AMD RyZen CPU Architecture for 2017

[Clukos]

Apparently Threadripper (and Epyc) will be on a new stepping (B2): https://www.techpowerup.com/234476/amd-readies-b2-stepping-of-the-ryzen-summit-ridge-silicon

PCIe and IMC improvements are expected. Wondering if/when we'll see that stepping on mainstream Ryzen CPUs

Although I can't complain, my 1700 seems to have a pretty good IMC, able to run my 3600CL16 ram at 3466CL14 with very tight timings and SOC voltage at 1.11-1.13 (limit for 24/7 is rated at 1.2 according to Asus) and it is 24/7 stable (tested with the linux Google stress app test for 2 hours).

 

[Clukos]

Ryzen re-review by Guru3D, they check AGESA 1.0.0.6 and faster memory: http://www.guru3d.com/articles_pages/amd_ryzen_7_agesa_1006_performance_update_review,1.html

 

[Cyan]

an in-depth article compared all the best current CPUs in the market and awarded the Ryzen 1600 as the best CPU for gaming.

https://www.pcgamesn.com/best-cpu-for-gaming

 

[Clukos]

Intel compiler magic:

https://compubench.com/device.jsp?D...estgroup=compute&benchmark=compu15d&var=score
https://compubench.com/device.jsp?D...estgroup=compute&benchmark=compu15d&var=score

 

[Kaotik]

Intel compiler magic:

https://compubench.com/device.jsp?D=Intel(R)+Ryzen+7+1700X+Eight-Core+Processor&os=Windows&api=cl&testgroup=compute&benchmark=compu15d&var=score
https://compubench.com/device.jsp?D...estgroup=compute&benchmark=compu15d&var=score

Erm, what's going on there, faked AMD CPU as Intel one and it performs better in CompuBench?

 

[Cyan]

Intel compiler magic:

https://compubench.com/device.jsp?D=Intel(R)+Ryzen+7+1700X+Eight-Core+Processor&os=Windows&api=cl&testgroup=compute&benchmark=compu15d&var=score
https://compubench.com/device.jsp?D...estgroup=compute&benchmark=compu15d&var=score

lol, is that for real? How can you make a Ryzen CPU pass up as an Intel one? No wonder how many Intel rabid fans are out there, bringing benchmarks all over the table and infinity fabric all over the place. I wouldn't change a Ryzen CPU for an Intel one nowadays, i5 and i7 have become slightly obsolete thanks to competition, if that didn't happen we'd be still with 4C-8T for some years..

 

[Pressure]

lol, is that for real? How can you make a Ryzen CPU pass up as an Intel one? No wonder how many Intel rabid fans are out there, bringing benchmarks all over the table and infinity fabric all over the place. I wouldn't change a Ryzen CPU for an Intel one nowadays, i5 and i7 have become slightly obsolete thanks to competition, if that didn't happen we'd be still with 4C-8T for some years..

Worse, at 4C-4T ... they have been milking that Core i5 for more than 5 years.

 

[Clukos]

Erm, what's going on there, faked AMD CPU as Intel one and it performs better in CompuBench?

Seems like that's what's happening, it's spoofed via VM (probably) and the bench thinks it's an Intel processor, running better code essentially.

http://www.agner.org/optimize/blog/read.php?i=49

However, the Intel CPU dispatcher does not only check which instruction set is supported by the CPU, it also checks the vendor ID string. If the vendor string says "GenuineIntel" then it uses the optimal code path. If the CPU is not from Intel then, in most cases, it will run the slowest possible version of the code, even if the CPU is fully compatible with a better version.

 

[Kaotik]

Seems like that's what's happening, it's spoofed via VM (probably) and the bench thinks it's an Intel processor, running better code essentially.

http://www.agner.org/optimize/blog/read.php?i=49

Or is it just AMD CPU running on Intel OpenCL Driver?

edit:
It really looks like it's nothing more than Intel OpenCL-driver running on Ryzen - the bigger question is how do you get it to run Intels instead of AMDs OpenCL-driver?

 

[Pressure]

Interesting read about the CPU dispatcher.

Workarounds
At present, we don't know if or when Intel will make a new compiler and new software libraries that do not check the vendor ID string. In the meantime, here is what we can do about the problem.

• Use another compiler. In my tests, the Gnu compiler for Linux has an optimizing performance similar to the Intel compiler, but the Gnu function library (glibc) is inferior. All other compilers gave lower performance in my tests. There is no other Windows compiler with a similar performance, not even the Gnu compiler for Windows.
  
• Use the Intel software and patch the CPU dispatcher. In my C++ manual, I have provided the code for alternative CPU dispatchers for Intel's compiler and function libraries and descriptions on how to patch them into your software. This, of course, relies on undocumented details of the Intel software. This dispatcher-patch can improve performance on non-Intel processors considerably in many cases.
  
• Never trust any benchmark unless it is open source and compiled with a neutral compiler, such as Gnu or Microsoft.
  
• It is possible to change the CPUID of AMD processors by using the AMD virtualization instructions. I hope that somebody will volunteer to make a program for this purpose. This will make it easy for anybody to check if their benchmark is fair and to improve the performance of software compiled with the Intel compiler on AMD processors.

I don't own an AMD processor and have no idea what compiler the commonly used software uses. However, the results would be very interesting.

 

[Kaotik]

From another forum, there's no shenanigans going on - it's simply the Intel OpenCL-driver running on Ryzen. Apparently OpenCL-drivers designed for x86 will run on any x86 CPU

 

[ArkeoTP]

From another forum, there's no shenanigans going on - it's simply the Intel OpenCL-driver running on Ryzen. Apparently OpenCL-drivers designed for x86 will run on any x86 CPU

So, you're telling me that Intel is actually helping AMD's performance here? Dear Lord, what the heck is up with AMD's OpenCL driver?

 

[Kaotik]

So, you're telling me that Intel is actually helping AMD's performance here? Dear Lord, what the heck is up with AMD's OpenCL driver?

Most likely AMD is focusing on their GPUs more than CPUs

 

[Cyan]

Ryzen 1600 dethrones i7 7700k as best seller on Amazon

 

[Esrever]

From another forum, there's no shenanigans going on - it's simply the Intel OpenCL-driver running on Ryzen. Apparently OpenCL-drivers designed for x86 will run on any x86 CPU

OpenCL driver's quoting the CPU could be how the benchmark is biased so that doesn't really proof that there isn't any shenanigans.

 

[sebbbi]

Suprising FP results... 32-core EPYC vs 28-core Skylake-SP Xeon:
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/21

EPYC seems very strong. Zen cores seem perfect fit for enterprise. Energy consumption also looks nice. And AMD has priced them aggressively (half price for same perf in many cases).

 

[Pressure]

It's also nice with 128 PCIe lanes.

 

[Clukos]

And just a bit easier to grasp

 

[sebbbi]

That 7007$ Intel quad core is a steal

 

[Gubbi]

EPYC seems very strong. Zen cores seem perfect fit for enterprise. Energy consumption also looks nice.

We need Albuquerque to weigh in here, he's the expert. To me it looks competitive: high performance, high RAM capacity (2TB/socket) and low power.

Cheers