AMD Ryzen DDR4 Memory Scaling Tests On Linux

[dungeon]

I'd be interested to see 3200MHz results. I heard that seems to make a significant performance difference.

I heard the opposite happen too See recent JayzTwoCents video he get that G.Skill memory designed for Ryzen... and he can go to 3200MHz but perf was actually worse than at 2993

But he count it as improvment after 3 weeks as he couldn't pass 2667 initially.

[RavFX]

I have G.Skill 3200 CL16-18-18-18-38, the kit designed for Intel. I did not really bench but here are the compile time difference between 2133 and 3200: (LLVM with radeon stuff)

Tue Mar 28 15:57:35 2017 >>> sys-devel/llvm-4.0.0 | 2133Mhz
merge time: 15 minutes and 15 seconds.

Wed Mar 29 20:09:37 2017 >>> sys-devel/llvm-4.0.0 | 3200Mhz
merge time: 13 minutes and 43 seconds.

About a minute and a half gain, so I say about 11.5%. (I compile everything in a tmpfs)
Motherboard is the Gigabyte GA-AB350 Gaming 3, Bios F5 (F6d is a disaster and got pulled off by Gigabyte)

[bridgman]

He, he, my Asus AM1M-A mobo get bios updates about 15 times for the first 2+ years . I didn't expected they will support cheap platform and cheap mobo nowhere near that much, but for some reason they did.

In comparison to others, mine experiance with Asus bios updates was really FineWine and much better than expected Of course, this is rule of thumb as that depends - it is irregular and depends per mobo, as not all models and particulary if i look at other vendors not get same amount of updates - majority from the other vendors actually didn't get any bios updates

Yep, I have had good experiences with Asus in the past as well, but I'm probably an easier customer than some because I haven't had time to do anything other than plug the parts in and run them at stock speeds.

[Herem]

I heard the opposite happen too See recent JayzTwoCents video he get that G.Skill memory designed for Ryzen... and he can go to 3200MHz but perf was actually worse than at 2993

It only ran slower during the initial run at 3200MHz. Later in the video (around 13:25) he re-ran the test at 3200MHz and it returned the highest score achieved at any memory setting.

[oleid]

@Michael: How about benchmarking those scientific CPU tests that behaved oddly before?

[dungeon]

It only ran slower during the initial run at 3200MHz. Later in the video (around 13:25) he re-ran the test at 3200MHz and it returned the highest score achieved at any memory setting.

Yeah i know... but after triple checking or some other bench, it might slow down again I just wanna point something weird might happen after 2933, just to people open more eyes after that... maybe throtling appears there sometimes, does not look stable to me anyway

[pixo]

Hmm... I'm waiting for the Asus Prime X370 I ordered to arrive - everything else is here. It was pretty much the only mobo available up here in the Great White North.

Maybe by the time it gets here there will be a BIOS update that helps

I have this board and it has one design flaw. If I put GPU into the 3rd slot it will cover the USB3.0 header.

Another problem I found with 3rd x16 slot is that if I put GCN based GPU into it and enable IOMMU I am unable to boot.
Output shows "AMD-Vi: Completion-Wait loop timed out" errors and boot will stop due to kernel panic at some point.
The IOMMU groups are workable. Still I wish that chipset devices and slots would have isolation.

The good thing is that if I disable CSM in bios, the GPU to boot from will be the chipset one.

I hope that the IOMMU problems can be and will be solved by BIOS update.

[khnazile]

Could someone make tests of Ryzen with 64 GBytes of RAM at different clockspeeds? There were some slides showing that the memory clock is limited to 1866MHz with 4 dual-ranked modules. It would be very disappointing if that's true.

[dungeon]

I have this board and it has one design flaw. If I put GPU into the 3rd slot it will cover the USB3.0 header.

Another problem I found with 3rd x16 slot is that if I put GCN based GPU into it and enable IOMMU I am unable to boot.
Output shows "AMD-Vi: Completion-Wait loop timed out" errors and boot will stop due to kernel panic at some point.
The IOMMU groups are workable. Still I wish that chipset devices and slots would have isolation.

The good thing is that if I disable CSM in bios, the GPU to boot from will be the chipset one.

I hope that the IOMMU problems can be and will be solved by BIOS update.

Well i don't think it is design flaw and i wouldn't put GPU there, as third slot is not full one - it is 2.0 only, work at max 4X mode and share bandwidth with PCIeX1es...

AMD Ryzen™ Processors
2 x PCIe 3.0/2.0 x16 (x16 or dual x8)
AMD 7th Generation A-series/Athlon™ Processors
1 x PCIe 3.0/2.0 x16 (x8 mode)
AMD X370 chipset
1 x PCIe 2.0 x16 (max at x4 mode) *²
AMD X370 chipset
3 x PCIe 2.0 x1

**2.PCIeX16_3 slot shares bandwidth with PCIeX1_1 and PCIeX1_3.

https://www.asus.com/Motherboards/PR...pecifications/

[L_A_G]

DO NOT GET THE ASUS. The ASUS mobos have been shit for memory support. Regretting my Asus Prime X370.

Can't imagine it being any worse than my Gigabyte GA-AB350 Gaming. That thing wouldn't POST on anything above 2400 MHz.

During the short time I owned it it somehow succeeded in giving me a worse experience than my Asus P4P800-SE. This is saying a lot as the Asus broke down piece-by-piece over the years I tried to use that machine and was one of the major factor why I switched to using Macs for my personal machines (still have a 2011 Macbook Pro as my laptop for browsing the web on the couch and working on the go).