Home / Component / CPU / AMD Zen engineering sample benchmark leaks

AMD Zen engineering sample benchmark leaks

So far, the only real performance numbers we've received for Zen have come from AMD's marketing and presentation slides. However, this week we may have got our first look at legitimate real-world performance thanks to the Ashes of Singularity benchmark database, in  which a benchmark for one of AMD's Zen engineering samples was uncovered.

This follows from last month, when the specs for several AMD Zen engineering samples made their way online. The Zen chip benchmarked here is identified as ‘1D2801A2M88E4_32/28_N', which is a 2.8GHz base/3.2GHz turbo clocked CPU. Engineering sample clocks are usually not final, so the final product could end up having a speed bump. This is the 8 core, 16 thread version of the Summit Ridge CPU, which will likely be the first one to market.

standard-1080p-zen-1d

The screenshot above shows that the Ashes of Singularity benchmark was running at the standard graphical preset at 1080p, with an RX 480. Here, the Zen engineering sample scored 58 frames per second on average, as shown by the CPU Framerate bar on the graph.

As wccftech points out, this is higher than an Intel core i5 4670k, which scores 52.6 frames per second. It is also a significant leap over the FX-8350, which scored 42 frames per second. It does lag behind the Intel Core i7 4790 though, which manages 65.4 frames per second in this benchmark. However, that could change between now and release.

So this appears to be our first real-world taste of what Zen is capable of. However, it might be a while before we can get our hands on the CPU as a recent AMD roadmap leak points to Zen arriving in 2017.

Discuss on our Facebook page, HERE.

KitGuru Says: These are early numbers but they are still very interesting to see. Hopefully we will get more information like this in the build up to Zen's eventual launch. 

Become a Patron!

Check Also

Intel’s x86S initiative has been abandoned

Intel has officially abandoned its plans for its own-developed x86S specification, a streamlined version of …

62 comments

  1. I did hear quite a while ago that Zen was around Haswell performance. Maybe it’s true./….

  2. Looks possible, if they get the stock clock up to 3.4Ghz then it should be a great performer. Also we have to see how it performs against the haswell-E series on multi threaded apps. I’m glad we finally have choice in CPU. Lets hope the price is 25% less than Intel.

  3. Kind of sad to be honest we only expect AMD to give us what Nvidia/Intel gave us 2 years ago for both GPU and CPU.

  4. The screenshot clearly says it has 8 physical cores, so we *do* know that it has 8 cores.

  5. Hmmm, at first I thought the 2D parts were quad, but it appears they may be older samples… RAMBUS just went final on the DDR4 PHY and they can get two more spins tested internally by NovDec which implies they will get another 20-30% including clockspeed, final chipsets and final mobos…

  6. The leaks were using RX480, so expect GPU limitations. Tis why Wccftech uses CPU framerate which is mainly CPU benchmark in their graphs. Any stock i7 5960X trounces this ES easily. If this ES is really 8c/16t as shown in the screenshot and runnin 2.8ghz on all cores then Zen is in dire trouble. The diff between 2.8ghz and 3ghz is only 7%, and yet i7 5960X CPU framerate scores are nearly double http://www.ashesofthesingularity.com/metaverse#/personas/5c2738bf-7190-40e4-b562-69f70e584ac1/match-details/2b1735bd-c511-4bc7-a933-c0fc89ea9d05 That CPU framerate score differs very little between different GPUs.

  7. It seems a bit disingenuous to compare what is expected to be a $200 Zen CPU with a $1,000 CPU (the VAST majority of PC users and OEMs won’t spend over $400 for a CPU). A better comparison would be against an i7-6700k which is retailing at $350 on Newegg. Even assuming a $250 price tag on Zen (which is a high estimate even including the slightly pricier Motherboards) you’re looking at a 28.5% difference in price. Which means that if Zen can reach within 25% of i7 performance (for single-thread, it seems like it would exceed in multi-thread). Then Zen is a solid product. Additionally, given that this is the first year in a long time that AMD will be on an equal manufacturing process as Intel, and that Intel has hit a wall of sorts in performance, it is far from impossible for AMD to accomplish.

