Overclock.net banner

1 - 5 of 5 Posts

1,120 Posts
Discussion Starter #1
So I'm writing this to document my experiences with installing my new Vertex 2 60GB SSD on my EVGA nForce 790i Ultra motherboard.

To begin, the drive in question is what I believe to be, from all research I've done and things I've checked, the 34nm Vertex 2. The firmware was upgraded to the latest version, 1.33.

My motherboard itself only supports SATA 3Gbps, or SATA II.

And our story begins:

I purchased my SSD during a Boxing Week sale and only got around to installing it a few days ago. In between now and the purchase time I had a few problems, the first being the possibility of possessing a 25nm Vertex 2. That was proved false, but then I was presented with the problem that my motherboard does not support AHCI mode and thus, it would seem, TRIM.

Some quick research proved this to be false, as I could still have TRIM if I used Microsoft's default IDE drivers. I was also told that I could use JMicron's SATA port on my motherboard as it did support AHCI. The drawback, however, is lack of dependabillity -- or so people say.

I was met with a bevy of options and no clear ideas as to what was going on, so my only choice was to run everything through a series of test installs and benchmarks to decide which option was the best.

The SSD was properly wiped and aligned. I began with a Windows 7 install (Ultimate edition). I then uninstalled the existing NVIDIA SATA controllers and replaced them with the Microsoft IDE controllers. Then I installed the newest nForce drivers (v15.58) from the NVIDIA website and tested. I didn't know my face from my caboose at this point, so I didn't know that what I just did made no sense so this test will be scrapped. I also didn't bench with AS SSD, either, so I have no good benchmarks anyways.

Realizing my mistake, I installed Win7 fresh and replaced the NVIDIA controllers with the MS IDE ones. Nothing else was installed on this computer but every available update from Windows, AS SSD and HDTune Pro. A quick test to see if TRIM was working (CMD -> fsutil behavior query DisableDeleteNotify -- if the number is 0, TRIM is working, 1 means it isn't; it's apparently near impossible to prove if TRIM is functioning or not, but this is one of the closest means) indicated it was. Results from AS SSD gave the following (some numbers rounded up to the nearest whole number), which was indicated to be in MS IDE mode by AS SSD:

Seq. Read: 146mb/s
Seq. Write: 96mb/s
4K Read: 21.76mb/s
4K Write: 64mb/s
4K/64Thrd Read: 25mb/s
4K/64Thrd Write: 80mb/s
Acc. Time Read: --- (didn't register a value)
Acc. Time Write: 0.220ms
TRIM: yes

Decent numbers, but rather low, seemingly. The only other problem noted was that in MS IDE mode, I could not plug and play with my other HDD's; in other words, I installed the OS with only the SSD plugged in, and then with the computer on plugged in the other HDD's but they wouldn't register.

This was the JMicron run. Everything else was done the same but the default controllers were used as AS SSD informed me things were now running in MSAHCI mode:

Seq. Read: 181mb/s
Seq. Write: 96mb/s
4K Read: 18mb/s
4K Write: 46mb/s
4K/64Thrd Read: 104mb/s
4K/64Thrd Write: 84mb/s
Acc. Time Read: 0.134ms
Acc. Time Write: 0.242ms
TRIM: yes

Better numbers than the last test. TRIM proved to be working. I decided to now installed the nForce drivers just to see, and there was no appreciable difference. The NVIDIA Controllers were then replaced with the MS IDE ones, and no appreciable change again was registered.

Back to the SATA ports. My initial concern was the fact that NVIDIA had no TRIM support through their drivers. However, from my last reformat I used the 15.53 version drivers, whereas the new ones were 15.58 and from May 2nd, 2011 no less. It would seem logical that NVIDIA, with such a late driver release into a market now saturated with SSDs, would include something. I could find no solid release notes to confirm or deny this but I would have to assume that something was done. Their SATA controllers were also supposed to be slower than the basic MS IDE ones, so I could only hope that that was remedied as well.

