AS SSD Benchmark

[nsa666]

Originally posted by: n7
I have no idea what sort of results i should be getting, but here's my Intel X25-M 80 GB numbers.

http://ense7en.com/pics/Intel%...hmark%2006.28.2009.JPG

it is clearly too slow for an intel ssd..

you should get something like this:
http://www.abload.de/img/asbench4lcju.png

(older version of as ssd bench)

 

[Rubycon]

Originally posted by: nsa666

Originally posted by: n7
I have no idea what sort of results i should be getting, but here's my Intel X25-M 80 GB numbers.

http://ense7en.com/pics/Intel%...hmark%2006.28.2009.JPG

it is clearly too slow for an intel ssd..

you should get something like this:
http://www.abload.de/img/asbench4lcju.png

(older version of as ssd bench)

OK I just ran the older one on the same setup as my previous post...

Text

So I presume like iometer these higher results on Intel SSDs only affect non desktop use (i.e. multiple user random requests).

 

[nsa666]

@n7: you should switch to AHCI mode.

 

[n7]

Already running AHCI mode.

I suspect if i want to fix performance in this bench i would need to run HDD erase & reinstall...

I get these from CrystalDiskMark: http://ense7en.com/pics/Intel%...formance%20enabled.JPG

 

[nsa666]

Strange that you getting different values. Set CDM to 1GB testsize and if it doesnt change try my benchmark again. 25MB/s seq. write is surely wrong.

 

[taltamir]

Originally posted by: Idontcare
Ah yes, its a privilege to bleed a little every now and then when handling hardware on the leading edge of a technology curve. (called the bleeding edge for a reason I suppose :laugh

ugh, i hate when this happens... plus it is really annoying to focus on not bleeding on the hardware.

 

[Rubycon]

Originally posted by: nsa666
@n7: you should switch to AHCI mode.

Mine shows IDE too yet I assure you the drive is running in AHCI mode too. On the Areca it says SCSI.

 

[nsa666]

Originally posted by: Rubycon

Originally posted by: nsa666
@n7: you should switch to AHCI mode.

Mine shows IDE too yet I assure you the drive is running in AHCI mode too. On the Areca it says SCSI.

Not because what it shows, but because of the results. the 64-threads tests is not faster than 1-thread tests->AHCI/NCQ is not working correctly.

 

[alcoholbob]

Originally posted by: n7
Already running AHCI mode.

I suspect if i want to fix performance in this bench i would need to run HDD erase & reinstall...

I get these from CrystalDiskMark: http://ense7en.com/pics/Intel%...formance%20enabled.JPG

Did you install Intel Storage Matrix Manager? I found The Intel X25-series pretty quirky performance-wise without that application installed.

 

[nsa666]

Standart windows AHCI-drivers should acutaly be enough.

 

[alcoholbob]

Maybe it's my chipset (X58) but on XP (IDE), Vista (IDE) and 7 (AHCI), with the X18-M 80G, X25-E 64G and X25-M 160G none of them could complete the random IOPS test with HD Tune Pro 3.5 without Intel SMM installed.

 

[rcpratt]

I ran the benchmark and was unhappy with my write speeds. I then realized I wasn't in AHCI, so I made the required regedits (windows 7) and then changed the settings to AHCI. Then all hell broke loose.

Benchmark

The write performance is now even worse than before. Something is quite wrong. I've updated the firmware. I also installed Intel Matrix Storage Manager. My whole system seems to be running slower, although I could just be paranoid. Also been trying to do a HDDErase, but I'm having issues with that as well. Any ideas?

 

[Old Hippie]

Also been trying to do a HDDErase, but I'm having issues with that as well. Any ideas?

You need to use HDDErase 3.3 for an Intel drive.

 

[rcpratt]

Thanks. I think I tried 3.1 and 4.0 last night, I'll give that a shot tonight.

I ran another one this morning and the write bench was much better (like 50/40/35/~0.100), but it really still feels like my system is launching apps slower than before.

