Force Series 3 performance with AMD

I did not see much info in the other Force series 3 threads regarding this so I started a new one. Yes I have seen the recall and I plan on going through the RMA process tomorrow but for now my two drives are working.

Motherboard is Gigabyte GA-890FXA-UD5 Rev 2.0

rest of specs in profile

I knew this was not the best sata controller but performance is worse than I expected. I have the drives plugged into the SATA 3 (SB850) ports and I have enabled and disabled SATA 3 mode.First I added the drives as data drives to my existing windows install. The sata controller is in RAID mode because I have another array.

Right out of the box I ran read tests (trying to write as little to the drives as possible so garbage collection can keep up) with several programs and the highest speed I could get was 268 MB/s (240 MB/s with SATA 3 disabled). I also tried running the ATTO and IOmeter test according to the corsair blog and received about 265MB/s on ATTO and only 10k IOPS in IOmeter. attached results.

I tested the RAID driver shipped with my board, one from a few months ago and the newest. I also tried a RAID 1 and RAID 0 netting approx 500 MB/s read.

I also did a clean install on the ssd's in a RAID 0 and saw no increase in performance.

So is this normal performance or do I have a driver issue? Also should I try a clean install on a single drive and set the controller to AHCI? I could not find any other BIOS settings that looked out of place. Or is the recall issue with the drives something that also hurts performance?

Thanks for any help!

ssd results.txt

Yellowbeard said in another thread that AMD controller may not be able to make drive perform at rated speeds. Not sure if that meant any/most/all AMD controllers though.

I am using a Gigabyte GA-770TA-UD3 MB. My benchmarks were not also not the greatest.

I am using AHCI on 6Gb SATA connection.

We'll see after I get my replacement F3.

Yes I have checked the drivers several times. I used the sb850 RAID driver from Gigabyte that came with the board, a few month old AMD driver, a new AMD driver from their website, and also the built in Win 7 driver on a clean install. All resulted in the same performance. And yes both drives perform exactly the same.

I verified that the drives are in the correct ports,(they show up in RAIDxpert as 6 Gbps) and I have tried them in 4 of the 6 SATA 3 ports. Also tested in the 2 SATA 2 ports of the other onboard controller which had much worse performance. Also have tried it with the CPU/RAM overclocked and at defaults.

Unfortunately I do not have access to another board with SATA 3 ports. So far the only possible problem I have found is that my SATA cables that came with the motherboard are only rated for SATA 2. Would an inadequate cable reduce performance or just not work.

I have heard that AMD may not produce the best performance, but I expected it to be at least 300-400 MB/s read compared to 500+ rated.

Really I just wanted to see if anybody thought it might be related to the RMA issue with the drives. Hopefully my new ones will work better, but I have a feeling the problem is on my end.

r9ball, what kind of benchmark numbers did you get?

According to controller specifications, the drive specifications, and internal testing, there are only 3 chipsets currently capable of running the Force 3 drives up to their specification. P67, H67, and Z68.

Yellowbeard, are there any consumer level pci-e add-in cards that would yield good performance numbers.

Also could you comment on whether my performance is on-par with tested AMD systems or is it even lower?

I personally have not tested any PCI-e cards so, you'd need to consult some other source for that information. You should be able to find some product reviews online.

Also, we knew before we started all this that P67/H67/Z68 were the only MOBO chipsets that would be able to run the Force 3 drives up to spec. Therefore, we have not tested on AMD.

I have not read it all but I think this AnandTech review will shed some light on your situation.

Thanks for the information so quickly. Anand actually was able to get over 300 MB/s read with that other sata 3 drive on 890fx, so I was expecting at least slightly better results than what I am seeing. He actually seemed to like the amd controller at that time granted the tested drives were not nearly as fast as the current sata 3 drives.

Seems I will have to do more research and possibly see if I can return the drives for a refund and get some ssd's more suited to my mobo or find an affordable add-in card.

Here are my marks from the Force 3 and another brand which can be figured out by the name of the pic.

It is odd that the Force 3 seems to be opening programs a little faster.

We'll see when replacement arrives. I still would have thought the Force 3 would have done better than it did. If benches don't change some with the replacement, then not much reason to keep it cause I'm not going to buy a new CPU and MB and redo everything just for the SSD. MY other Corsair items have been great and it is my fault for not digging deeper to see how the SSD would be in an AMD system.

http://www.nielsensbilliards.net/rick/f3.jpghttp://www.nielsensbilliards.net/rick/c300.jpg

r9ball,

Your motherboard (GA-770TA-UD3) uses the AMD SB710 chipset. This is a SATA II chipset (not SATA III).

