Question Userbenchmark GPU 0th percentile and NVME 4th?

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

ju1c3

Honorable
May 7, 2018
136
4
10,585
I wanna start off by saying, Yes I know userbenchmark is not the best, but 0th percentile??? could this be a sign of improper installation? I just installed the m.2 a few days ago and the GPU a few hours ago.

Here's a link to the bench:
Userbenchmark Test
Let me know if this is normal
 
Solution
Yep, probably, the only limitation I can see is the motherboard. If you wanna get full use of your components you'd be better off upgrading.
As for the low sequentials without direct access to the system it's hard to know. Their score is only slightly lower than my PCIe Gen 3.0 drive in sequentials, and pretty much the same for randoms, hence not bringing it up. It IS a different drive but same settings in CrystalDiskMark. Could be firmware weirdness on the Sammy drive I don't know (I'd expected it to be faster as well).
That's why I suggested looking at CrystalDiskInfo because the sequentials look like the drive is running PCIe 3.0 x2 instead of PCIe 3.0x4 since the randoms are all right where they should be. That being said the new board should resolve this entirely so it's just academic at this point.
Yes I know, I bought a dd4 motherboard, not ddr5, Next time I upgrade I'll be going with ddr5. Also I have 4 sticks installed as of now. My XMP is enabled. You think the ram mismatch could be limiting my m.2 speed? damaging my ram? or are you just saying the ram could be held back by that?
You have two sticks of 3200 and two of 3600 so all 4 are running 3200 and without you looking it's impossible to know what latencies they're running at. The Memory tab of CPU-Z will show what your DRAM is currently running at latencies wise. Higher frequency and lower latency is better and will improve CPU performance, but it will have nothing to do with your M.2.

Here's mine as an example (15-17-17-34 being the relevant numbers):
HUs77OA.jpg
 
Last edited:
That's why I suggested looking at CrystalDiskInfo because the sequentials look like the drive is running PCIe 3.0 x2 instead of PCIe 3.0x4 since the randoms are all right where they should be. That being said the new board should resolve this entirely so it's just academic at this point.

You have two sticks of 3200 and two of 3600 so all 4 are running 3200 and without you looking it's impossible to know what latencies they're running at. The Memory tab of CPU-Z will show what your DRAM is currently running at latencies wise. Higher frequency and lower latency is better and will improve CPU performance, but it will have nothing to do with your M.2.

Here's mine as an example (15-17-17-34 being the relevant numbers):
HUs77OA.jpg
d5d5ac4e860b541d9ad0a1c7e72da05c.png

https://gyazo.com/d5d5ac4e860b541d9ad0a1c7e72da05c
Here is a picture of mine. Tell me if anything is off.
 
d5d5ac4e860b541d9ad0a1c7e72da05c.png

https://gyazo.com/d5d5ac4e860b541d9ad0a1c7e72da05c
Here is a picture of mine. Tell me if anything is off.
Looks to me like it's running at 3200 with the latency from the 3600 kit so you're getting the downside of 3600 latency wise, but not getting the extra bandwidth.

If you look at the SPD tab it will show you what each kit has for XMP and JEDEC profiles (since you have two different kits you'd need to check each). This would tell you if the 3200 kit has notably better latency (I'm guessing it is going to be 16-18-18-34 or so). You could try to manually configure the latency to match your 3200 kit as it has lower latency than your DRAM is currently running. This won't be as big an improvement as just running 3600 would be, but I'm not sure 4 DIMMs would do 3600 period.

My system doesn't play well with DRAM over 3200 so rather than try to go higher I manually tuned a 3600 kit latency down with no issue.
 
Looks to me like it's running at 3200 with the latency from the 3600 kit so you're getting the downside of 3600 latency wise, but not getting the extra bandwidth.

If you look at the SPD tab it will show you what each kit has for XMP and JEDEC profiles (since you have two different kits you'd need to check each). This would tell you if the 3200 kit has notably better latency (I'm guessing it is going to be 16-18-18-34 or so). You could try to manually configure the latency to match your 3200 kit as it has lower latency than your DRAM is currently running. This won't be as big an improvement as just running 3600 would be, but I'm not sure 4 DIMMs would do 3600 period.

My system doesn't play well with DRAM over 3200 so rather than try to go higher I manually tuned a 3600 kit latency down with no issue.
10f17fa055094ce5dba0c54410470eb0.png

https://gyazo.com/10f17fa055094ce5dba0c54410470eb0
^^ 3200 16gb stick

d62e4c53a27e7633e296068d69587196.png

https://gyazo.com/d62e4c53a27e7633e296068d69587196
3600 8gb sitck
 
Interesting how the currently running timings don't match either kit of DRAM. With the latencies listed on them I wouldn't expect that you would be able to get much better timings given that they're also different memory IC on top of the other differences. I wouldn't bother messing with the settings on what you have now.

Depending on when you plan on upgrading your entire platform you could always look at a 2x 16GB 3600 CL16 kit ($80-100 in the US) which would net you better latencies and bandwidth. I wouldn't bother unless you're planning on keeping this one for another 3+ years though as the performance difference isn't going to be big.
 
Interesting how the currently running timings don't match either kit of DRAM. With the latencies listed on them I wouldn't expect that you would be able to get much better timings given that they're also different memory IC on top of the other differences. I wouldn't bother messing with the settings on what you have now.

Depending on when you plan on upgrading your entire platform you could always look at a 2x 16GB 3600 CL16 kit ($80-100 in the US) which would net you better latencies and bandwidth. I wouldn't bother unless you're planning on keeping this one for another 3+ years though as the performance difference isn't going to be big.
This upgrade of motherboard I'm doing will only mainly effect my M.2 running slow correct? are they any other benefits worth noting besides maybe 1% better performance on GPU etc?
 
This upgrade of motherboard I'm doing will only mainly effect my M.2 running slow correct? are they any other benefits worth noting besides maybe 1% better performance on GPU etc?
Yeah it should resolve the M.2 issue and might even positively impact the video card slightly (PCIe 3.0 vs 4.0 on the primary slot). It looks like the VRM is better, but it doesn't seem like you have any issues with the current board there. Internal case temperatures might go down as the VRM won't run as hot but I doubt it will lead to notable performance anything.
By the way I'm pretty sure my M.2 is in the wrong slot and that's why the speed is even lower than normal PCIe 3.0... It's the spot closest to the psu instead of closest to the CPU (bottom instead of top).
491a748ae96547d82b3f2b036a22fbe0.png

https://gyazo.com/491a748ae96547d82b3f2b036a22fbe0
You're right that is why the sequential performance is lower the slot is only x2.
 
You're using a GPU that many people likely won't have run through UBM. Check where your setup stands in 3DMark if you want to compare to others since you can search by GPU and CPU type rather than just comparing to everyone who happens to have tested that GPU. Everyone's setup is obviously different, but it will give you a much better idea than UBM.