AMD Ryzen CPU Core Scaling Performance

[indepe]

I don't think this has anything to do with AMD, as there are also strange results with i7 vs i5 when running Dota2/Vulkan. On the plus side, it may mean that Vulkan performance is much better than current test results (with Talos Principle currently out of the picture).

Quoting myself for a correction: That was actually i5 vs i3, here: http://www.phoronix.com/scan.php?pag...ogl-i3i5&num=1

i5 was always faster than i3, except with two cards in Dota2/Vulkan, in one case by 30% (0.7x).

EDIT:
In fact, the two cards were RX 460 and R9 Fury in that case, and R9 Fury in this case (with Mesa 17.0 and Mesa 17.1).

[L_A_G]

Hmm... I wonder what the problem here really is? Is it a BIOS problem?

Because apparently some reviews have found that Asus and MSI boards suffer from quite a few problems even under Windows with vendor drivers. This would suggest that boards have been pushed out the door before they're finished. Another thing that speaks in favour of the motherboards being at fault is the fact that right now AM4 boards are about as rare as hen's teeth outside of those receiving review kits. If they really were facing unexpected problems that weren't being solved in time for launch then hardware vendors would delay the start of actual production as much as possible.

One interesting tidbit that's cropped up during testing is that Ryzen chips have pretty bad latency to RAM and I'd put my money on this being the problem as the issue seems to be in games, which are more about latency rather than throughput. The problem so bad that Ryzen chips even fall behind AMD's old Excavator parts in tests measuring latency to RAM while doing just fine in tests measuring latency to the different levels of cache memory. Bandwidth tests are also showing some pretty good results for RAM and all levels of cache. In general heavy compute applications you're usually going to be accessing data that's stored sequentially in memory, meaning that each time you fill a new cache line, which contains multiple values, those additional values are going to be used, thus reducing accesses to RAM. The guess I'd make on this subject is that this is either something on the motherboard or then the MMU (which sits in the CPU) not being able to deliver on accesses to RAM in a timely fashion.

TL;DR: It could be an issue with either the CPUs themselves or the motherboards, which have suffered from loads of issues and have almost universally been delayed past the last minute.

[bridgman]

Take all of this with a grain of salt but...

Because apparently some reviews have found that Asus and MSI boards suffer from quite a few problems even under Windows with vendor drivers. This would suggest that boards have been pushed out the door before they're finished. Another thing that speaks in favour of the motherboards being at fault is the fact that right now AM4 boards are about as rare as hen's teeth outside of those receiving review kits. If they really were facing unexpected problems that weren't being solved in time for launch then hardware vendors would delay the start of actual production as much as possible.

My impression is that the mobo testing & associated BIOS fixes were orchestrated to be ready for product launch but they were not all in place a few weeks earlier when we started sourcing mobos to send out to pre-launch reviewers. The result was different reviewers having different experiences, although it appears that latest BIOSes bring all the mobos up to roughly the same level.

That said, any new CPU launch seems to result in a flurry of additional BIOS updates for the first month or two after launch so I wouldn't expect Ryzen to be any different.

One interesting tidbit that's cropped up during testing is that Ryzen chips have pretty bad latency to RAM and I'd put my money on this being the problem as the issue seems to be in games, which are more about latency rather than throughput. The problem so bad that Ryzen chips even fall behind AMD's old Excavator parts in tests measuring latency to RAM while doing just fine in tests measuring latency to the different levels of cache memory. Bandwidth tests are also showing some pretty good results for RAM and all levels of cache. In general heavy compute applications you're usually going to be accessing data that's stored sequentially in memory, meaning that each time you fill a new cache line, which contains multiple values, those additional values are going to be used, thus reducing accesses to RAM. The guess I'd make on this subject is that this is either something on the motherboard or then the MMU (which sits in the CPU) not being able to deliver on accesses to RAM in a timely fashion.

The Aida64 developers did not have access to a Ryzen before launch; there seems to be some consensus that the latency numbers are incorrect as a result. I don't know the specifics but keep an eye on this over the next week or so.

[sordna]

@phoronix: If the mobo doesn't support disabling SMT, how about following this:

http://nicolas.limare.net/pro/notes/2014/11/26_ht_freq_scal/

[L_A_G]

My impression is that the mobo testing & associated BIOS fixes were orchestrated to be ready for product launch but they were not all in place a few weeks earlier when we started sourcing mobos to send out to pre-launch reviewers. The result was different reviewers having different experiences, although it appears that latest BIOSes bring all the mobos up to roughly the same level.

