HP EX920 wayyy underperforming. How do I go about getting it fixed?

I purchased the HP EX920 from newegg.ca roughly around 11/21/2018 and the first thing I noticed was that it ran and benchmarked very slow around only 1600mb/s seq read speeds.
Didn't make sense since pcie 2.1 x4 should be able to do 2000mb/s. Either way, I passed it off thinking maybe my older motherboard may be the bottleneck.

Now I have a brand new laptop with good specs (Y740), which came with a Samsung PM981 which should benchmark significantly lower than the EX920.
I benchmarked this Samsung drive extensively (see here: https://imgur.com/a/0HR9HtQ) with my system in it's stock factory state, then took it out, installed my HP EX920 after running a Secure Erase command from parted magic to reset this drive back to factory performance. This EX920 has only had light use and used with 50% over provisioning in my desktop as a secondary drive.
I did a clean install of windows and all the latest drivers I could find. I double checked for firmware updates to make sure it already had the latest version SVN139B (it did) then benchmarked it in two controller modes, ACHI mode using the official driver from your website, then in RST/RAID mode using the latest RST drivers from Intel (see here https://imgur.com/a/AYFSZIg).

On this same laptop where the samsung PM981 easily hit 2300mb/s this HP EX920 couldn't go anywhere past the same 1600ish mb/s bottleneck that I saw on my desktop regardless of what drivers or controller mode I used. To further verify if I had an exceptionally slow EX920 I my SSD scores in usermark from when I had the PM981 to now with the EX920. Which shows my HP EX920 wayyy under performing and the PM981 performing above average in the same laptop.


UserMark results (EX920): https://www.userbenchmark.com/UserRun/20703947
UserMark results (PM981): https://www.userbenchmark.com/UserRun/20699487

I was excited to use this SSD when I first got it only to be disappointed but giving it the benefit of the doubt and blaming my computer, now that I have a new computer and what should be an inferior SSD that's outperforming the EX920 I have, I'm now disappointed again after verifying it was not my computer but this drive that was underperforming. I was meticulous in my testing to make sure it was not my system that was the bottleneck. I'm quite sure that I've had a bad unit and have been using this bad EX920 this whole time.

Is there anyway to have this fixed or replaced under warranty or something? The HP website just leads me to a manual and no proper contact, no matter how hard I search, and I cant register my EX920 or check it's warranty because the website cant find anything with my serial number.

 

PCIe 2.0 x4 max ~1650MB/s because of 8/10 encoding, even if you were to turn off Cstate, LPM and ASPM you won't get to 2000MB/s. also don't trust too much in CDM, its wildly inaccurate when comes to it's number and should only be used to see if you get similar numbers with others who have the same drive.

Y740 is 8/9th gen cpu? assuming SSD is via chipset, it's limited by DMI 3.0 so ~3600MB/s. maybe your EX920 stuck at 2.0 for some reason or defect. i have seen weird defects like that on storage in general for both PCIe, SATA and even USB.

also, why do you think EX920 is under performing with those numbers? it might be what it's normal speed at QD1 you never know. do you have someone with another EX920 tested their numbers against yours? try running your EX920 in CDM with sequential read at Q2 or Q4 and see if speed goes above 2000MB/s if it does then theres no problem.

 

If that's the case I can understand it performing like that in my desktop but doesn't make sense in my new laptop. In all the benchmarks I've seen of this drive it easily hits over 3k mb/a in seq read tests. Look here for instance https://www.tomshardware.com/reviews/hp-ex920-ssd,5527-2.html it easily hits 3.2k mb/s in a QD2 256kb read test. That's a way harder test than the QD1 1mb test I'm running too.. my PM981 actually performs faster than most other PM981 drives from what I've seen by a slight difference, easily hitting 2.3k+ mb/s the same test my EX920 can't even go past 1.7k mb/s in. If it's stick in PCIe 2.0 mode how do I fix it.

Also I made a mistake, my desktop supported PCIe 3.0, and was using 4x slot, so there shouldn't have been a 1600 mb/s bottleneck like there was.

 

first, understand that laptop/desktop isn't the problem here. if your drive caps you at 2.0, you are stuck regardless of which new laptop/desktop you have. if it's not a defective drive, then we look at your CPU/chipset, laptop/desktop again is irrelevant here unless its something 4+ yrs old.

now assuming it's not a defect with your EX920. i edited the graph a bit so look at the edited part.




see how HP EX920 1TB at QD1 starts at 2000 MB/s? thats exactly what CDM Q1T1 means. EX920 only hits 3000MB/s once it goes past QD2. again, if your issue isnt a defective drive capping you at PCIe 2.0 speed, then this is normal.

this is your EX920 graph, see the SEQ and then Q1T1? if you change that to Q2T1 and test it and you are not restricted by defect or PCIE 2.0 then u'll hit over 3000MB/s prob.





you may think samsung is fast, that is because they do multiple flash channels and OC their controller for higher QD1 sequential bandwidth, but that causes them to suffer in random read/write performance, which they use the help of firmware + Dram and SLC layer to mask up that poor random speed.

not saying EX 920 doesn't have DRAM but most drives that uses TLC has this problem.

 

While that's fair, I'd agree if I was being capped at 2k at QD1, but I'm not. Besides, usermark shows me that my EX920 is performing in the BOTTOM 9% of ALL EX920s tested in 3350 EX920 benchmarks, which is probably the most alarming thing. I still have my Q8 tests on my PM981 which scored around 3.3k mb/s read. I'll run tests again on my ex920 at higher queue depths to see if you're right.

