come out.

        AMD isn’t exactly innocent either, but historically speaking Intel has the played the whole “benchmark optimization game” time and time again.

          • Klimax
          • 1 year ago

          Sorry, but I don’t see anything that could be called evidence…
          And ICC ceased to be bad for AMD since around 2012. (If I find link again, I’ll edit it in – Behardware’s archive is gone so archive.org has to be used)

    • strangerguy
    • 1 year ago

    $300 is now the new best performance/price mark with the 8700K/2700X. If I’m cash strapped I would rather just buy a <$200 Haswell i5 OEM refurbs which won’t be winning Cinebench scores but are still ridiculously fast for games than bother with $200 new current gen CPUs.

      • Kretschmer
      • 1 year ago

      I mean, if you want to save $100 the i5 with six cores is a pretty safe bet. The mainstream gaming market is finally getting around to four cores, it’s likely that you’ll be safe with six real cores for quite some time.

        • strangerguy
        • 1 year ago

        By the same logic why not just spend $100 more for a 8700K? I’m gonna get the most of the $100 back anyway when it comes to the resale down the road. People who settle for the 6 core i5s are being penny wise, pound foolish.

        Once again, If $100 is a big deal to anyone in the days of uber-expensive DDR4 and GPUs why are they even looking at new parts to begin with?

    • Krogoth
    • 1 year ago

    8700K influx comes from Sandy Bridge-Haswell crowd who ignored Skylake/Kaby Lake and were waiting for a mainstream six-core SKU from Intel.

    • blastdoor
    • 1 year ago

    My impression is that gaming is popular among a large segment of DIYers, and of course games tend to benefit more from better per-thread performance than from more threads. Improving IPC or clock speeds is clearly something AMD needs to do in order to get a larger share of the gamer market.

    [s<]Of course, even if AMD somehow manages to do that, intel always has its insurmountable process lead to fall back on. [/s<]

      • jihadjoe
      • 1 year ago

      AMD is also a lot more popular among DIYers than with people who buy pre-builts.

      • Krogoth
      • 1 year ago

      PC gaming market isn’t really that big though in the grand scheme of things. AMD is after the enterprise, SMB markets with its Ryzen Pro/Epyc platforms. There’s far more revenue there.

      Intel makes far more revenue from its mainstream, SMB and enterprises crowds then the DYI segment.

      Besides, GPU matters far more in PC gaming then the CPU. The CPU only matters in certain gaming genres/niches (grand strategy, RTS. emulators and [b<]dwarf fortress[/b<])

        • Kretschmer
        • 1 year ago

        CPU bottlenecks are real when you’re running at 100+Hz, especially if you’re using strobing. While a steady 80FPS with a given processor seems fantastic, it’ll be disappointing to a gamer that runs his strobing display at a fixed 120Hz.

        There are games out there where CPUs heavily differentiate:
        [url<]https://www.anandtech.com/show/12625/amd-second-generation-ryzen-7-2700x-2700-ryzen-5-2600x-2600/19[/url<] Also, I get the feeling that the "walk around a quiet area" method of benchmarking CPUs on many sites understates the CPU load of games when you have large battles or other taxing scenarios. I understand that repeatable benchmarks necessitate simple scenarios, but I've seen instances where a given processor is more limiting in real play than frame times suggest. In light of all of these, I tend to error on the side of better IPC and clock speeds when building a gaming PC.

          • Krogoth
          • 1 year ago

          Nope, the majority of gaming tiles are still GPU-limited outside of running low resolutions (720p or lower). GPU is still by far the largest factor in gaming performance.

          Your cherry-picked example shows that 8700K is only marginally faster than 2700K at GTA:V a game known to be CPU-bound. In Anand’s entire bench suite, the only game where the 8700K shows an noticeable difference over the 2700K is Civilization VI a 4x strategy game.

            • Ifalna
            • 1 year ago

            Ever tried playing a MMO?
            Your CPU will be the limiting factor, when many people are around, esp when you do ingame parsing of combat data, while your GPU twiddles it’s thumbs.

        • jihadjoe
        • 1 year ago

        I suspect AMD takes an even smaller pie out of those markets because purchasing decisions have to be approved by PHBs. They’re pretty popular with the DIY crowd who are relatively well-informed, and as chuckula said they’re also quite popular in Germany. Mindfactory is a melding of those two groups, and if AMD can’t win over there they probably aren’t winning anywhere else.

