Why does my Intel based SSD have such slow random read/write performance?

996GT2

Diamond Member
Jun 23, 2005
5,212
0
76
I have a 40GB Kingston SSDNow V series, which is based on the 34mm Intel G2 architecture. With half the NAND of an Intel X-25M 80GB, random 4k read/write performance should be similar to the actual Intel drives. However, when I benchmarked mine in AS, the performance seemed to be MUCH lower than what AT and other sites were getting:

My drive has 30% free space.

BeforeTonyTRIM.png


Compared to AT's results:

randomwrite.png


randomread.png


Am I doing something wrong here? This is my second SSD (first was an OCZ Summit), and both times I'm not getting the performance that reviews show.
 
Last edited:

deanx0r

Senior member
Oct 1, 2002
890
20
76
I posted a similar problem in another thread. But I am in the same boat as yours. My results don't quite match Anandtech's results in the random read/write department. Can someone shed some light on this?

4160398367_6304e87568_o.jpg
 

wgoldfarb

Senior member
Aug 26, 2006
239
0
0
From your screen shots, it seems as if both of your drives are in IDE mode, instead of AHCI. I believe that to get the most performance out of these drives you should be in AHCI mode.
 

deanx0r

Senior member
Oct 1, 2002
890
20
76
Negatory, I am running this drive in AHCI mode.
I've no idea why the benchmark shows IDE.
The motherboard is an Asus P5B Deluxe, running AHCI Enhance mode on the ICH8R. I haven't tried it on the JMicron controller (not planning to).
 
Last edited:

Voo

Golden Member
Feb 27, 2009
1,684
0
76
I'm not sure where you're seeing problems but imho the AS SSD screen looks as it should for a 160gb G2 with firmware update. I think anand never posten pictures of that particular benchmark and I wouldn't try to compare different benchmarks just by the numbers..
 

EarthwormJim

Diamond Member
Oct 15, 2003
3,239
0
76
So this is a used drive?

Could need a refresh or whatever it is you do to non-trim supporting drives.

Why don't you try out iometer, that's what Anand used.
 

wgoldfarb

Senior member
Aug 26, 2006
239
0
0
Negatory, I am running this drive in AHCI mode.
I've no idea why the benchmark shows IDE.
The motherboard is an Asus P5B Deluxe, running AHCI Enhance mode on the ICH8R. I haven't tried it on the JMicron controller (not planning to).

I also have the P5B deluxe, and I am also in AHCI enhanced on the ICH8R (my JMicron controller is disabled in the BIOS). When I run AS SSD it shows "msahci" as the mode. I assume "msahci" refers to Microsoft AHCI drivers, which I am using (I did not install Intel's Matrix Storage Manager drivers because TRIM doesn't work if you do that). So, the fact that AS SSD shows "IDE" for you might be part of the problem -- perhaps it is an issue with the driver? Did you install Intel's Matrix Storage Manager driver?
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
you know your drive is working in ACHI mode because the 4k-64thrd results are not the same as your 4k results. ACHI is what makes there be a difference.
 

996GT2

Diamond Member
Jun 23, 2005
5,212
0
76
I have my Thinkpad T61's BIOS set to AHCI mode for the SATA controller, and I'm pretty sure Windows 7 pre-installed the MS AHCI drivers since it shows "Standard AHCI 1.0 Serial ATA Controller" in Device Manager.

That said, though, what am I doing wrong to get such low performance? The random read/write performance of my Kingston drive is really no better than an Indilinx drive (it actually may be worse). This is very strange since the big selling point of the Kingston drive was that it used the 34nm Intel G2 architecture.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
59
91
I don't know of any laptop setups that actually deliver the kind of sata efficiency that desktop counterparts deliver. Same with ram, you rarely see aggressive ram bandwidth and timings on a laptop.

If you want to see if your Kingston performance is really lacking I'd recommend first testing it on a desktop rig to confirm the bench results aren't just being limited by your laptop's design.
 

Rubycon

Madame President
Aug 10, 2005
17,768
485
126
I don't know of any laptop setups that actually deliver the kind of sata efficiency that desktop counterparts deliver. Same with ram, you rarely see aggressive ram bandwidth and timings on a laptop.

