11 ms·
High End CPUs – Intel vs. AMD
- rgbrenner 9y ago1 user reported the score: https://www.cpubenchmark.net/cpu.php?cpu=AMD+Ryzen+Threadripper+1950X&id=3058 https://www.cpubenchmark.net/cpu.php?cpu=AMD+Ryzen+Threadrip... Edit: just noticed the title was changed. Originally it said something about Ryzen beating Intel's processors.
- sjsotelo 9y agoThis article also favors AMD http://www.techradar.com/news/ryzen-threadripper-16-core-processor-impresses-with-a-beefy-overclock-to-41ghz http://www.techradar.com/news/ryzen-threadripper-16-core-pro...
- baobrain 9y agoTechradar's reporting...has had some issues. Initially the graphs on their threadripper review were misleading (link to imgur album [0]) and later changed. Anandtech[1] (prior discussion[2]) and Gamers Nexus[3] are much more reputable and go further in depth. [0] https://imgur.com/a/SuzY9 https://imgur.com/a/SuzY9 [1] http://www.anandtech.com/show/11697/the-amd-ryzen-threadripper-1950x-and-1920x-review http://www.anandtech.com/show/11697/the-amd-ryzen-threadripp... [2] https://news.ycombinator.com/item?id=14979151 https://news.ycombinator.com/item?id=14979151 [3] http://www.gamersnexus.net/hwreviews/3015-amd-threadripper-1950x-1920x-review http://www.gamersnexus.net/hwreviews/3015-amd-threadripper-1... Edit: clarification
- prodikl 9y agoyeah i was gonna say. give it a week or a month to even out and so we can get more samples before we crown anyone king
- xmichael99 9y agoWow! Hard to believe I would ever, ever, ever see AMD on the top of that list! Amazing! Forget about the price, just amazing to see AMD at the top, now factor in the price and wow Intel is screwed. Not because of this one release, but because they are getting pounded from all angles. Intel's strategy of limiting the PCI throughput to hold GPU manufactures back is over, and these AMD cpu's are going to be paired super well with GPU makers, mostly Nvidia but of course their own ATI, which is really going to making Intel look sad soon. Boat loads of major players have been irked by Intel holding back PCI throughput, AMD let it rip, and "thread ripped" too!
- thethirdone 9y ago> Intel's strategy of limiting the PCI throughput to hold GPU manufactures back is over, and these AMD cpu's are going to be paired super well with AMD, which is really making Intel look sad soon. Boat loads of major players have been irked by Intel holding back PCI throughput, AMD let it rip, and "thread ripped" too! In what way does the increase in PCI lanes stop holding gpu manufacturers back? To the best of my knowledge a single GPU doesn't use more than 16 lanes. Only multi-gpu systems are held back by the PCI lanes. So I can see threadripper being great for supercomputers, but I don't see how it affects Nvidia or AMDs GPU section.
- MattSteelblade 9y agoThe vast majority of Intel's lineup--at least for the last few years--is only 16 lanes.
- thethirdone 9y agoI get that, but given that GPUs currently can't use (for physical connector) / don't benefit from more than 16 lanes. Unless you are using multiple GPUs or another PCI device, you would not benefit from more than 16 lanes. How does that hold GPU manufacturers back?
- siberianbear 9y agoSome parallel computing problems are best solved by one CPU and a lot of GPUs. In some of these problems, inter-GPU communication is important. If you have each GPU using a single PCI lane, then the compute power can be bound by iner-GPU communication bandwidth.
- thethirdone 9y agoYeah, there are definitely cases for threadripper's PCI lanes. But a single GPU is not one of those cases.
- 9y ago
- axaxs 9y agoI don't mean to diminish the efforts of anyone involved, but I truly feel one man more or less moves the direction of the CPU industry: Jim Keller. He, among others, invented x86_64 at AMD during its previous glory days. AMD dominated the competition. He led the Apple chips at A4, and Apple chips then and now dominate the mobile competition. He came back to AMD and helped create Zen, to obvious results. Apparently he now works for Tesla. In any event, this guy seems to have the Midas touch wrt CPUs, it's a shame there isn't more written about him or more importantly, by him.
- thehobgoblin 9y agoI remember how excited I was when I heard the announcement that Keller was to work on a new project at AMD a few years back. And I can certainly relate to the latter point: As an electronic engineer that likes to dabble with IC design, I'd love to consume some of the wisdom that the man could impart.