Your motherboard does have a Marvell 9128 chip for SATA III support - but this is not an AMD chipset - your SATA III support is from the Marvell controller.

Here is my result for f3 (i had only one full run, all other were incomplete because bsod's).

http://dl.dropbox.com/u/15434121/meh.png

I have asrock 890fx deluxe3, f3 was connected to sb850 sata3 port. I think i had latest amd drivers when i ran that test.

r9ball,

 

Your motherboard (GA-770TA-UD3) uses the AMD SB710 chipset. This is a SATA II chipset (not SATA III).

 

Your motherboard does have a Marvell 9128 chip for SATA III support - but this is not an AMD chipset - your SATA III support is from the Marvell controller.

This was my initial thought looking at his dropdown specs but........

Motherboard is Gigabyte GA-890FXA-UD5 Rev 2.0

rest of specs in profile

Here is my result for f3 (i had only one full run, all other were incomplete because bsod's).

 

I have asrock 890fx deluxe3, f3 was connected to sb850 sata3 port. I think i had latest amd drivers when i ran that test.

Thanks for sharing the information. It's good to see that there is some hope for using these drives on my motherboard. Now I just have to figure out what the issue is.

Do you know if your SATA cables are rated as SATA 3 6Gbps or are they standard SATA 2 cables. Also was your controller in AHCI or RAID mode, I may try AHCI tonight and forget about the RAID 0 for now.

Yellowbeard, It looks like you may have mixed up posts/systems between me and r9ball. I overlooked that his AMD system had a different controller.

Oh BTW, still no BSODs or missing drives. been testing for a several hours for the last few nights.

Thanks for sharing the information. It's good to see that there is some hope for using these drives on my motherboard. Now I just have to figure out what the issue is.

 

Do you know if your SATA cables are rated as SATA 3 6Gbps or are they standard SATA 2 cables. Also was your controller in AHCI or RAID mode, I may try AHCI tonight and forget about the RAID 0 for now.

 

Yellowbeard, It looks like you may have mixed up posts/systems between me and r9ball. I overlooked that his AMD system had a different controller.

 

Oh BTW, still no BSODs or missing drives. been testing for a several hours for the last few nights.

I think those SATA cables are SATA 3 6Gbps rated, because they came with that motherboard.

I was running that F3 in AHCI mode (bios and windows enabled AHCI), no experience from RAID mode.

Yeah I didn't think about my SATA3 using the Marvell controller

I am using SATA 3 cable and AHCI.

The other brand is performing at spec or a little better so I think my system is fine. We will see with replacement.

I might also consider upgrading to newer AMD MB if needed.

Diari's benchmarks look pretty good!

I personally have not tested any PCI-e cards so, you'd need to consult some other source for that information. You should be able to find some product reviews online.

 

Also, we knew before we started all this that P67/H67/Z68 were the only MOBO chipsets that would be able to run the Force 3 drives up to spec. Therefore, we have not tested on AMD.

 

I have not read it all but I think this AnandTech review will shed some light on your situation.

Given the amount of people using AMD boards, shouldn't Corsair think about testing their drives with AMD. It is a small outlay for a board and a processor - maybe an hour or two of setup and nominal testing. Nothing thorough, but enough to get some actual insight.

Like I said in an earlier post, I have seen results posted on other forums from people using AMD 850 chipsets with the new Sandforce drives that rival the Intel scores.

Might want to take a look elsewhere or do some actual testing before you discount the AMD boards. They really are returning some excellent results - respectfully suggested.

Really appreciate your and Corsair's responsiveness on these forums by the way, regards.

Given the amount of people using AMD boards, shouldn't Corsair think about testing their drives with AMD. It is a small outlay for a board and a processor - maybe an hour or two of setup and nominal testing. Nothing thorough, but enough to get some actual insight.

 

Like I said in an earlier post, I have seen results posted on other forums from people using AMD 850 chipsets with the new Sandforce drives that rival the Intel scores.

 

Might want to take a look elsewhere or do some actual testing before you discount the AMD boards. They really are returning some excellent results - respectfully suggested.

 

Really appreciate your and Corsair's responsiveness on these forums by the way, regards.