  8. Non-K Skylake i7 6700+RX480 is also much faster http://www.ashesofthesingularity.com/metaverse#/personas/b70de3e3-9c89-43d6-8c94-d6d6ccf7f379/match-details/1c09302c-ba69-4989-8693-c7aa26111d80 CPU framerate 81.9fps. Like worlds apart….. 81.9fps is over 41% faster, and thats just 4c/8t only….. Intel 14nm is still superior and not equal to GF 14nm….. Notice that GF only has 14nm LPP…. low power process typically for mobile chips, so expect low frequency….. Also look at RX480 power draw and frequency limits…..

  9. The 6600K, according to these benches, will destroy ZEN.

    It’s less than $250.

  10. A Intel 6600K CPU is on par with the i7 4790 that obliterated the ZEN 8 core 16 thread ZEN. It costs as low as $220.

  11. Post doom and gloom elsewhere… Mobos aren’t finalized and neither is the DDR4 IMC… We can expect another 20% or so in Dec…

  12. Engr samples are not final… Useless…

  13. James Stephen Edge

    The Zen in this benchmark is an engineering model, the final version will have 50-100% more performance which means they should trade blows with more expensive Intel CPUs.

  14. This is a GPU limited benchmark and must be taken with a grain of salt. As such, it is limited to four threads and zen’s ability to scale to 16 threads is negligible in such a setting. Keeping these constraints in mind; the performance is impressive!

  15. CPU framerate isn’t limited by GPU. This is framerate without using rendering, only game engine tasks (AI, physics, scripts etc), so it is CPU benchmark.

  16. The variant of the AOS benchmark used here is to test GPU performance and not CPU. Ergo, it is a GPU limited benchmark.

  17. I think it looks good. Keep in mind that these were run using AOS GPU benchmarks and does not assess multiscore scaling across 16 threads. It is limited to performance across four threads. That being said and being that this is an engineering sample…that is impressive! I would not jump to conclusions and wait for proper tests done in a more controlled setting with an optimal mobo and memory.

  18. Is GPU limitation only make the CPU result closer to each other? GPU-bottleneck only makes the CPUs waste time idling so even as low as Athlon X4 can compete with i7 haswell on GPU-bottlenecked system. So i believe that’s not the case here.

  19. Till we see some cpu only benchmark’s of final products its all debatable

  20. actually, the 5960x turbos to 3.5Ghz
    it has lots of cache and running in quad channel memory mode
    we don’t know much about the system Summit Ridge was being benched on
    Summit Ridge will compete against Intel’s mainstream CPU die, which 95% of consumers use
    Intel’s mainstream CPU is maxed out at 4C/8T

  21. Michael Kreitzer

    To be fair Intel is hasn’t done any better. Single threaded performance from Intel has been more or less flat since Haswell with only Skylake offering a bit of a bump for some applications at the cost of I/O throughput.

  22. Turbos are for single core only, when all cores are used then base 3ghz kicks in.

  23. CPU frame rate is CPU benchmark so its not GPU limited. See….
    http://i.imgur.com/VIEnqqd.png

  24. When tis came out http://wccftech.com/amd-bulldozer-essample-leaked-benchmarked-tested-asus-sabertooth-990fx-am3/ AMD sheeple could not believe it….. Final release with higher frequency could not help Faildozer also….

  25. If mobo makers can’t be finalize their mobos then Zen is dorked. Maybe tis why we don’t see any mobos shown from big mobo makers close to ‘launch’? MSI have shown Kabylake mobo already, so where are the AM4 mobos?

  26. Shut up… Zen is not Bulldozer and Intel can’t play FMAC games this time… A SAMPLE running right behind Haswell at 1GHz less is no joke…

  27. They couldn’t make them until the socket and chip were done MORON… This is a totally new arch that has to FULLY share sockets with CPUs and APUs…

  28. That’s not true. The full load turbo (100% load on all cores and threads) of the 5960x is 3.3Ghz. So it was running at a minimum of 3.3Ghz unless thermal- or power-throttled.
    Even though a lot of cores are used in Ashes of the Singularity, it will by far not fully load 16 threads. According to reviews, having more than 8 threads is completely useless on Ashes of the Singularity.