If you want to see if your Kingston performance is really lacking I'd recommend first testing it on a desktop rig to confirm the bench results aren't just being limited by your laptop's design.

Notebook chipsets prefer (energy) efficiency over performance. There are gaming type notebooks that practically rival desktop performance and their (lack of) battery run time proves it!
 

996GT2

Diamond Member
Jun 23, 2005
5,212
0
76
I don't know of any laptop setups that actually deliver the kind of sata efficiency that desktop counterparts deliver. Same with ram, you rarely see aggressive ram bandwidth and timings on a laptop.

If you want to see if your Kingston performance is really lacking I'd recommend first testing it on a desktop rig to confirm the bench results aren't just being limited by your laptop's design.

Hmm, my T61 is a fairly recent Thinkpad with the Intel 965M + ICH8R chipset, a 2.4 GHz Penryn C2D and 4 GB RAM. Would that setup really give about 1/2 the random 4k read/write performance of a desktop?

Also, since notebooks are optimized for lower power consumption, would plugging it into AC power and turning the power management to "Max performance" affect anything?
 

EarthwormJim

Diamond Member
Oct 15, 2003
3,239
0
76
Hmm, my T61 is a fairly recent Thinkpad with the Intel 965M + ICH8R chipset, a 2.4 GHz Penryn C2D and 4 GB RAM. Would that setup really give about 1/2 the random 4k read/write performance of a desktop?

Also, since notebooks are optimized for lower power consumption, would plugging it into AC power and turning the power management to "Max performance" affect anything?

It does on my netbook.
 

jimhsu

Senior member
Mar 22, 2009
705
0
76
Please note anand uses 4K asynchronous writes with a queue length of 4 (I think it's 4). If the Kingston drive is somehow not as capable due to lack of channels (see below), perhaps queue lengths above 4 do not help further increase speeds (as opposed to real Intel drives which plateau at 32 or so). That explains about 2/3rd of the variability.

The other 1/3rd probably still has something to do with free space. I don't know how much unallocated space there is on a 40GB kingston, but if it's less than the X-25M percentage wise, then 30% free space might not be enough to get maximum performance (the intel starts suffering performance related issues at 20%).

Also note that half the NAND means half the # of channels (presumably). Channels are directly related to performance. That of course doesn't explain why you don't get anand's results. http://www.futurelooks.com/wordpres...ngston_SSDNow_V-Series_40GB_Boot_Drive_09.jpg
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,786
136
Hmm, my T61 is a fairly recent Thinkpad with the Intel 965M + ICH8R chipset, a 2.4 GHz Penryn C2D and 4 GB RAM. Would that setup really give about 1/2 the random 4k read/write performance of a desktop?

Also, since notebooks are optimized for lower power consumption, would plugging it into AC power and turning the power management to "Max performance" affect anything?

Your sequential reads are only 130MB/s. That's in line with SATAI-150 speed. You should be getting 170MB/s on the 40GB Kingston drive(yea the Intel based one). I think if you search about X25-M and the Thinkpad T series the issue with supporting only SATA-150 is shown.

deanx0r's result seems fine. Intel controllers are known for their random writes, not random reads. If you use CrystalDiskMark you'll see similar ~20MB/s random 4K reads.

For 996GT2's, result I think its your laptop.
 

996GT2

Diamond Member
Jun 23, 2005
5,212
0
76
Your sequential reads are only 130MB/s. That's in line with SATAI-150 speed. You should be getting 170MB/s on the 40GB Kingston drive(yea the Intel based one). I think if you search about X25-M and the Thinkpad T series the issue with supporting only SATA-150 is shown.

deanx0r's result seems fine. Intel controllers are known for their random writes, not random reads. If you use CrystalDiskMark you'll see similar ~20MB/s random 4K reads.

For 996GT2's, result I think its your laptop.

Yeah I knew that all Thinkpads prior to the T400/X200 series were capped at SATA I speeds even though the chipset could support faster. However, since the SATA bus still supports AHCI, NCQ, and all of the other technologies, would it really make my random 4k read/write performance so low? I mean I'm literally getting about 1/3 of what most review sites are getting...
 

fletch101

Junior Member
Nov 15, 2009
4
0
0
18-22 MB/s for 4k random reads is what I see from end users benching with Crystal and AS SSD.