 

[rcpratt]

Originally posted by: Old Hippie

Also been trying to do a HDDErase, but I'm having issues with that as well. Any ideas?

You need to use HDDErase 3.3 for an Intel drive.

How do I burn that? It doesn't have an .iso.

 

[rcpratt]

edit: Got it

 

[rcpratt]

Happy days, finally got HDDErase 3.3 up and running, and correctly switched it to AHCI. The results look spot on: http://img513.imageshack.us/im...benchintelssdsa2mh.png

 

[Old Hippie]

Originally posted by: rcpratt
Happy days, finally got HDDErase 3.3 up and running, and correctly switched it to AHCI. The results look spot on: http://img513.imageshack.us/im...benchintelssdsa2mh.png

Good deal! :thumbsup:

 

[ZeroBurn]

i was under the impression hdderase 4.0 would work if you had the new 8820 firmware?

http://www.pcper.com/article.p...=691&type=expert&pid=9

 

[rcpratt]

I hadn't read that, but I thought I had tried it anyways. Oh well, 3.3 worked just fine.

 

[nsa666]

Update:
1.2.3487.20916
* IOPS-View

 

[Emulex]

i find a good test is the iometer config commonly used in the vmware forum under speed/SAN discussion. it's right up there with the sql server beatdown.

brings my 12 sata and 12 15K SAS san's to its knees but shows great sequential.

also notes the CPU utilized (for fcoe/iscsi/aoe/das/etc).

one dude ran a vertex (or three x25-e) and it smoked many older FC san's not only in sequential but in random iops.

really i think benchmarks are way overhyping the product. in some cases yeah that device is 2-3x faster than a decent WD black 1TB but in some cases it is not.

i'm not sure if the ocz can handle multithreaded NCQ as well. you get a few tasks going and it seems the vertex drops faster than the intel.

either way i love ssd it is super dope. i wish they'd make a IDE(PATA) vertex. i wouldn't need to upgrade a fleet of old machines if they could push ATA-133/100 all day long with near-zero random access compared to the cheapo 7200rpm 2gb drives that are way old and nearing their lifecycle. same with old notebooks. pata notebook with slow azz 5400rpm would really get new life with a stupid fast SSD.

anyone seen anyone market to this segment(PATA)?

 

[jimhsu]

Suggestion: Can you figure out a way to bench interleaved random read/writes? See my thread here:

http://www.hardforum.com/showt...d.php?t=1436928&page=9

The reason is that it's very hard for the controller to handle such a mixed workload - it would be a good stress test. Typically 50% read + 50% write MB/sec is quite a lot worse than either 100% read or 100% write workloads. How much worse depends on the controller.

Some especially innovative companies that make really expensive enterprise SSDs like Sandforce and Violin Memory claim to have solved the problem by "their non blocking RAID RAM cache architecture enables a read to be done from a virtual flash block while an erase-write is still in operation on the same flash block. " (see http://www.storagesearch.com/ssd-law-1.html). Would be interesting to see how the consumer SSDs fare.

 

[Idontcare]

Originally posted by: jimhsu
The reason is that it's very hard for the controller to handle such a mixed workload - it would be a good stress test. Typically 50% read + 50% write MB/sec is quite a lot worse than either 100% read or 100% write workloads. How much worse depends on the controller.

jimhsu, just curious but can you give me some kind of idea of what kind of applications or usage scenarios create these situations where the harddrive controller is handling mixed io like you propose.

(just to be clear I am not challenging your request for the feature, I like the idea, just am curious where life mimics this scenario)

Is an example of this scenario say when I have TMPGenc transcoding a video file where both video source and output file locations are on the same physical drive? Or maybe I have TMPGenc transcoding and at the same time I use nero recode to read another previously transcoded file and burn it onto a DVD?

 

[jimhsu]

Plenty. Mirroring a source code repository, copying over mp3s/isos while opening word, running multiple torrents while doing a virus scan, handling multiple VMs for a multi-user server, etc etc. Basically many many multitasking scenarios.