  29. So you’re tellin me Zen is not complete? Do I smell more delays comin up? So that OEM AM4 mobo with A12 Bristol Ridge leaks are not true? C’mon….. Stop spurtin nonsense…..

  30. Its lookin more like another Faildozer with this leak…. Similarity, 8 cores still can’t beat 4c/8t Intel chip….. AMD’s becomin quite a joke in the CPU circles… Reminds me of this, Intel beatin AMD despite havin half the number of cores….
    https://www.youtube.com/watch?v=mRtFNxNShts

  31. That’s lyin… 5960X goes to 3ghz on all cores, only the 5820K goes to 3.3ghz on all cores…. Also try explainin tis… used up to all 10c/20t also….

    http://core0.staticworld.net/images/article/2016/05/broadwell_e_core_i7_6950x_ashes_of_the_singularity_beta_cpu_benchmark-100663335-orig.png

  32. Well, it hasn’t been released yet and is 11% away from i7 while being almost 1GHz lower clock… You’re the joke…

  33. Kaby Lake will launch months before Summit Ridge.
    AMD has always said “late 2016 launch” from the very beginning and already in 2015 earning calls said to not expect any relevant revenue from Summit Ridge in 2016, but instead focus on 2017 for revenue. This has pointed to a end of the year launch in low volume.

    So if they launch in December 2016, it’s not delayed at all. If they launch in January 2017, it’s a minor delay. Later than that would be a significant delay, but still no signs of that happening.

  34. And yet AMD has not given any dates or month, with just few months away from launch… just vague timelines like end of Q4 or startin in 2017…. Also Kabylake ES has been out for months already, so that is expected….. But Zen ES has been very late….

  35. Look at 5960X result http://www.ashesofthesingularity.com/metaverse#/personas/5c2738bf-7190-40e4-b562-69f70e584ac1/match-details/2b1735bd-c511-4bc7-a933-c0fc89ea9d05 CPU framerate 92.7fps, almost double…. And thats 3ghz on all cores. From 2.8ghz to 3ghz is only 7%, but from 58fps to 92.7fps is 60%……

  36. I’m not “lying” if anything, I would be misinformed. But I’m not, I just double checked. 5960x has 4 turbo stages. Full load turbo is 3.3Ghz, as I already explained in my previous reply.

    What exactly do you want me to explain in this chart?
    Let’s start from the bottom:

    6 cores (i7-4690x) are faster than the 4 cores (i7-6700k) because the game takes advantage of more than 4 threads and hyperthreading is never as fast as a real additional core. Also, the massive amounts of additional cache should give quite the boost in performance.

    The additional 2 cores of the 5960x help a bit, but scaling is already decreasing significantly at this point (as I said, it scales to a maximum of 8 threads). Also, the Haswell cores are quite a bit faster than the Ivy Bridge ones, but this is offset by the lower clock speed.

    Finally, the i7-6950x is an even faster core (Broadwell) and also clocked significantly higher than the 5960x (up to 4Ghz).

    If you want to talk about thread scaling, you should look at some thread scaling reviews, not some random 4 completely CPUs.

    For example this, which also clearly shows the 8 thread limit:
    https://cms-images.idgesg.net/images/article/2016/02/dx12_cpu_ashes_of_the_singularity_beta_2_average_cpu_frame_rate_high_quality_19x10-100647718-orig.png

  37. We’ll see in DecJan… I just want something faster than my 8370… That plays every game @ 1200p with Ultra or high… The first leak is 35% faster than that… Final mobos, IMC, and RAM clock will probably be good for 20%+…

    You must be like 15…

  38. Yes, but Kaby Lake should launch like in September or so.
    It was never expected for Zen to launch in September already.
    No signs pointing to big delays of Zen yet. Wouldn’t be surprised if AMD pushed it out to 2017 though.

  39. Don’t forget Skylake and soon Kabylake. Those are even faster than 4c/8t Haswell at this bench. Not forgettin Broadwell-E also. Remember that this CPU frame rate is mainly CPU bench…. Just like Cinebench, AMD FX 8 core was faster than Core i5 also. The big diff is Zen will have its own FPU in each real core, unlike the AMD FX fake cores. So that increase is not surprisin, but still losin to 4c/8t i7 is simply alarmin…. And yer sounding deluded from reality, when I shown yer all the results to compare. Deja vu all over again when Zen finally comes out, by then from delays it may already be outdated.