        IIRC MindFactory was also the first, and only retailer that reported an AMD marketshare win back during the Ryzen launch.

      • shank15217
      • 1 year ago

      I don’t know if your IPC story holds true after the performance changes due to spectre/meltdown patches.

        • K-L-Waster
        • 1 year ago

        Perhaps you missed this?

        [url<]https://techreport.com/review/33299/recent-pcs-have-little-to-fear-from-intel-spectre-microcode-updates[/url<]

          • chuckula
          • 1 year ago

          Or this: [url<]https://techreport.com/review/33531/amd-ryzen-7-2700x-and-ryzen-5-2600x-cpus-reviewed/7[/url<]

            • shank15217
            • 1 year ago

            Dont 8700K turbo upto 4.7Ghz? Is that an ipc comparison? The Ryzen CPU also got a slight ipc bump with summit ridge while coffee lake got a slight ipc drop. The gap has narrowed about 6-7%, not as cut and dry as you think.

    • adamlongwalker
    • 1 year ago

    Though nice to have, I would rather use the findings from Amazon over this company. Simply the fact that Mind factory is more regional and Amazon is more global.

      • Jeff Kampman
      • 1 year ago

      [url<]https://www.amazon.com/Best-Sellers-Electronics-Computer-CPU-Processors/zgbs/electronics/229189[/url<] This isn't broken down to raw numbers but you can kind of get a sense of where the market is with it.

      • chuckula
      • 1 year ago

      Historically AMD has been popular in Germany since that’s where basically all of AMD’s CPU parts were manufactured [in Dresden] up until last year.

        • Sahrin
        • 1 year ago

        AMD’s parts have been made at Fab 36 (now Fab 8) in New York for quite some time (6 years)

          • chuckula
          • 1 year ago

          Funny how the usual suspects downthumbed me while upthumbing you. But I suppose you have the “correct” emotional attitude about AMD so literally spreading factually inaccurate information gets a pass.

          Anyway, according to Wikipedia the New York fab didn’t go into mass production until until [url=https://en.wikipedia.org/wiki/GlobalFoundries#Fab_8<]2012[/url<] and there's no record whatosever that it produces the 32nm SOI process that all AMD CPUs used until 2017. That means every single Bulldozer and Piledriver part. The New York operation might have produced some of the 28 nm APUs along with Dresden, but in an article about the sales of high-end Ryzens it's blatantly obvious that the correct context is to note that [b<]literally every[/b<] Bulldozer/Piledriver chip came out of Dresden and AMD only starting selling non-Dresden produced CPUs in 2017.

            • just brew it!
            • 1 year ago

            [quote<]the New York fab didn't go into mass production until until 2012 and there's no record whatosever that it produces the 32nm SOI process[/quote<] *Ahem*... [url<]https://semiengineering.com/32nm-soi-glofo-fab-8%E2%80%B2s-1st-silicon/[/url<] The money quote: [quote<]"GlobalFoundries’ Fab 8 running IBM’s 32nm SOI technology represents a terrific opportunity for accelerating SOI innovation into broader markets,” explains Horacio Mendez, Executive Director of the SOI Consortium.[/quote<]

            • chuckula
            • 1 year ago

            Yes, it says [b<]IBM[/b<] over and over while noting that AMD's products are made at other GloFo facilities.

            • just brew it!
            • 1 year ago

            It says IBM all over because IBM invented SOI. [i<]It's inherently an IBM process.[/i<] It also specifically says that Fab 8 was producing 32nm SOI wafers. Your claim that the article I linked specifically states that AMD CPUs are produced elsewhere is simply untrue. That said, nowhere do they state that they [i<]were[/i<] producing AMD Bulldozer/Piledriver cores there, and as Antimatter points out, the heatspreaders for the FX parts all (AFAIK) say "Diffused in Germany". So I'd certainly be willing to believe that all of the FX cores were done in Dresden. OTOH, at least some (if not all) Ryzen parts say "Diffused in USA"... so those likely [i<]are[/i<] coming from Fab 8.

            • chuckula
            • 1 year ago

            Please try to use some critical reading skills.