- posguy 9y agoThe timeline looks about right, SA says he moved back to AMD in 2012[1], so figure 4 years to design a new core and get a few test runs in, then a year to ramp enough so you can sell said chip. That being said, I think it has more to do with an overall culture change, the technical talent to build a good core has been at AMD this whole time. Jim Keller said as much back in 2014[2], and there is nothing special about Zen's architecture from a technical standpoint. What Zen is is a well executed conventional high performance CPU core. By not splitting up resources like previous leadership at AMD, they were able to build a solid, competitive core. This type of investment takes 5+ years to come to fruition, from thought to test runs of silicon. 1 - https://semiaccurate.com/2012/08/01/apples-cpu-architect-jim-keller-moves-back-to-amd/ https://semiaccurate.com/2012/08/01/apples-cpu-architect-jim... 2 - https://www.youtube.com/watch?v=SOTFE7sJY-Q https://www.youtube.com/watch?v=SOTFE7sJY-Q
- axaxs 9y agoWell...I don't know the entire history, but only my memory...which admittedly is never an accurate source. At that time, it was stated Jim Keller was brought back to AMD specifically to work on something which became codenamed 'Zen', since AMD was getting pounded in their previous ventures(Bulldozer et al). I was excited then, then forgot, and then Ryzen was born.
- eatbitseveryday 9y agoWhy aren't Intel Xeon E7 processors considered in these benchmarks? Clearly they are "high-end" and carry a $5k+ price tag to show for it.
- cmrx64 9y agothey are user-contributed.
- rocky1138 9y agoThe top CPU is almost 4 times faster than my CPU (i5 6600K). Is it just down to the GHz and number of cores? Besides those two elements, what makes this processor so much faster?
- axaxs 9y agoIt's really neither of those things. Example: see Mediatek. They throw a billion cores and good cpu speed at a problem to no real results. Also see, AMD of the last 5ish years. IPC is where the speed comes from, in conjunction with cores/hyperthreading. I won't attempt to explain it as I'm not qualified, but it's interesting reading.
- Brockenstein 9y agoWell yeah, they're all part of the whole. Neither thing is worth a damn without the other. A 1mhz single core CPU with 1000 IPC isn't going to come close to a high end CPU in 2017. Having a decent IPC coupled with a decent clock-speed, a decent architecture, and a good number of cores will get you a solid CPU. Ideally you have great numbers in every category that contributes to performance. But if your numbers are crummy in a category, poor IPC, low cores or poor clock speed that's going to have some serious performance implications unless your competition is also suffering from some sort of deficiency that keeps you on par. Ryzen addresses a lot of Bulldozer's weaknesses. But even Bulldozer was fairly competitive in highly threaded workloads compared to an i5. Having a truly competitive CPU with a high number of cores is just a different ballgame. And it's forced an Intel response. We've had about 7-8 years of the Core i series, and now suddenly when Ryzen appears we get an i9 with a serious number of cores. Because with multithreaded applications Intel's best 4 or 6 core i7's can't compete. Their per core performance isn't nearly superior enough to outperform 2-4x as many cores on a Ryzen CPU.
- bassman9000 9y agoL1/2/3 cache sizes, better instruction sets, improved instruction pipelines, more and larger registries... Some of the improvements come from having better SW too (e.g. better/more modern compilers). EDIT: grammar
- convery 9y agoSo, based on that performance test then AMD should have the same performance as my two E5-2660v2 at half the price. That's pretty impressive.
- O5vYtytb 9y agoWonder what Epyc will look like.
- fulafel 9y agoWhat makes PassMark a representative CPU benchmark? These one-company CPU benchmarks tend to be quite problematic (cf. GeekBench). SPEC just came out with CPU2017. In SPEC there's at least a bunch of peer review, transparency and attention from academics. Here are Anandtech's AMD vs Intel CPU2006 numbers: http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/14 http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-... http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/15 http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-...
- gcp 9y agoSPEC is very useful, but it's harder to get results, much less people who have access to the suite, and there is a strong influence of the compiler (SPEC2017 will tone this down early on, but there's almost no results out). I believe AMD released AOCC partially to get a competitive position for SPEC. (Remember the controversial slides with the non-reportable gcc 6.3 results) Don't get me wrong, I would love CPU2017 results for TR vs 7820X etc.
- onli 9y ago> What makes PassMark a representative CPU benchmark? Sadly nothing. PassMark is infamous for being not representative, and you can easily check that by looking at the position of the FX processors, which are rated way too high. I run a pc hardware builder with a recommender function at its core, which means that the real core is the meta benchmark powering the recommender. It works by taking professional reviews of real life workloads, and I think it is a lot more realistic, though absolute ordering is of course not that easy. I recently made the benchmark results visible, there is one for games: https://www.pc-kombo.com/benchmark/games/cpu https://www.pc-kombo.com/benchmark/games/cpu and one for applications: https://www.pc-kombo.com/benchmark/apps/cpu https://www.pc-kombo.com/benchmark/apps/cpu . Threadripper is in there and the big model leads the app benchmark, but I still need to add more benchmarks with the threadripper processors to make the result more certain. At the beginning I even started with the passmark results, but had to move to better data when user complained the results were not good. They were right.
