Meh, back when Skylake-X was detailed I called the design decision of heavily investing into wider AVX misguided and and completely unsuitable for consumer processors, and I still stand by it. I'd rather AMD revives its heterogeneous computing initiative and find a way to transparently uses GPGPU for wider AVX SIMD instructions.
For comparisons, we are talking about 28% higher IPC at best while its cores (incl. L3 cache) are over 64% bigger, both at 14nm where Intel still has a significant advantage.
Wasn't Skylake 12-13% ahead(excl. 256b) in your Ryzen Strictly technical post? Doesn't seem that surprising IMO that SKL-X is slightly ahead. It is the best performer in most applications except games and a couple of latency sensitive tasks.According to my latest test, Skylake-X has 14.6 - 28% higher IPC than Ryzen (post Spectre & Meltdown fixes).
Wider AVX allows Intel to try and entice AI and ML guys into writing code for the CPU, and let's face it, there far more people who are are able to write x86 code than CUDA. Of course AMD wants their GPUs to do the same stuff so I expect AMD to not be as aggressive in pushing wide-AVX in their CPUs. Let's see what they do with Zen 2.Meh, back when Skylake-X was detailed I called the design decision of heavily investing into wider AVX misguided and completely unsuitable for consumer processors, and I still stand by it. I'd rather AMD revives its heterogeneous computing initiative and find a way to transparently uses GPGPU for wider AVX SIMD instructions.
For comparisons, we are talking about 28% higher IPC at best while its cores (incl. L3 cache) are over 64% bigger, both at 14nm where Intel still has a significant advantage.
Sure, but in perf/watt it will always fight a losing battle against (semi-)dedicated accelerators. The mixed code scenario is the only thing it has going for it, and that's with the negative caveat of throttling the actual CPU whenever in actual use. The consumer oriented i9 series already showcased the extend of the issue of having to handle what's essentially a different platform with completely different thermal requirements.Wider AVX allows Intel to try and entice AI and ML guys into writing code for the CPU, and let's face it, there far more people who are are able to write x86 code than CUDA.
Meh, back when Skylake-X was detailed I called the design decision of heavily investing into wider AVX misguided and completely unsuitable for consumer processors, and I still stand by it. I'd rather AMD revives its heterogeneous computing initiative and find a way to transparently uses GPGPU for wider AVX SIMD instructions.
For comparisons, we are talking about 28% higher IPC at best while its cores (incl. L3 cache) are over 64% bigger, both at 14nm where Intel still has a significant advantage.
Another leak showing only a modest improvement in gaming. Overall 5% improvement will make it still in the neighborhood of i5 8400 at sub 200.00. I mean you can look at the high end of improvement and tout the 14%, but OTOH, many games show no improvement at all.Anyway, what is interesting in this graph is that Ryzen 7 2700X see the biggest improvement in games that Ryzen 7 1800X does the poorest (compare Intel's counterpart).
I am assuming that this is from the higher boost clock.
What is most notable about these games (that Ryzen 7 2700X improved upon) is that they are all poorly threaded.
So this means that while the average FPS increase is only 5%, this is quite misleading and that users would see big improvements (up to 14%) in games that perform poorly on Ryzen 7 2700X.
Unless GPU bound i find this really hard to believe, simply because the 2/3/4/5 core turbo is much higher then the 1800X. Add that many games see good improvements from improved memory latency, i would be very skeptical of 0% improvement. I doesn't make sense, gaming was the one area where Zen was weakest, the boost and latency changes should have biggest impact to gaming. Only games that can max 16 threads would be at risk of not seeing large improvements, but those already performed well on Ryzen. But oddly to that statement AOTS is supposed to see big gains so that must all be down to better latency.......Another leak showing only a modest improvement in gaming. Overall 5% improvement will make it still in the neighborhood of i5 8400 at sub 200.00. I mean you can look at the high end of improvement and tout the 14%, but OTOH, many games show no improvement at all.
Because most of those are GPU bound.Another leak showing only a modest improvement in gaming. Overall 5% improvement will make it still in the neighborhood of i5 8400 at sub 200.00. I mean you can look at the high end of improvement and tout the 14%, but OTOH, many games show no improvement at all.
My opinion is that OpenCL on GPUs is more suited for ML. Intel is really good with their OpenCL and clDNN.Wider AVX allows Intel to try and entice AI and ML guys into writing code for the CPU, and let's face it, there far more people who are are able to write x86 code than CUDA. Of course AMD wants their GPUs to do the same stuff so I expect AMD to not be as aggressive in pushing wide-AVX in their CPUs. Let's see what they do with Zen 2.
Wasn't Skylake 12-13% ahead(excl. 256b) in your Ryzen Strictly technical post? Doesn't seem that surprising IMO that SKL-X is slightly ahead. It is the best performer in most applications except games and a couple of latency sensitive tasks.
Seeing the listing of geil sniper x ram, we can confirm that PR can clock ram much higher than RR, can't wait to see the results with 4000 ram, and how much ipc gain using much faster ram.
Aida 64
if CL19 DDR4 4000MT/s.
~55,6ns
Or even better stuff if 3600CL16
~ 59ns
Yeah and with that latency then PR will have latency parity with intel Skylake, maybe in games it will be like 3~5 %
While official 3400 MHz support is a nice improvement over Ryzen 1xxx (3200 MHz), where do you get so low latencies from?
Users, that can try do as I did, put 4-3,8GHz on i7 8700K and put NB to 1,6GHz and you will be amazed by tech from AMD.
Ryzen 1K-series officially supports up to 2666MHz (1 DPC SR) and 2K-series is supposed to officially support 2933MHz.
Why would you do that? AMD FClk and Intel LLC/ring frequency don't even remotely do the same thing.
Does it really matter?
If high speed memory support is good then PR will automatically become very competitive in gaming.
AMD Zen L3 cache operates same as core frequency. Intel Nehalem uncore, and Haswell and later LLC/ring frequency is decoupled from core frequency.To try out stuff how they behave under different settings and compare "core"
to get 3.7 GHz core, 1.6 GHz LLC/ring, and compare to AMD Zen 3.7 GHz core and L3 cache, that test can't be called "real core IPC benchmarks".Users, that can try do as I did, put 4-3,8GHz on i7 8700K and put NB to 1,6GHz and you will be amazed by tech from AMD.
Holy Hell! , different between 3200 & 3200 LL Is huge.Nope.
And nope.
High speed memory won't automatically make Ryzen better in gaming.
The improvements from higher than 3200MHz MEMCLK are already diminishing.
At or above 3200MHz lower latency matters much more.
AMD Zen L3 cache operates same as core frequency. Intel Nehalem uncore, and Haswell and later LLC/ring frequency is decoupled from core frequency.
So, if one does:
to get 3.7 GHz core, 1.6 GHz LLC/ring, and compare to AMD Zen 3.7 GHz core and L3 cache, that test can't be called "real core IPC benchmarks".