Samsung 850 Evo Low IOPS

[Thentilian]

I have been on a search for AHCI drivers. The drivers supplied by Asus won't work with this board and version of Windows. I'm running Windows 10 Pro 64 bit. I'm selecting the proper OS but even going into each of the driver files for SATA and such, none of them will work.

The generic windows driver gets me these results:


While digging around in the driver update section I manually installed the only AMD driver I could find and it gave me these results:


Windows doesn't recommend using that file, so I went back to the generic windows driver. But as you can see the IOPS aren't even close to what the SSD is capable of. I ran the tests because it seemed like my boot times were pretty slow compared to when I had this drive in my old AM3 build. That PC booted faster, despite it only having SATA 2 interface and not SATA 3 like on my current Asus Prime X370-Pro motherboard.

I've even gone to AMD to get drivers, even looked at their RAID drivers, none of them will work. I remember there being AMD specific AHCI drivers, did they just stop updating it?

 

[Johan45]

http://support.amd.com/en-us/download/chipset?os=Windows 10 - 64

 

[Thentilian]

http://support.amd.com/en-us/download/chipset?os=Windows 10 - 64

I did download those, it doesn't install the AHCI driver. I even looked for it in the folder manually when saying 'I have the disk' and navigate to the folder, and nothing worked. Maybe I didn't find them, there is a lot of folders. I'll try again, thanks.

EDIT: Here is the list of drivers it wants to install... no mention of AHCI in there, or SATA for that matter.


I found them under a folder called 'hseries', I just wonder why they aren't installed to begin with. Windows doesn't want to use them, but they are digitally signed. I think it's the same driver I found before, performance was on the 69k read and 60k write. I guess that's just about as good as it's going to get. I heard AMD AHCI drivers aren't nearly as good as Intels. Also boot time went from 37 to 40 seconds to 27 to 30, so I am gonna stick with it unless I notice anything odd going on.

 

[Johan45]

Pretty sure it's the GPIO drivers. General Purpose I/O

 

[Thentilian]

Pretty sure it's the GPIO drivers. General Purpose I/O

Tried to use the GPIO drivers just now, none of them were compatible. Only the hseries folder with the AHCI sub folders had drivers that would install. I took a caption of what it said when trying to use the drivers in the GPIO folders.


At least the AMD driver I did find bumped performance up quite a bit. Boot times have gone as low as 24 seconds now, instead of 37 to 45 before using that driver. The IOPS are still not where they should be, but I think that's just an AMD driver/chipset issue from what I saw on forums of others complaining about this issue. Not something I generally research before buying a motherboard, but I guess it's interesting information to keep in mind when choosing a motherboard. I couldn't say for sure just how badly this affects load times with other things, but it does feel snappier since installing the AMD driver and boot times definitely are a lot faster.

 

[Johan45]

Well that's just odd. I never noticed anything like that when I installed the drivers but that was a while ago.

 

[Thentilian]

Well that's just odd. I never noticed anything like that when I installed the drivers but that was a while ago.

It's all new to me. I was wondering what the GPIO drivers were to begin with. Still wondering since they don't seem to work with the as the AHCI or SATA drivers. At least the AMD one I found is from 2015 instead of the MS generic one from 2006. Be nice if AMD updates this driver though, as IOPS performance is lacking by about 25 to 30% of where the drive is capable. I bought a speedy SSD, would be nice to get all the performance out of it that I can.

 

[Johan45]

I just checked an my system is using the MS drivers as well seems AMD got a bit lazy.

 

[Thentilian]

I just checked an my system is using the MS drivers as well seems AMD got a bit lazy.

What SSD are you using, and is your IOPS also suffering in performance?

 

[||Console||]

your results do seem low .

 

[Thentilian]

your results do seem low .
View attachment 193409

Yeah, Intel chipset vs AMD chipset. Looks like the driver for AMD is two years old. I think AMD needs to get on the ball and update some of their aged drivers. I had noticed results like this from others complaining that were on The SB8xx chipsets as well. I never paid much attention to this when I was on my AM3 board as it was SATA II, so I knew performance was going to be cut down. I expected to see much better performance being on a new chipset with full SATA III support, but the driver seems to be where the lack of performance is coming from, unless the chipset itself has issues... And I wouldn't even begin to figure out how to test what the issue actually is.

 

[EarthDog]

Check writeback cache. If its disabled, enable it, if it's enabled, disable it. See if that helps.

How full is the drive?

 

[Thentilian]

Check writeback cache. If its disabled, enable it, if it's enabled, disable it. See if that helps.

How full is the drive?

If you mean this:


Yes, it was enabled... I disabled it, and the test took over 3 minutes to get to 1%... so I canceled it. I reticked it and ran the test again and ended up with this:



The SSD has 114GB of free space with 9GB of it over-provisioned. It was exhibiting the same behavior before I even filled the drive 50%. It's running faster with the AMD driver I found, but still a bit slow on the IOPS. Not sure what kind of performance loss this is going to be, if any, but I did notice my boot times go down quite a bit after using the AMD driver instead of the Windows generic one.

 

[EarthDog]

Checj the turn off write back cache buffer...

 

[Thentilian]

Checj the turn off write back cache buffer...

Ok, I turned that off, results were slightly better, but I don't know if it's worth having that ticked off or not.

 

[Alaric]

I'm following this thread with interest, but Samsung Magician seems to have a problem on my rig. I ran the benchmark and got what has to be skewed results, so I installed the latest version and ran it again. These are the results:
Earlier version (4.5)


Current version



Any ideas? Those numbers would be nice, but I don't think they're right. LOL

 

[EarthDog]

Clearly not right unless you are running cache?

Old version, new drive, potentially unrealisitc results are possible!

 

[Thentilian]

I'm following this thread with interest, but Samsung Magician seems to have a problem on my rig. I ran the benchmark and got what has to be skewed results, so I installed the latest version and ran it again. These are the results:
Earlier version (4.5)
View attachment 193450

Current version
View attachment 193451


Any ideas? Those numbers would be nice, but I don't think they're right. LOL

Do you have RAPID Mode turned on? It will give crazy Read/Write numbers because of the cache drive it makes... Sustained Read and Write operations that exhaust the huge buffer will slow down to more realistic speeds. IOPS, I'm not sure about, but clearly something is giving you a massive jump in the numbers.

 

[Alaric]

Yeah, I completely forgot about Rapid mode. I knew it was turned on but didn't know it could do that to the benchmark.

 

[Thentilian]

Yeah, I completely forgot about Rapid mode. I knew it was turned on but didn't know it could do that to the benchmark.

Did you turn it off and retest? I'm curious as to what the numbers would be after that. I think the issue with my PC is AMD hasn't updated that AHCI driver since 2015. That's pretty old at this point. Besides the time on my PC always behind... that's becoming a real annoyance and I hate not being able to figure things out.