Bad performance from Intel 160GB SSD X25M G2 for an SSD

[Lord Banshee]

I just got my new work laptop(HP 8450w, i5 ~2.6GHz, 4GB of RAM) and it seems so much faster than my old laptop, but something just did not feel as fast as it should per "wow factor" of ssd.

So i downloaded AS SSD Benchmark and was pretty sad when the results for Seq was 1/4 than what intel advertises for read and are 1/5 for writes! talking 60MB/s for reads and around 18MB/s for writes.

I've updated the Controller driver to the 3/10 Intel Rapid Storage Tech but that only increased my Read time slightly. I will try firmware update on Monday, but this did not seem to help my bosses SSD with is a 80GB G1 drive .

Access times were around 0.250ms which isn't bad but 1/2 what it should be. I have to get back to you on the random 4K and 4K-64 times but i believe they are underwhelming also(but great compared to a normal HD though).

So i am wondering what is causing this performance problem and the only thing i can think of is that we are required to use PGP hard disk encryption on our drives, i know it would make an impact but did not believe it would be this bad. Any body else with PGP or other hard disk encryption software running on an SSD have seen similar things?

oh also this is on Win7 and as far as i know a clean install of win7.

thanks

 

[erdemali]

---------------------------------------------------------------

 

[sub.mesa]

Could we see a screenshot of AS SSD and CrystalDiskMark, with all benchmarks being performed? Specifically, i am interested in 4K Random 32/64 queue scores; so download a recent version of both benchmarks as older versions do not list this.

Post the image on some image site and link it here; imageshack.us for example.

 

[Zap]

So i am wondering what is causing this performance problem and the only thing i can think of is that we are required to use PGP hard disk encryption on our drives, i know it would make an impact but did not believe it would be this bad. Any body else with PGP or other hard disk encryption software running on an SSD have seen similar things?

That's certainly interesting. I'm looking forward to someone doing before/after benchmarks with/without PGP.

In the meantime you can console yourself with the fact that your setup is still faster than a VelociRaptor.

 

[Old Hippie]

Thinkpad T61p

If that's the machine you're using it's definately SATA1 limited but if you check out this thread it seems to be capped a little lower....but not as slow as yours.

After reading more, it seems there's two parts to the linked post. Here's the continuation of the first thread and the first post just about says it all.

I'd try different modes (IDE-SATA-SATA+AHCI) and check the alignment but I have no idea how badly encryption hurts speeds.

Have you tried this update?

Some mobile chipsets never seem to work quite as well as desktops.

 

[Lord Banshee]

Oh the laptop is not the T61p that's my personal laptop.

It is a HP 8450w, i5 ~2.6GHz, 4GB of RAM, etc.. Currently it is setup as SATA-AHCI.

I will run crystalmark and get the screenshots today.

About the alignment, from googling it sounds like i would have to reformat which is pretty risky on the work laptop, is there other means?

 

[Voo]

A AS SSD screenshot would also provide us with the used driver (maybe something went wrong there?), the alignment and if NCQ works - maybe something of that can point us in the right direction, though nothing should have those kinds of performance penalties and I know of several people that had problems with notebooks and SSDs.

And no you don't have to reformat the partition to get it working, there are other ways to ensure that (I'd use GParted and a Linux live cd but I'm sure there's a windows tool that does the same work)

 

[Lord Banshee]

Reran as-ssd again interesting is my SeqWr is now 2x faster ? So weird

The the alignment problem i haven't research it much but the OS partition starts from the beginning of the drive so it should be fine right?

 

[Old Hippie]

The the alignment problem i haven't research it much but the OS partition starts from the beginning of the drive so it should be fine right?

There's a little more to it than that but that green number that says

1024 K - OK

indicates that the alignment is OK.

 

[erdemali]

Can you check if TRIM is enabled?

DisableDeleteNotify = 1 (Windows TRIM commands are disabled)
DisableDeleteNotify = 0 (Windows TRIM commands are enabled)

at the command prompt type the following:

fsutil behavior query disabledeletenotify

 

[Lord Banshee]

Trim is enabled (console reports 0)

I believe i found where the extra seqwr 20MB/s came from on as-ssd. Apparently after running crystaldiskmark for some reason as-ssd gain that performance. Maybe firmware say the 1GB transfer that crystaldiskmark did and optimized for large transfers(?)

Also i found some extra random performance by running the intel ssd toolbox optimizer cleanup.

This is as-ssd right after intel cleanup:

This is crystal disk mark as intel cleanup:

This is as-ssd right after crystal disk mark:

 

[erdemali]

Have you got Crystal Disk Info and can you post a screenshot?
http://crystalmark.info/software/CrystalDiskInfo/index-e.html

 

[Lord Banshee]

 

[erdemali]

I think you were initially right about the encryption software.
Your firmware is not the latest (02HD).

 

[Old Hippie]

I think you were initially right about the encryption software.

That's kinda scary.

I've never used encryption but can you look at the task manager while running the test and see if your CPU usage is high....or wouldn't that mean anything?

 

[Voo]

Hmm ok, we can disqualify alignment and firmware (2cvi has TRIM and all the other goodies like faster sequential write for the 160gb version), the intel driver is also there - strange. Either some problem with the mb or it's really the encryption software, though usually that shouldn't have such a tremendous effect..

 

[Rubycon]

That drive by itself should be >400 in AS SSD.
They scale well in RAID 0 - just four of 'em > 1600 with bad alignment! Even with good alignment the score is not any better.

 

[erdemali]

SATA300 cannot go past 289Mb/s I think

 

[Rubycon]

SATA300 cannot go past 289Mb/s I think

This is not an interface bandwidth problem it's an iops problem. Some notebook sata controllers "hold back" severely with these drives as well. They should ALWAYS be in AHCI mode.

 

[Lord Banshee]

Wow Rubycon that is nice setup.

About the CPU Usage during crystal disk benchmark.. it peeks at 28% for seq/512K/4K and 38% in the 4K QD32 (on a i5 with 2 cores 4 threads)

So it looks like the "encryption software" may be the problem but i will still try and update to the o2HD firmware on Monday and see how it goes (already have a burned disk at work for the iso, don't have an blanks at home)

 

[Lord Banshee]

This blog shows similar finding

http://anthonyvance.com/blog/security/ssd_encryption/

 

[Voo]

Well the blog also used a SF drive, which for encryption (= random data for the controller) loses much of its appeal, but I'm shocked that even the Intel drive loses that much performance.. strange - why? A modern CPU should be able to keep up with the transfer rates, so where's the bottleneck? After all at least for the Intel SSD what kind of data it's writing/reading is completely unimportant..

 

[Old Hippie]

About the CPU Usage during crystal disk benchmark.. it peeks at 28% for seq/512K/4K and 38% in the 4K QD32 (on a i5 with 2 cores 4 threads)

I'd say you found your problem.

 

[Lord Banshee]

I'd say you found your problem.

Just wondering does anyone have cpu usage on ssd during this benchmark without any kind of encryption?

 

[Rubycon]

Just wondering does anyone have cpu usage on ssd during this benchmark without any kind of encryption?

Just tried it and it spikes a bit as you can see in this video but nothing spectacular. It's probably not the most accurate way to do this.

http://www.youtube.com/watch?v=FmtXgbUQ4hU