-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Interesting, thanks for the share.
This is what mine was at about a month back.
NUMA 3200 C14-13-13-28-42-256-1T tFAW 18 tRDRD & tRWRW SCL 2
Thanks for the share.
Interesting results. You have higher mem bandwidth all round, but lower CPU cache scores.
What difference does 18 TFAW make? I think i am using 24. What about TRDRD too? I am already on SL 2 & 2.
Soldato
- Joined
- 19 Oct 2008
- Posts
- 6,033
Remembered to do this. 1950X stock scored 166 cb single core in Cinebench if that's still of any help. That's using the Asus X399-A and 3200 mems and SVM disabled (virtualisation).Yes definitely not enabled that. I am using a more recent bios version i think, but as i have Raid Nvme now i can only roll back so far and no earlier bios that supports Raid has helped so far.
It's not a huge issue, but now i know performance has dropped it is annoying me.
Yeah i did get a score or two near that, and i think one even beat that but now it has dropped a bit. Ah, think i will just give up worrying about it for now.
Ryzen Master Tool for TR is super neat btw. I've now ditched CPU overclocking via the Bios and just ramp up clocks temporarily while gaming as needed.
I actually cant remember what they do now. IIRC, TRDRD & TRWRW affect read and write latencies. I've never been able to figure out why my cache data rates are slower than the 1950x but looking at @sandys result, they are lower than mine so maybe it's related to the number of cores.
Out of interest do you see comparatively low firestrike combined scores compared with the 1800x?
Don't think so, but combined scores seems to be all about low driver overheard and CPU IPC so not that favourable for Ryzen/AMD when running standard firestrike. Less of an issue at Firestrike Extreme and 4K as the bottleneck is the GPU. Firestrike standard is more a CPU test these days.
Got to the bottom of my useless Threadripper RAID0 NVMe Array, well not too the bottom as such but worked around it, it would appear the issue is some combo of the AMD Raid driver or perhaps my MSI board BIOS and driver, assuming no one else has a problem, if you don't have two arrays on the system a problem may not manifest, I wanted to copy data between two NVMe arrays on the RAID controller, on doing so the system would die on its arse.
I've done numerous re-install, switching cards around etc, decided just to use the drives as they come and thrash 3 at the same time outside of RAID just using them like normal disks and they worked fine, so set one up as a RAID and it was slow again, so I have just disabled AMD RAID and created a RAID0 in software through windows and the thing is flawless, OK it doesn't have the headline speeds of the TR RAID setup stand alone but I can do my transfers and see at least 2Gb/s between discs (the other is a single drive not raided yet, I was just testing)
So I guess that is how I will use my system for now, unless someone has any insights on why the AMD RAID wasn't performing.
In some ways I prefer it this way, the system boots quicker when not having to do RAID initialization and I can still use driver monitoring software, which you can't do on the RAID controller.
It'll have its downsides like lack of trim but I'll deal with that should I notice a operformance degradation, perhaps by them there will be new drivers/bios.
Cheers
I've done numerous re-install, switching cards around etc, decided just to use the drives as they come and thrash 3 at the same time outside of RAID just using them like normal disks and they worked fine, so set one up as a RAID and it was slow again, so I have just disabled AMD RAID and created a RAID0 in software through windows and the thing is flawless, OK it doesn't have the headline speeds of the TR RAID setup stand alone but I can do my transfers and see at least 2Gb/s between discs (the other is a single drive not raided yet, I was just testing)
So I guess that is how I will use my system for now, unless someone has any insights on why the AMD RAID wasn't performing.
In some ways I prefer it this way, the system boots quicker when not having to do RAID initialization and I can still use driver monitoring software, which you can't do on the RAID controller.
It'll have its downsides like lack of trim but I'll deal with that should I notice a operformance degradation, perhaps by them there will be new drivers/bios.
Cheers
You have to assign the NVMe Boot disk as a Volume on the array and then load the raid drivers when doing the install, or at least that is what worked for me.
On that note, I am seeing poor performance when trying to use my new raid array and transferring between two disks on the controller.
Independently the disks are great but transferring between them which is surely why you want nice fast discs it can be slower than a normal SSD/HDD, any tips.
Top two tests are independently run, bottom two tests two disks, (1 drive, 1 Raid0)thrashed at the same time, spu is not being hammered so am unsure why I have this bottleneck, should be no bottle neck with all the lanes?
Last edited:
Ok been dicking about some more, couldn't leave it alone, finally got the AMD Raid drive to work once I installed it in Windows rather than do it at boot time, not sure why perhaps my system had a hardware/software conflict that doesn't get resolved until all software is upto date., So yeh, raid NVMe that is faster than a normal HDD in use, yay.
But interestingly comparing AMD RAID0 speeds with my Windows RAID0 using same drives slots but Samsung NVMe driver rather than AMD Raid driver, AMD RAID seems a touch slower, not that I can feel it, but see it in reported benchmark numbers.
AMD RAID benches
Windows RAID benches, same drives, same slots etc.
Interesting
But interestingly comparing AMD RAID0 speeds with my Windows RAID0 using same drives slots but Samsung NVMe driver rather than AMD Raid driver, AMD RAID seems a touch slower, not that I can feel it, but see it in reported benchmark numbers.
AMD RAID benches
Windows RAID benches, same drives, same slots etc.
Interesting
Last edited:
Yup, they perform well despite being the cheaper OEM disks, I wonder if the windows raid numbers are high due to double counting as both disks show up as D: when striped in windows rather than displaying as 1 drive but then the sequential reads/writes show up in the expected range rather than double so it doesn't seem to be the case, dunno, both seem as quick as each other in use as I hit the max of my other disk, not sure if there is another way to test them.
Last edited:
Associate
- Joined
- 29 Jun 2016
- Posts
- 543
Seeing the same behaviour.
Have CMR32GX4M4C3466C16 (B-die) in the Aorus 7, where that RAM was well cheap on amazon compared to anything GSkill.
Using XMP, giving 3466 w/ VDDR @ 1.35 and VSoC @ 1.1 the PC is not stable under prime/memtest, but runs surprisingly well.
Dropped the frequency to 3200 w/ the same voltages and it's smooth sailing.
Wonder what the major factors are in that 3200MHz being a common achievable frequency, if its the signalling, IMC or something to do with the die-die interconnects?
I've got no idea, but it seems a lot of people can get 3333-3466Mhz stable. Guess we're just unlucky.