Ok results are in. I'd like to note, I have nothing running in the background but intel RST and had the laptop set to performance mode and -0.15 undervolt on the 9750h, something I didn't do when testing the samsung. I also have faster Kingston HyperX Impact C15 ram slotted in, in place of my old SK Hynix C19 ram that I had in during my testing of the PM981.

Samsung PM981:
[ Read]
Sequential 1MiB (Q= 8, T= 1): 3313.208 MB/s [ 3159.7 IOPS] < 2513.70 us>
Sequential 1MiB (Q= 1, T= 1): 2207.550 MB/s [ 2105.3 IOPS] < 474.68 us>
Random 4KiB (Q= 32, T=16): 1717.898 MB/s [ 419408.7 IOPS] < 1180.66 us>
Random 4KiB (Q= 1, T= 1): 47.003 MB/s [ 11475.3 IOPS] < 86.98 us>

[ Write]
Sequential 1MiB (Q= 8, T= 1): 2399.406 MB/s [ 2288.3 IOPS] < 3452.32 us>
Sequential 1MiB (Q= 1, T= 1): 2393.720 MB/s [ 2282.8 IOPS] < 437.68 us>
Random 4KiB (Q= 32, T=16): 2036.358 MB/s [ 497157.7 IOPS] < 996.41 us>
Random 4KiB (Q= 1, T= 1): 138.229 MB/s [ 33747.3 IOPS] < 29.48 us>

HP EX920:
[ Read]
Sequential 1MiB (Q= 8, T= 1): 1941.713 MB/s [ 1851.8 IOPS] < 4315.94 us>
Sequential 1MiB (Q= 1, T= 1): 1869.219 MB/s [ 1782.6 IOPS] < 560.51 us>
Random 4KiB (Q= 32, T=16): 1431.368 MB/s [ 349455.1 IOPS] < 1446.69 us>
Random 4KiB (Q= 1, T= 1): 63.444 MB/s [ 15489.3 IOPS] < 64.42 us>

[ Write]
Sequential 1MiB (Q= 8, T= 1): 1661.418 MB/s [ 1584.5 IOPS] < 5016.56 us>
Sequential 1MiB (Q= 1, T= 1): 1707.533 MB/s [ 1628.4 IOPS] < 613.55 us>
Random 4KiB (Q= 32, T=16): 804.184 MB/s [ 196334.0 IOPS] < 2600.69 us>
Random 4KiB (Q= 1, T= 1): 139.483 MB/s [ 34053.5 IOPS] < 29.22 us>

See the difference? I'm capped at 2k for some reason no matter what QD I test at. According to the same review benchmark you used as example, I should be able to easily hit 3k+ mb/s. I wasnt able to do that even on my original desktop with this drive when it was brand new even though my motherboard was pcie 3.0 spec.

PS - I opened up my laptop and reseated the nvme. Still capped at 1800-1900 mb/s regardless of queue depth.

 

then its a drive defect? if you have problem hitting over 3000 MB/s once you get past 4Q on sequential read then its not same as advertised. if you purchased an OEM drive then that could be a reason, either defect or purposely limited speed.

if not and should be same as all other EX920 then return it and get a new one.

usually if you get one of these problem u'll check drive in at least 2 computers, then check on your existing OS and a new OS with microsoft's default NVMe driver. then u'd start limiting down what the problem is.

 

Definitely a drive defect. I've contacted the company that handles support and warranty for the EX920 already. They said they're looking into it to see if they can reproduce the issue and find a fix instead of having me ship it back for a new one.

 

Are there any other tricks to get drive speed up? I never trusted those advertised number much, but software limitations are obviously the biggest hurdle.

 

more speed = less power saving, you can set window's high performance and edit it's power settings then toggle to battery profile with window's default settings.

settings to change, i recommend jon's from this page https://www.tweaktown.com/guides/68...sd-performance-installation-guide/index8.html

as for cstate, you can use throttlestop to toggle it on/off. on when you want better cpu/storage performance, off when you're using battery profile. (assuming intel cpu)

anything else it'll be on bios/hardware level which theres no profile for it.

 

With C states, make sure you leave C1 auto demotion and undedemotion checked. Unchecking them seems to give way worse performance, especially on q1t1 usage. Not sure about every other options as there doesnt seem to be much of a benefit otherwise. Perhaps there is some other way of checking it?

 

think of it like this. bios has it's power saving feature such as cstate, pstate and other things like ASPM, LPM on chipset/pcie/dmi.

windows has core parking and power saving etc and you can access those option by tweaking registry to change demotion/promotion etc.

within windows if you use TS, there is speedstep and cstate which you also need to turn them off.

basically just turning one of them off may or may not work as each computer is different so to ensure they are all not controlling it, you disable them all. then you do a test you'll get maximum performance, anything else beyond that would be hardware side, SSD firmware, OS versions with it's NVMe drivers.

overclocking CPU also wins because CPU is the central of everything so u get boost from pretty much everything which is why turning cstate off affects SSD performance.

 

It certainly varies from computer to computer. I noticed that I must keep all of those checked as I get microstutter in other applications. Are you aware of anything in regards to "Interrupt response limit"? I cant find any particular documentation as to doing what with what in it actually does.

 

not sure what interrupt response limit is. but it makes no sense to have micro stutter if you turn off power saving infact it should be the opposite, less micro stutter/ lag/ delay once you turn them off, because you get more performance.

only thing i can think of is if you are really getting microstuter having turning power saving feature off then it could be because power/temp throttling either in CPU/storage causes stuttering.

 

We are discussing that in throttestop forum now. I dont know either, so more testing is on the way.