            I could point out the specific passages of that article that at no point say that the NY fab is making a single AMD product and point out how reading the statement that GloFo is making all of AMD’s 32nm chips has jack-all to do with a fab that’s dedicated to making Cell processors and a few IBM Power chips. But instead I’ll focus on this:
            [quote<]GF’s been turning out 32nm SOI chips at its other fabs since June ’11. [/quote<] Which kinda makes sense since you're effectively alleging TIME TRAVEL by having a fab that opened in 2012 be responsible for making the Bulldozer silicon that was allegedly on sale and being reviewed by TR in 2011. If AMD could really do that then they would have launched 7nm chips last year! But the real kicker to all of this is our friend Google who says you are wrong. Here's a fun bet. Click these links. Every legitimate picture of an FX CPU that says "Diffused in Germany" on the heatspreader? You pay me $100. Bonus $1000 for every TR article (the ones that showed the actual chip at least) that says "Diffused in Germany". Every one that was made in the "USA" or "New York"? I'll pay you back $200. Looks like you're buying me a nice 28 core Skylake-X system! I'll even start off with our ultra-premium FX-9590: [url<]https://www.google.com/search?q=fx+9590&source=lnms&tbm=isch&sa=X&ved=0ahUKEwjyvrKt3YHcAhUH6YMKHUJ6DA4Q_AUIDCgD&biw=3413&bih=1796[/url<] [url<]https://www.google.com/search?q=fx+8150&source=lnms&tbm=isch&sa=X&sqi=2&ved=0ahUKEwiO6bW73YHcAhUo04MKHah-DcUQ_AUIDCgD&biw=3413&bih=1796[/url<] [url<]https://www.google.com/search?q=fx+8350&source=lnms&tbm=isch&sa=X&sqi=2&ved=0ahUKEwjfxMLH3YHcAhUKyoMKHRv5CwIQ_AUIDSgE&biw=3413&bih=1796[/url<] [url<]https://www.google.com/search?q=FX-8370&source=lnms&tbm=isch&sa=X&ved=0ahUKEwiB0pnW3YHcAhUpw4MKHaqDANwQ_AUIDSgE&biw=3413&bih=1796[/url<] Hell, it turns out the APUs on both the 32nm SOI and 28nm bulk process were all made in Germany too: Here's our pal Llano: [url<]https://www.google.com/search?biw=3413&bih=1796&tbm=isch&sa=1&ei=7so6W7vGGqa9jwTVjYrYCA&q=AMD+llano&oq=AMD+llano&gs_l=img.3..35i39k1j0i24k1l3.1245.3352.0.3456.17.14.3.0.0.0.137.861.11j1.12.0....0...1c.1.64.img..2.13.752...0j0i67k1j0i30k1j0i5i30k1j0i8i30k1.0.oUCf6ExNL60[/url<] And Trinity: [url<]https://www.google.com/search?biw=3413&bih=1796&tbm=isch&sa=1&ei=0so6W5GGCcXajwTYorTwDg&q=AMD+Trinity&oq=AMD+Trinity&gs_l=img.3..0i30k1j0i24k1l8.118699.120083.0.121484.11.10.0.1.1.0.123.688.7j2.9.0....0...1c.1.64.img..1.10.692...0j35i39k1j0i67k1j0i8i30k1.0.GYC4C1QwPQQ[/url<] And our pal Kaveri: [url<]https://www.google.com/search?biw=3413&bih=1796&tbm=isch&sa=1&ei=Jss6W-yBFcTKjwSBuImIBA&q=AMD+Kaveri&oq=AMD+Kaveri&gs_l=img.3..35i39k1j0l2j0i30k1l6j0i24k1.99068.101096.0.101218.7.7.0.0.0.0.113.556.5j2.7.0....0...1c.1.64.img..0.7.552...0i67k1j0i8i30k1.0.q55noISvcnQ[/url<] And this statement literally agrees with what I said in my first post but tries to phrase it in a way that thinks you're outsmarting me: [quote<]OTOH, at least some (if not all) Ryzen parts say "Diffused in USA"... so those likely are coming from Fab 8.[/quote<] Read what I wrote. I said "until last year". You know, 2017, the year that RyZen finally got out the gates? Oh, and this is flat-out wrong too: [quote<]It says IBM all over because IBM invented SOI.[/quote<] Wrong for lack of any supporting evidence and contradicted by Wikipedia: [quote<]In 1990, Peregrine Semiconductor began development of an SOI process technology utilizing a standard 0.5 μm CMOS node and an enhanced sapphire substrate. Its patented silicon on sapphire (SOS) process is widely used in high-performance RF applications. The intrinsic benefits of the insulating sapphire substrate allow for high isolation, high linearity and electro-static discharge (ESD) tolerance. Multiple other companies have also applied SOI technology to successful RF applications in smartphones and cellular radios. [/quote<] That fact that a full decade later IBM commercialized one version of SOI that even AMD abandoned as fast as it could has zippo to with IBM "inventing" anything. I seem to recall literally nobody claiming that Intel "invented" FinFets when here we are 6 years later and literally everybody uses them.