- Brockenstein 9y ago
- cafxx 9y agoThat explains the name. They ripped Intel a new one.
- ldev1 9y agoWell, I use CPU for games (do not compile C++, I'm using C# and javaScript mainly). And AMD is not anywhere close to Intel for me.
- 0xbear 9y agoIn case someone from AMD is reading this: guys, you need to fix stability issues on Linux, or your Epyc is DOA. People are having some serious lock-up trouble with Ryzen, even with the latest AGESA updates and kernels. This is the only issue preventing me from recommending to purchase threadripper workstations at work. We do need that pcie bandwidth for GPU, but we absolutely can't tolerate instability.
- foepys 9y agoEpyc has another stepping than Ryzen and, as far as I know, nobody could reproduce the "kill-ryzen" SEGFAULT on Epyc or Threadripper.
- 0xbear 9y agoI sure hope so, but given the epic fail of Ryzen, I'm going to wait for others to beta test it for me, and/or for some sort of explanation and fix from AMD.
- letsgetphysITal 9y agoWell then you likely won't get the information you need as quickly as you need it. Buy one, test it in your environment, file bug reports if you come across any. You don't need to replace your entire fleet in one go; Plan a phased rollout, starting with one workstation running the most aggressive of your workloads.
- randallsquared 9y agoI had a lot of issues: obviously Ubuntu 17.04 doesn't boot, but Fedora 26 crashed a lot seemingly no matter what I changed. Windows 10 also crashed fairly often. I was about say that it wasn't as bad as F26, but given the automatic restart on crash that I kept seeing, it might have been as bad. Currently using Arch/4.12-zen/nvidia with Gigabyte Gaming K5/Ryzen 7 1700X/GTX1060 and since kicking nouveau out, I have had no crashes at all. I am hopeful that this means my Ryzen woes are over.
- cosmolev 9y agoSingle Thread Performance https://www.cpubenchmark.net/singleThread.html https://www.cpubenchmark.net/singleThread.html
- akerro 9y agoI would like to point out that even in some benchmarks AMD is not on the top, it doesn't mean it's no the best buy product. There are other factors like heat production and power consumption, which Ryzen has even 3 times lower than equivalent Intel CPU.
- jrs95 9y agoSingle core performance on the i9 is still going to be significantly higher though, so it depends on your use case. Testing I've seen so far has had the i9 getting about 20% higher framerates in games, for example.
- jrs95 9y agoThere are also more i9 chips coming out this month and next. The $1200 12 core 7920X will probably have roughly the same multicore performance as the TR 1950X, but while maintaining significantly better single core performance. So, in my opinion, if you're looking for an affordable badass workstation chip, but you also want top of the line game performance, wait until the end of August and get an i9 7920X.
- arca_vorago 9y agoOne thing I think is that AMD has even more room to shine as software gets better at parallelization. I saw this trend back during my days managing 250gb+ data generation/day at a genetics company, and eventually got to build a 4 cpu, 64 core AMD Opteron system for physics computation. I am super-excited about the new server line of CPU's, because the Opteron line wasn't perfect and I expect they learned a lot from it. Also, I only need to get 2 cpu's to get that 64 core count again! I dream about supermicro or someone doing a 4 cpu board for the new line... 128 cores... (hey, I can dream!)
- rbanffy 9y agoI'd love to see how POWER8/9, SPARC64 XII and SPARC M7 stack up. If someone could throw in some z14 PU benchmarks, I'd be more than happy. Are AMD's server-grade EPYC parts available? (edit: I get it. These are mostly desktop processors with some low-end server parts thrown in. It's not a comprehensive high-end CPU benchmark, as it misses the whole E7 family)
- sp332 9y agoWhy is the "AMD Ryzen 7 PRO 1700X" testing significantly higher than the 1800X? And why don't any of the AMD chips have clock speeds listed?
- hbogert 9y agoAmazed by the little difference between the following two Ryzens and their scores: - Ryzen 5 1600x - Passmark: 13130; single: 1942 - Ryzen 7 1700 - Passmark: 13819; single: 1760 Why would you go for octa-core which is only marginally faster in multi-core workloads, but considerably slower in single-threaded workloads. On top of that, it's 25% more expensive.
- lostmsu 9y agoWhat this review does not mention in its price tag - is the cost of the cooling system. AMD claims liquid cooling is required for Threadripper.