I wouldn't be surprised if that's already been planned in the future. GPUs would benefit from such a design far better than CPUs.
I'm ready , bring it on. AMD finally will catch up to Intel ipc with this cpu , i have heard from my sources in the industry ( sorry cannot mention names ) , that this cpus in fact will overclock like crazy and they run cool as a cucumber. Guys i will tell you something i have been running , testing , overclocking , benchmarking , running games on my i9 9900K versus my Golden Ryzen 2700x overclock at 4.4Ghz 24/7 with memory at DDR4-3600Mhz cas14 and the difference clock for clock between this two cpus is negligible , absolutely a waste of money on the i9 9900K , i bought it only to run benchmarks for hwbot points , soon to run on liquid nitrogen , where the Intel cpu scores better because of the overhead in the overclocking department , i9900K runs hotter than my Gem 2700x at the same clocks :/ , really it is that hot. Bare in mind AMD SMT is miles ahead of Intel HT. Testing both systems side by side at the same time ( 2700x vs 9900k ) real hands on experience i can tell you an honest answer , i know for a fact once the new Ryzen 3700x are release is game over for the costly , very expensive , hard to find 9900K ! Heck , the 2700x is so close in performance it makes the 9900k a very hard buy at this price point.
On that news it really looks like my new build for 2019 will be AMD? For sure the glamor on intel's 9 series is beginning evaporate or has!
Ah, ok, so in that case we'll say a general IPC of +15%, combined with the +12% increased clockspeed, this gives 1.15*1.12 = 29% increase in general performance vs current gen Ryzen. Well, 29% increase in per core performance vs current gen Ryzen, that's still a big increase. I think that puts it above Intel for per core performance. See here (https://www.guru3d.com/articles_pages/battlefield_v_pc_performance_benchmarks,4.html) the difference between 2700X and 9900K at 1080p in BFV is 'only' a 16% lead for the 9900K. We can already see that if Zen 2 is gonna be 29% higher per core performance, then this is gonna even outperform the 9900K in gaming - cool! EDIT: as a side note, I think some games do actually make use of AVX. For instance, in BF1 my CPU will kick up to AVX style increased voltage level, so I think BF1 uses AVX to name just one game. And it was AVX that you say AMD have specifically optimised for when it comes to their IPC increase. Perhaps it's possible that this means gaming IPC related performance gains could be more like the 29% rather than the general 15% you mentioned - in which case the calculation in my previous post would hold true and gaming performance increase per core could be closer to the 43% increase, rather than the 29% determined in the previous paragraph. (Sorry for the all the %'ages and numbers, hard for me to keep track of what I'm talking about, let alone anyone else!)
Not an expert in assembler but physics engines could use (and they probably already do) AVX instructions.
AMD stated about 25% increase of performance over prev. generation - so 29% higher IPC doesn't sound logical to me. Zen+ over Zen had +3% IPC (cache latency optimization mainly). Therefore +29% over Zen would mean ca. 25-26% more IPC than Zen+, which would mean absolutely no clock increase over last generation (12nm -> 7nm). No way! Last legit rumors before the official announcement mentioned engineering samples (unstable) of Zen2 tested on 4Ghz base clock with 4.5 Ghz turbo having avg. IPC increase of ca. 13%. Gaming perf. beeing on par with that of i7 8700 "already". They usually need another 1 or 2 steppings to sort out leakage problems, which result in a more stable and in avg. 0.2-0.3Ghz higher clocked operation (but no IPC increase per se). These are my assumptions of course, but that "news" (+29% instead of more likely +13-15% IPC) just gives false hopes for a lot of people as i see it! I'm pretty sure Zen2 will be great, - and i'm going to buy one next year - but not THAT great...
Hard to say, since they were on about "last Gen" and they are focusing on the data centre which had no Zen+ i would say 25% but then again it appears to be specific to the type of tasks it's doing that the gain maybe 29% but since the original estimates was 13% as a worst case, i guess they are stating it averages to around 25% and there maybe only a limited amount of tasks where it may fall to "only" 13% gain. I am personally sticking to 13% IPC as my expectation... that way you are never really disappointed and if they exceed it all the better! For me the platform looks great it pull plop into my Crosshair VI so i can by the processor when available then wait on reviews of the best X570 motherboard and replace it. After all the X570 is coming with PCIe 4.0 so increased through put from the CPU to the chipset and device communication so a new motherboard would be in peoples best interest but at least PCI3.0 and 4.0 are backwards compatible. From cache latency there is loads of work AMD can do maybe they gone with the higher latency cache because it reduces cost and means future versions can come with lower latency and throughput as prices come down and gain some impressive speeds while being reasonably priced. So Zen is all about scalability.
So e.g a 2700X with a 13-15% increase in IPC should be around 167-170pts(CB15 G3D) vs Intel's current CL i7/i9 152-153 @3.5GHz! Nice! Can't wait to see!
I wonder if i will be able to thow a 3600 on my b350 mb msi pc mate hope i will be able to do that! Go go amd!
Remember, they delivered numbers for server chips. That means comparing old generation with new generation EPYC. So, take 32C/64T EPYC, and check its stock (non-OC) clock (performance) under full load. Then expect new generation to achieve 1.25x higher performance. EPYC 7601 is: 2.2GHz base; 2.7GHz all-cores loaded; 3.2GHz Turbo So, if we take 2.7GHz as all-cores loaded, 1.25x statement can be achieved by moving clock under full load to 3.4GHz. In other words: "For now it is uncertain what part of improved performance of EPYC comes from IPC and what comes from clock improvement."
Only 6 months to wait. That will go by in a jiffy. And I can go from B350 chipset and Ryzen5 1600 to new platform. Skipped the Zen+.
Some further details on the IPC claims: https://www.notebookcheck.net/Updat...than-Zen-1-in-certain-workloads.359556.0.html
From what I recall, AMD was vamping up the AVX instructions, so I wouldn't be surprised if that contributes toward a lot of that benchmark-specific performance measurements. Not that that's a bad thing - AVX is slowly being applied in more and more applications.
Never thought it would really be 29% IPC improvement. Realistically if they improve about 10% IPC and have higher clock, we can see that 25% performance increase, which is still very nice. On a side note though, I think that AMD should temper the expectations as it's better to be surprised on release than to be disappointed. They often go big on promises without clarifying what their numbers actually mean.