It's not a matter of testing. The AMD boards themselves don't have the same performance so even if Corsair tested their drives with them it's not really going to change anything.

With that being said the difference in performance is pretty small between the chipsets, this I will agree. What makes a bigger difference is the actual controller you're using. For the most part, the Intel controllers have better performance than Marvell, JMicron, and Silicon Image. It has nothing to do with brand biasing, but Intel is clearly the leader in performance.

It's not a matter of testing. The AMD boards themselves don't have the same performance so even if Corsair tested their drives with them it's not really going to change anything.

 

With that being said the difference in performance is pretty small between the chipsets, this I will agree. What makes a bigger difference is the actual controller you're using. For the most part, the Intel controllers have better performance than Marvell, JMicron, and Silicon Image. It has nothing to do with brand biasing, but Intel is clearly the leader in performance.

I absolutely understand that Intel generally gives better performance. What I am saying here is that when another company initially brought the new sandforce drives out, there were quite a few benchmarks posted on , mostly Intel, but also a few AMD boards in another forum. Those benchmarks show the AMD boards VERY close in performance to Intel in AS SSD marks as well as ATTO. If it wasn't a competitors site I would post a link.

The first page of the reviews has an Asus 850 board AS SSD score of 697, shortly after that is a post with a Asus P67 board - AS SSD score of 692.

The ATTO scores are within a heartbeat of one another also.

Nothing to do here with brand biasing either, but the results pretty much state the case. I was a bit supprised because all of my SATA II drives are markedly slower for me than what I see posted on Intel SATA II boards.

Attached are my AS SSD and ATTO benchmark scores running on my PC listed in my profile in AHCI mode using the latest AMD chipset 8.71 drivers. I used the 6Gbps SATA cable that came with my motherboard.

I'm using 32-bit Windows XP Media Center Edition. Maybe I need to upgrade to 64-bit Windows 7 to get the good numbers Diari posted? My mobo is using the same SB850 south bridge as Diari's mobo, so in theory I ought to be able to reach the same speed.

I have an update. I have narrowed the issue down to my motherboard and in particular the BIOS settings.

I have found that the poor performance happens when my sb850 is in RAID mode with the RAID 5 option disabled, ,which is how I had it set up with my old drives. Also I found that just switching to AHCI as was done at the beginning of this post does not restore performance. To get good performance I have to reset the board to default first then set it to either AHCI or RAID with RAID 5 enabled. I am not sure why RAID 5 enabled/disabled changes performance and why switching from disabled to enabled with out a reset to default does not restore performance.

In AHCI mode after setting to defaults I now get read speeds up to 500 MB/s on both drives (in a used state after a few windows reinstalls). I have not run any other tests yet. In a RAID 0 config I now get up to 760 MB/s instead of 500MB/s. So performance is still not as high as the Intel systems, but it is at least acceptable now. Oh, interestingly When in IDE mode I got 340 MB/s read.

I don't have any SATA III drives any longer but I do have three other pairs of SSD's in raid 0 - I enabled the raid5 just to see what would happen (didn't get as thorough as you going to default and back however) I had a 10% increase on the AS SSD scores for one set - the others remained about the same. The ATTO results tended to favor write increases at the expense of reads, but improvements with all three sets - nothing nearly as dramatic as yours however.

I have noticed that the bios settings for my 850 chipset have some quirks also - I had a set of (not corsair) SATA III drives - I had SATA III enabled initially, but had to disable then enable it for the SATA III to function. On my first run of ATTO the drives individually had reads that topped out around 244,000. After doing the disable/enable they were over the 300K mark.

Aggiebroz,

You did get up to 500MBs in AHCI?

Was that in a type of RAID or was that as stand alone drives?

I only have a single drive with AHCI and the SATA III.

My MB has RAID and can change BIOS settings but not sure if that would help since I am not RAIDing the drive...Gigabyte GA-770TA-UD3

You did get up to 500MBs in AHCI?

Was that in a type of RAID or was that as stand alone drives?

I only have a single drive with AHCI and the SATA III.

My MB has RAID and can change BIOS settings but not sure if that would help since I am not RAIDing the drive...Gigabyte GA-770TA-UD3

Yes I saw up to 500 MB/s on each drive in AHCI mode and RAID mode with no raid array set up.

Also as said before, the sata 3 controller on your motherboard is not the same as mine. My controller is part of the AMD SB850 (which your board does not have) and your sata 3 controller is from marvel. So my settings will not be applicable to your system.