  40. Lookit http://www.pcworld.com/article/3075433/hardware/intel-broadwell-e-core-i7-6950x-review-the-first-10-core-enthusiast-cpu-is-a-monster.html?page=2 says ‘Even better, Oxide provided us with a beta version of Ashes of the Singularity that adds a new mode specifically to test CPU performance, rather than GPU performance’
    …. Turbos and boost are for single cores or a few cores within TDP, but does not kick in when ALL CORES are used. And wheres that 3.3ghz? http://ark.intel.com/products/82930/Intel-Core-i7-5960X-Processor-Extreme-Edition-20M-Cache-up-to-3_50-GHz In every review, it says 3ghz like tis one…….
    http://www.legitreviews.com/wp-content/uploads/2014/08/5960x-cpu-features.jpg

  41. Way into 2017 is more like it, judgin by new leaked roadmaps with no specific time. Notice also AMD been rather quiet about Zen, only speaks bout them rarely usually at investor conferences or earnins time….. Very unlike previous CPUs before they launch, besides Faildozer…..

  42. There are full load turbos. AMD specifies them exactly and Intel likes to keep their customers uninformed, but that doesn’t mean they don’t exist.
    Turbo is mentioned in several reviews like:

    http://www.tomshardware.com/reviews/overclocking-retail-intel-core-i7-5960x-cpu,4237.html
    “we had little trouble taking all five to 4.5GHz (a 1.2GHz overclock over the chip’s 3.3GHz stock Turbo Boost setting with eight cores active)”

    http://www.legitreviews.com/intel-core-i7-5960x-8-core-haswell-e-processor-review_149560
    “8 physical cores with 16 total threads, a clock speed of 3.0GHz at base and 3.3GHz at Turbo”

    Or even better, here’s a whole article about the issue with all the full load turbos.
    https://www.pugetsystems.com/blog/2015/07/09/Actual-CPU-Speeds—What-You-See-Is-Not-Always-What-You-Get-675/

  43. And don’t forget that there are two more Zen versions AT lest in the works now and REALLY don’t forget that AMD caught up at 15W with lowly Bulldozer… That implies they can do the same 40% increase as with Kaveri to Bristol Ridge..

    Intel is barely getting 5% increases anymore…

  44. Wait for them forever if you want….. Current mobile AMD chips are simply awful, even them 15W ones. Lookit http://www.notebookcheck.net/Bristol-Ridge-in-Review-AMDs-A10-9600P-Against-the-Competition.168477.0.html ‘With a 41 Wh battery, the Pavilion 15 is able to run a normalized WLAN test for 3.5 hours. This is significantly better than the Carrizo model, but it falls short of similar Skylake models’ and more there……

  45. They are similar to those used for laptop CPUs, and don’t last long and comes in short burst sometimes whenever TDP allows…. ‘The amount it goes up depends on how many of the total number of cores are active, as well as temperature; if the CPU gets too hot, it will back off toward the base clock speed to keep from overheating’…. So, when all cores are on full load and heat builds up, then can expect 3ghz base frequency…..

  46. You are kind of pathetic… Every site a few months ago posted how OEMs were gimping Carrizo… But they all showed 7h+ battery life…

  47. You shared an article from PCworld, unfortunately this now makes anything you say obsolete.

  48. Yes, like I said initially. If power or thermal throttled, it will downclock. This usually isn’t the case in desktop environments though, unless it’s full of dust or high ambient temperature etc.

    AotS will not load such a CPU anywhere near 100%. I don’t know the exact numbers, but definitely less than 50% average. So the relevant clocks should be anywhere between 3.3Ghz and 3.5Ghz

  49. How on earth do you land on that 200 dollar figure?
    Me I’m guessing 450-500 USD for a 8 core Zen (chip only, no motherboard) making it compete with a Intel 6 core in price.