Everything was cleared again, up until the controllers. They were left at default (which AS SSD informs me was nvstor mode) and tested:

Seq. Read: 183mb/s
Seq. Write: 90mb/s
4K Read: 20.52mb/s
4K Write: 58mb/s
4K/64Thrd Read: 92mb/s
4K/64Thrd Write: 58mb/s
Acc. Time Read: 0.114ms
Acc. Time Write: 0.238ms
TRIM: yes

Next, they were updated to the most recent version (v11.1.0.43) by the 15.58 drivers and then everything was ran (in nvstor64 mode, according to AS SSD):

Seq. Read: 183mb/s
Seq. Write: 53mb/s
4K Read: 20mb/s
4K Write: 56mb/s
4K/64Thrd Read: 91mb/s
4K/64Thrd Write: 53mb/s
Acc. Time Read: 0.118ms
Acc. Time Write: 0.236ms
TRIM: yes

No appreciable differences between the two tests, except the sequential write speeds. Could this be because I'd just run a write test moments before? Possibly. However, no one buys SSDs for their write speeds.

The numbers, however, would seem to indicate that NVIDIA has solved some problems with their original drivers. The numbers compete with that of the JMicron controller in AHCI mode and beats the MS IDE mode numbers handily. And through all the tests, not once was it proved to be that TRIM was not functioning.

My conclusions on the experiment are thus:

TRIM is now roundly supported by nForce chipsets.

Using the NVIDIA SATA controllers trumps the use of the generic Microsoft MS IDE controllers and competes with using JMicron's SATA controller through AHCI mode.

The numbers still are not stellar, but for a previously unsupporting motherboard on an old LGA 775 socket, the numbers aren't too bad. The drive, or the drive on this motherboard, seems to lack in the write department, but again, that's not why people usually purchase SSDs.

Aside: my internet connection would occasionally drop at time of high usage and would take forever to load up upon restart or resume from sleep. I found that by not installing NVIDIA's ethernet driver, these problems were fixed.

To compare my data to other data, I'm not going to go too all-out. In the June edition of Maximum PC, they tested a bunch of new SSDs and had some numbers on the Vertex 2 100GB for comparison. This was running through SATA 6Gbps (SATA III), and here are the AS SSD numbers:

Seq. Read: 209mb/s
Seq. Write: 72mb/s
4K Random Read: 5249 IOPS
4K Random Write: 12910 IOPS
Acc. Time Read: 0.130ms
Acc. Time Write: 0.235ms

The numbers immediately seem a little bit higher. A quick bench on my numbers before writing this showed:

Seq. Read: 185mb/s
Seq. Write: 63mb/s
4K Random Read: 5271 IOPS
4K Random Write: 14612 IOPS
Acc. Time Read: 0.118ms
Acc. Time Write: 0.235ms

Those last numbers running through the nvstor64 mode from test #4, still, after a few days of use. Not bad, and definitely comparable. My write times even went about by about 10mb/s, so it could've been sequential writes that burned me earlier. I'm happy, for now, at least, with the performance of my 775 board and my SATA II. When I do upgrade, I'm sure I'll be even happier, but for now I don't see any incredibly demanding need to ship out a board and a chip that I haven't even come close to pushing to the limit yet.

One final quirky bit, however. In the attached image you'll see a run of AS SSD I did back at the end of March to see whether or not my SSD was running on 25nm. I posted it in the forums here, but I'll share the picture anyways.


Odd. My read times are slightly higher and my access time is very, very low. My write times also blow my other configurations out of the water. The differences that could account for this: the firmware, as you can see, is 1.25. Could this have affected it? Maybe. I'm not an SSD doctor. The other thing was that this SSD was plugged in as a secondary drive, not my primary drive with an OS installed on it. I plugged it into an open port and ran a quick check to see if it needed to be RMA'd or not. Could this account for the difference? You tell me, as I haven't the faintest idea, but am hoping yes. I'm just a man with some numbers that's passing them on to people who hopefully know what to do with them.

And sorry for the lack of pictures. I didn't take any AS SSD screenshots because I never thought to publish this information until after my successful conclusion.

Hope this helped somebody.

1,120 Posts
Discussion Starter #3
Thanks Krogoth. Hopefully this helps other people still hanging onto their nForce's.

166 Posts
nice. how did you check for trim support?

ill give a test with this drivers, but i think my problem is more Bios related.

1,120 Posts
Discussion Starter #5
I stated it very briefly in the write-up. Apparently it's hard to test if TRIM is actually working since it's hard to catch it in the split second it does its thing, but this is the closest method:

- Open up CMD (Start, CMD; run as admin if necessary)
- Type in (without quotations): "fsutil behavior query DisableDeleteNotify"

If the number is 0, TRIM is supposed to be working. If it's 1, it's not.
1 - 5 of 5 Posts