Sign in to follow this  
TSullivan

256GB Crucial m4 Review Discussion and Test Lab Preview

Recommended Posts

The 256GB Crucial m4 just showed up on our doorstep a bit ago, so I figure what better to do than post about it and start sharing some performance stats ;)

Since we will have some lag between benchmarks being finished and charts being created, I am going to just dump our IOMeter files here for those who are curious. Any questions or comments on this drive will be answered as I see them.

Share this post


Link to post
Share on other sites

We haven't made the transition to the LSI 9211 yet, but here are some shots of the m4 through that. The review will still be using the standardized tests through the 9260 for the 6.0Gbps speeds.

post-70131-0-76879800-1300918294_thumb.p post-70131-0-64648800-1300918293_thumb.p

Share this post


Link to post
Share on other sites

Nice, although a bit disappointed in general of the 4KB results.

(expectations were too high I guess)

I'd appreciate if you could create a zip file of the csv files from your iometer testing.

Share this post


Link to post
Share on other sites

Will do, right now our main benchmarks are wrapped up, and we ran into some surprising results from our real-world tests. Verdict so far is under certain conditions (Productivity) I/O speeds plumment to levels below even the C300. We are in the process of talking with Crucial to find out what might be causing this. In other tests like Gaming, scores are through the roof, above the 510 and just slightly below the Vertex 3. On the scores that are down... its not pretty at all. In some cases its 1/3 of the Intel 510.

Share this post


Link to post
Share on other sites

They still have the time to make firmware adjustments. (release is set to end of April according to Crucial forums)

Looks like it's using a different rev of the controller, 88SS9174-BLD2.

Share this post


Link to post
Share on other sites

Hi,

I would love to see some IOmeter numbers from the M4 on ICH10R and 9211.

I'm especially interrested in 512B, 4KB, and 8KB random IOPS, pure read/write and mixed, 512B alligned, 4KB alligned, and 8KB alligned where they are equal to or less than block size. This is because of M4's 8KB native page size due to 25nm IMFT NAND. If you have a C300 256GB laying around for comparison, that would be great. And a pure read, pure write, and 70/30 R/W (workstation mix) for 100% random IOs, corresponding to the ATTO layout just for random IOPS.

If you want, i can make the IOmeter file and post it here as a zip.

Also, i would like to see torture test degrading by time for the M4, and for C300 for comparison if possible.

4KB, 100% random, 100% write, 4KB alligned, QD 32, entire LBA space for test file. 1 Second resolution for first minute to look at burst, then 1 minute resolution for an hour for sustained. If degrading of sustained random write is still going on at 1 hour, then 15 minutes resolution for an over-night test.

The C300 proved to be really good at sustained random writes over longer time spans. Where opponents dropped to steady-state at 5-30 minutes, it took hours for C300.

*when torture-testing SandForce SSDs, make sure to use latest IOmeter build with incompressible data, or it will just de-dupe the empty data.

I know these are time consuming requests, so i totally understand if you don't feel like it.

Share this post


Link to post
Share on other sites

They still have the time to make firmware adjustments. (release is set to end of April according to Crucial forums)

Looks like it's using a different rev of the controller, 88SS9174-BLD2.

April 26th, and yes, that's the right processor.

Share this post


Link to post
Share on other sites

Hi,

I would love to see some IOmeter numbers from the M4 on ICH10R and 9211.

I'm especially interrested in 512B, 4KB, and 8KB random IOPS, pure read/write and mixed, 512B alligned, 4KB alligned, and 8KB alligned where they are equal to or less than block size. This is because of M4's 8KB native page size due to 25nm IMFT NAND. If you have a C300 256GB laying around for comparison, that would be great. And a pure read, pure write, and 70/30 R/W (workstation mix) for 100% random IOs, corresponding to the ATTO layout just for random IOPS.

If you want, i can make the IOmeter file and post it here as a zip.

Also, i would like to see torture test degrading by time for the M4, and for C300 for comparison if possible.

4KB, 100% random, 100% write, 4KB alligned, QD 32, entire LBA space for test file. 1 Second resolution for first minute to look at burst, then 1 minute resolution for an hour for sustained. If degrading of sustained random write is still going on at 1 hour, then 15 minutes resolution for an over-night test.

The C300 proved to be really good at sustained random writes over longer time spans. Where opponents dropped to steady-state at 5-30 minutes, it took hours for C300.

*when torture-testing SandForce SSDs, make sure to use latest IOmeter build with incompressible data, or it will just de-dupe the empty data.

I know these are time consuming requests, so i totally understand if you don't feel like it.

Shoot me the file at tsullivan@storagereview.com and I will see what I can do. The longer tests might be a bit of a problem, since we are talking a ton of data flowing to the drive. If we can snag another sample though it would ease some of our concerns.

Share this post


Link to post
Share on other sites

How's about the review :)

Seriously, I've got a few m4's and I'm having some strange performance issues having performed random writes.

What was the issues that kept you from publishing?

Share this post


Link to post
Share on other sites

Our review is back on track, benchmarks are off to our graphics guy, and is in the writing phase now.

The main point of contension surrounded our productivity trace. That was originally recorded from a Sony TZ notebook running Vista. It was an awesome trace that stressed drives in ways more other tests couldn't, although it had non-4k alignment.

Up until now every SSD running that trace had no problem handing the 512-byte alignment, and took it in stride. This can be seen in every review we have published so far. Now with the move to 8k-pages on the m4/c400, Micron dropped the legacy 512-byte performance caching in favor of stronger 4k/8k performance. This caused the m4 to see ~30-40% drops in that trace compared to the C300 and perform well under other next-gen drives. In our 4k aware traces it didn't have this problem and surpassed the C300 in performance. On a 4k-aware OS there shouldnt be performance issues on this drive, in fact it should be stronger than the C300. On legacy platforms though you would be much better off getting another drive like the C300.

What sort of problems were you running into?

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this