Some reviews did talk about having to apply BIOS updates to sort out problems, but with the ExtremeTech review the problems went so far that you apparently sent them a Gigabyte X370 motherboard to replace the MSI one that wouldn't work properly even with BIOS updates. Then again these days it's more or less the norm for firmware and other software for hardware to actually be production ready very close to launch or after so it wouldn't surprise me if a lot of the problems were caused by reviewers simply not bothering to update their BIOSes to launch versions.

The Aida64 developers did not have access to a Ryzen before launch; there seems to be some consensus that the latency numbers are incorrect as a result. I don't know the specifics but keep an eye on this over the next week or so.

Well that would definitely explain the rather worrying latency figures, but on the other hand it would also make the worryingly low performance in CPU demanding real time applications, i.e games run at 1080p or below, a bit more baffling than what it is by removing one of the potential causes of it.

Mind you, I'm not the super demanding type, I actually expect to get slightly singed whenever I'm an early adopter of anything. I don't mind if it takes a software update or two to get my hardware running at full potential.

[sordna]

Phoronix: AMD Ryzen CPU Core Scaling Performance

Curious how Ryzen scales across its CPU cores and SMT? Here are some Ubuntu Linux benchmarks testing a Ryzen 7 1700 with different core/thread counts.

phoronix can you disable SMT by offlining the virtual cores? Google for "Disabling Hyper-threading and Frequency Scaling" and click the first link. I'm new and can't post URLs yet... however linux does allow you to effectively disable SMT and test the 8 physical cores.

[bridgman]

Well that would definitely explain the rather worrying latency figures, but on the other hand it would also make the worryingly low performance in CPU demanding real time applications, i.e games run at 1080p or below, a bit more baffling than what it is by removing one of the potential causes of it.

The performance numbers I'm seeing do smell a bit like they are memory bandwidth related, but some of that may just be that the BIOSes out there don't have all the fine-grained timing values for different memory speeds yet. Some of the review systems seem to be running memory more slowly than I would have expected (eg Michael mentioned running at 2133) but from what little digging I have had time to do at least one of the cases seems to be the BIOS having timings for two speeds, one fairly slow and one just a bit too fast.

That would fit with what I'm seeing - that 2666 sticks sometimes do better than 3000's right now because the 2666's actually run at 2666 (or whatever the number is, 2666 isn't looking right when I type it) while the 3000's actually run at something lower because they can't run at 3000 yet and they don't have the in-between timings yet. It would be nice if I could find some benchmarks against a 6900K running dual channel rather than quad channel - my guess is that gaming performance would be a lot closer and faster RAM clocks would help to close the gap with quad channel.

Again, take with a big grain of salt. I don't have a system to play with yet (CPUs are in stock around here but not mobos, which is the opposite of what I expected) so all I can do right now is look for patterns across a bunch of reviews like everyone else.

[leipero]

Mind you, I'm not the super demanding type, I actually expect to get slightly singed whenever I'm an early adopter of anything. I don't mind if it takes a software update or two to get my hardware running at full potential.

I don't know how trustwordy are those gaming benchmarks those youtubers do, I assume, same as Michel, they report numbers they get. One video I watched shows performance of [email protected] vs. 7700k@5GHz with 720p resolution and low graphic preser (to eliminate GPU bothleneck), and it seems that Ryzen falls exactly where it should be, at 1.1GHz disadvantage, it is pretty much same level of IPC performance, even Guru3D done all CPU's @equal GHz, and it suggests that Ryzen have slightly better IPC compared to Intel 7000 parts. So clearly, that position of "poor single threaded performance" is inaccurate. Furthermore, on those benchmarks at 720P low setings you can clearly see where i7 7700k becomes bottleneck and gets some stutters (yes even at 5GHz), while 1700 breaks no sweat at all, to be fair, due to the 1GHz+ higher frequency, it pushes more max. FPS, and that is completely expected.

Someone already posted the video in threads before, but if you can't find it i can post it (it's "something productions" video).

[bridgman]

Someone already posted the video in threads before, but if you can't find it i can post it (it's "something productions" video).

Joker productions ?

https://www.youtube.com/watch?v=nsDjx-tW_WQ&t=1s

Interesting... Joker is running memory at 3000 MHz, that might help to explain the difference in gaming results. I noticed the "MEM" usage on the HUD is different between 1700 and 7700K, not sure how to interpret that.

Hmm... I wasn't able to find any kind of summary of benchmark numbers - all of the comments *about* these videos on other forums suggest that the numbers were pretty close, but there has to be a better way than watching the video and scribbling down numbers.

[Mark Rose]

Performance aside, that MotherBoard looks skookum as frig!

Spotted the AvE fan.