            • just brew it!
            • 1 year ago

            IBM/AMD’s SOI is a different process from Peregrine’s, in that it uses a silicon base with an insulating oxide layer as the substrate instead of sapphire. Silicon-on-sapphire is expensive and defect-prone; getting rid of the sapphire substrate is what opened the door to SOI being used in mainstream products.

            And yes, I mis-remembered some of the details. AMD’s original SOI manufacturing process was actually co-developed with Motorola (not IBM), but used IBM’s SOI design libraries.

            Mea culpa.

            Edit: IMO it is still fair to say that IBM was instrumental in enabling/promoting widespread commercial adoption of SOI.

            • YellaChicken
            • 1 year ago

            You ironically started that post with [quote<]Please try to use some critical reading skills[/quote<] JBI actually acknowledged quite openly that you were right about pre-ryzen parts being made in Dresden. He never disputed that. Read all his replies to you again, carefully this time.

            • Antimatter
            • 1 year ago

            The heat spreaders on Bulldozer and Piledriver actually say “Diffused in Germany, Made in Malaysia”.

            • freebird
            • 1 year ago

            I enjoy down voting Count Chuckula usually (but couldn’t this time 😉 , but he is correct AMD’s 32nm SOI process was produced in Dresden, Germany at Fab 38. Fab 36 is also there under GF ownership now. The only thing that probably wasn’t built there was some APUs that moved to 28nm Bulk process back in the Rory Read days I believe. I think there was some news about even TSMC making some of them, but manufacturing being moved to Fab 8 probably as capabilities permitted.

            That and I believe all of the IBM SOI chips were/are produce in the East Fishkill Fab that GF got from buying IBM Foundry business in 2014?

            Not sure what all the p!ssing is about over old process tech any how… the bigger question is Intel gonna get 10nm server parts out before AMD can steal a large and very SIGNIFICANT chunk of Intel’s Server business not to mention very high margin revenue. This is why I’m convinced AMD is pulling out all the stops and going to produce 7nm Zen 2 Epyc CPUs at both TSMC & GF (when it comes on line there) because AMD has a great opportunity to PAC MAN (munch) a whole lot of Intel’s server business in 2019 depending on how many 7nm Epyc chips they can produce. AMD already assumed 7nm Epyc was going to have to go toe-to-toe with 10nm Ice Lake in 2019 (which seems to have evaporated like dry ice) with Cascade Lake coming in 2H2018 and now supposedly Cooper Lake on 14++ in 2019.

    • chuckula
    • 1 year ago

    Oh sure, the Germans might like Covfefe Lake, but remember… they also [url=https://www.youtube.com/watch?v=SwpM0afTteE<]love David Hasselhoff.[/url<] I think that by the end of the month we'll have [b<]32[/b<] reasons why nobody will be buying Covfefe lake anymore. Even the man-perm loving Germans.

      • Concupiscence
      • 1 year ago

      Nearly 20 years hence, I still remember the look of visceral disgust a German exchange student at my high school flashed when somebody mentioned David Hasselhoff. “Ugh, he’s awful! Just this bellowing old man making music for old people, I can’t stand him.”

        • meerkt
        • 1 year ago

        But…
        [url<]http://mentalfloss.com/article/77591/berlin-has-hidden-david-hasselhoff-museum[/url<]

      • Kretschmer
      • 1 year ago

      Is this parody? I think it’s parody.

Pin It on Pinterest

Share This