Vertex 2 60 AS SSD Bench

vroom_skies

[H]ard|Gawd
Joined
Sep 25, 2005
Messages
1,390
Could I get some feedback on this. Unless I'm mistaken (I haven't had the time to research it much) this can't be right.

SSDBench.jpg


- Win 7 64
- CMD says TRIM is enabled
- SSD Tweaker (auto configure)
- Auto defrag is off

Thanks in advance.
 
uncompressable data on sandforce controler = those numbers. hammering the drive with constant benchmarks will make the drive slower. looks about normal to me.
 
Sandforce's quoted specifications are inflated. A benchmark like AS-SSD gives you an idea of the actual performance you will obtain with realistic data.

Also, the Sandforce controllers will, after some undocumented amount of writes, lower the performance even further, and the performance cannot be recovered even with TRIM. OCZ says that the performance may take a week to recover.

In the future, you might want to avoid Sandforce SSDs. If you want a reliable SSD that performs as the manufacturer specifies, with consistent performance, I suggest Intel X25M SSDs. If you want a little bit higher performance than the Intel (which is already quite good), and also consistent performance, then you might want to look at the Crucial C300 series.
 
The Samsung 470 (New samsung controller) should also outperform the Intel shouldn't it? The writes on C300 drives are quite low until you get up into the 128MB+ range which gets pricey.
 
There have been few reviews of the Samsung 470 series. After I read the storagereview take on it, I thought it looked pretty good for anyone who needs high sequential write speeds. But techreport just did a review

http://techreport.com/articles.x/20087

and I am disappointed to see that the Samsung 470 performance is inconsistent. It seems that it has a problem that looks similar in symptoms to the Sandforce, namely that after being heavily used, the performance drops, and it cannot be recovered even with TRIM.

For good, consistent performance, I think that the best choices are Intel X25-M or Crucial C300.
 
So both the Sandforce and Samsung 470 controllers are essentially trimming at a very slow pace? I think that's what I'm reading in the TechReport brief but maybe I'm missing something.

Doesn't it seem like a firmware flash to decrease the length of time between TRIM/clearing would help the issue? Seems like an odd problem with both the SF and Samsung controllers that could be avoided or alleviated...
 
It is not clear exactly why Sandforce and Samsung 470 drives have their performance degrade, even after TRIM. From OCZ, I have heard both that Sandforce's garbage collection implementation is not very aggressive, and that Sandforce deliberately throttles the performance of the drive for about a week after heavy use.

For Samsung 470, you know as much as I do, after reading the techreport article. As they said in the article, Samsung would not answer their questions.

My guess, and it is only a guess, is that the Samsung 470 achieves its high sequential write speeds only when it has a lot of pre-erased flash blocks to write to. After the drive has been heavily used, there will be few pre-erased blocks. Even if the drive has been sent TRIM commands, the drive still has to use some sort of garbage collection algorithm to collect the unused ("invalid") pages together into block-sized groups and then pre-erase the blocks. Evidently, any GC that the Samsung 470 has was not doing its job after the techreport workload, even after several hours. If the Samsung 470 does have idle-time GC, then perhaps it could be tweaked to be more aggressive. But I would not count on that happening, since Samsung has always been quite conservative with their SSD firmware updates.
 
Thanks for the info you guys.
Glad to know those numbers are in the average ballpark. I'm not really concerned with performance as the drive is running quite nice, I was just concerned with the numbers at first. I know that the manufacturers numbers are always inflated, but I thought what I was getting was a bit low.

Thanks for clearing that up for me. I don't have spare cash to invest in a new SSD, but this was my laptops HDD and is only being used in my desktop as a temp fix. So later down the line I'll grad a "real" one for my tower.
 
If the write performance does not come back up to 70+ MB/s after a week of light use (no benchmarking for the week, since Sandforce drives don't like to be benchmarked, OCZ says), then you can bump this thread with a new benchmark post.
 
Seems low to me.


Less than a week experience with SSDs here and I have learned that they are very touchy. Seems like set up has to be perfect to get the Vertex 2 to perform right.
 
If the write performance does not come back up to 70+ MB/s after a week of light use (no benchmarking for the week, since Sandforce drives don't like to be benchmarked, OCZ says), then you can bump this thread with a new benchmark post.

Understood.
This was only the second time the SSD was benched. I ran it before SSD Tweaker and then once after. The benchmark after was a smidgen faster, but nothing huge. What's odd is that the bench mark results were a little faster, but system responsiveness felt slower.
 
Back
Top