  50. Did you even check the link given? That’s review is about the latest Bristol Ridge, not Carrizo…. What gimpin? Read the review, here’s a sample ‘However, the test APU cannot perform at the upper ends of this range continuously due to the limiting TDP value. This behavior exhibited in the Carrizo models as well. The APU often starts the benchmarks at a high clock speed. Sadly, these high clock speeds are reached by surpassing the 15-watt TDP limit, which means that after a short period of time (around a minute), the performance drops by up to 20%.’….. They also tested it with dual channel RAM…. Lookit….

  51. http://seekingalpha.com/article/3998662-amd-zen-dead-arrival?li_source=LI&li_medium=liftigniter-widget

  52. if this is real, then zen is good as promised.

  53. These guys just don’t get it. I remember when Bulldozer engineering benchmark samples were leaked and they showed them trailing INTEL chips. AMD fanboys got on every tech blog and shouted that it wasn’t fair to assume these leaked benchmarks are the final product. They thought of every explanation they could to excuse away the terrible scores from AMD. In the end Bulldozer turned out to be AMD’s long slow undoing. ZEN is finishing them off. Granted AMD did not have NEAR the R&D Budget that intel had! There was no way for them to truely compete. AMD is DEAD and the markets will soon devour what is left of them once the official benchmarks come out. Sorry guys, INTEL WON! No more competition here in the DESKTOP CPU MARKET.

  54. This bench is fake. Where can Mr.Feng (or somebody else) get a mobo fully operational for Zen? There are nothing around yet. It is a fake.

  55. Let’s stop the speculation. Clearly this benchmark has a strange smell. First and foremost, where can Mr.Feng (or somebody else) get a mobo fully operational for Zen? There are nothing around yet. It is a fake.

  56. Judgin by his name, possibile from from far east…. places like Taiwan or China, where most mobo makers are located….. GG

  57. Aye mate! Zen starting to look like a stinking pile of turd.

  58. As long as they bring something to the table that has I5 performance for 120-150$ they’re guaranteed to get a large portion of the market. Make it unlocked and be sure to give it atleast 4-cores (Not threads). Possibly an 4core/8Thread model for 200$ that competes with/trails slightly behind the I5 in gaming but excels in tasks such as Rendering and streaming and this insane obnoxious 8core/16Thread model for 350$ to compete with the I7 and beat it in rendering/recording or market it for it’s multitasking.

    We’ll see how these predictions hold up :p. Just hoping they won’t forget about the I only have 500$ to spend on a new PC and would prefer something beefier than an Pentium/I3 demographic.

  59. the final version will have 50-100% Wut, where did you get those numbers, because everybody was saying the same thing during bulldozer at that time the final model was quite similar to the engineering models. Here’s some leaked info on the FX-6100, which ended up being only like .2 Ghz faster when it came out. http://wccftech.com/amd-bulldozer-zambezifx-benchmarks-leaked/

    Just so you don’t wait like a year for it like I did back in 2011

  60. Yeah fuck knows where people are getting 200 USD from..

    AMD actually need to move their asses and make these chips. Haswell IPC was good when there was only haswell.. Now three more intel gens are here even after huge 14nm delays come on AMD dont let it be obsolete before its even out…

  61. I call BS, I just went and looked at two other websites and the 5960X @ 3.5 GHz turbo core and an RX 480 only hit 54.3 fps with the same settings.
    That’s why Zen overclocked @ 3.0 GHZ beat the Intel Core i7 6900K down clocked to 3.0 GHZ in the benchmark they ran at the press conference. This writer always tries to find ways to bash AMD on their CPUs and cards.

  62. Well, remember that Intels high end offering is always one generation behind their mainstream offering in terms of architecture.
    And while we got things pointing in a lot of different directions with Zen as far as its performance is concerned I’m guessing something just at the heels of Broadwell on average, beating it in a few workloads, losing out in others in terms of single threaded performance but throwing in a couple of extra cores for the price.

    Anyway, Haswell is *still* a good CPU that’s more then fast enough for anything you can throw at it these days, and this will be kind of like buying a discounted old xeon or some such in that you’re getting more multi-threaded performance for your money but might be a tiny bit behind in single threaded performance compared to the latest and greatest from Intel.
    But honestly even if it was worse then Haswell in single threaded performance it could still be a success.

    Some people who do have a clue about computers are *still* buying piledriver based processors even now in 2016.