It's also similar to my results with a G2 X25 160GB, I am coming in at like 20.85 on average.

The speeds shown there must be the result of the specific IO meter setup. Maybe you should try to test with the exact same setup and let us know if that explains the disparity you are seeing.

But I think you will see that those numbers are the result of his test setup.
 

chrisf6969

Member
Mar 16, 2009
82
0
0
run Crystal Mark as I've seen more results from that benchmark

Have you run the Intel SSD Toolbox to TRIM your drive?
 

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
18-22 MB/s for 4k random reads is what I see from end users benching with Crystal and AS SSD.

It's also similar to my results with a G2 X25 160GB, I am coming in at like 20.85 on average.

The speeds shown there must be the result of the specific IO meter setup. Maybe you should try to test with the exact same setup and let us know if that explains the disparity you are seeing.

But I think you will see that those numbers are the result of his test setup.

Thats odd. I used Crystal with my 80GB G2 and got scores right in line with the graphs at the top of the thread. ~39MB Randon write and ~60MB random read. Was TRIM enabled on the drives you tested?
 

fletch101

Junior Member
Nov 15, 2009
4
0
0
Thats odd. I used Crystal with my 80GB G2 and got scores right in line with the graphs at the top of the thread. ~39MB Randon write and ~60MB random read. Was TRIM enabled on the drives you tested?

Exar-

Can you run AS SSD Bench and tell me your results on default random read 4k?

I'll dig up some links to what I was referring to in my response.

As for TRIM, the answer is no. However, it was a brand new drive at the time I benched it. It only had Win 7 x64 and drivers at the time. No games or apps were installed yet, so it should not have been suffering from SSD slow down yet. Unless, of course, it was a refurb that the vendor snuck in on me. Possible, I guess.

For your Crystal, did you do anything different or just the default settings? I just ran it here. 22.77 read and 67.08 write for the 4k test. I have read 1-2 posts where people mention that they don't believe the Crystal results for reads on the X25 G2, but I've yet to be able to refute them in personal experience while using it or other benches (specifically AS SSD).

Let me know, please.
 
Last edited:

SolMiester

Diamond Member
Dec 19, 2004
5,330
17
76
Where can I get AS SSD, I have googled but got nothing, I would like to test an A-Data 64GB SSD
 

fletch101

Junior Member
Nov 15, 2009
4
0
0
Exar-

Can you run AS SSD Bench and tell me your results on default random read 4k?

I'll dig up some links to what I was referring to in my response.

As for TRIM, the answer is no. However, it was a brand new drive at the time I benched it. It only had Win 7 x64 and drivers at the time. No games or apps were installed yet, so it should not have been suffering from SSD slow down yet. Unless, of course, it was a refurb that the vendor snuck in on me. Possible, I guess.

For your Crystal, did you do anything different or just the default settings? I just ran it here. 22.77 read and 67.08 write for the 4k test. I have read 1-2 posts where people mention that they don't believe the Crystal results for reads on the X25 G2, but I've yet to be able to refute them in personal experience while using it or other benches (specifically AS SSD).

Let me know, please.

Here's some links to what I was talking about with the low to mid 20's for random 4k read:
http://www.legitreviews.com/article/1022/11/
http://www.hardforum.com/showthread.php?t=1439930
 
Last edited:

exar333

Diamond Member
Feb 7, 2004
8,518
8
91
Exar-

Can you run AS SSD Bench and tell me your results on default random read 4k?

I'll dig up some links to what I was referring to in my response.

As for TRIM, the answer is no. However, it was a brand new drive at the time I benched it. It only had Win 7 x64 and drivers at the time. No games or apps were installed yet, so it should not have been suffering from SSD slow down yet. Unless, of course, it was a refurb that the vendor snuck in on me. Possible, I guess.

For your Crystal, did you do anything different or just the default settings? I just ran it here. 22.77 read and 67.08 write for the 4k test. I have read 1-2 posts where people mention that they don't believe the Crystal results for reads on the X25 G2, but I've yet to be able to refute them in personal experience while using it or other benches (specifically AS SSD).

Let me know, please.

Sorry, didn't see this until today; I will run when I get home from work and post results.