New x25-m firmware

[Glenn]

Here are my results using two X25M 80GBs in Raid 0. I just cloned my Raid Volume to an extra Sata drive I had sitting here and then cloned back after the firmware update. I thought I had some good numbers before but the new ones are way better!

Just to test real world use, I transferred a document file of 1.8GBs with 3820 Files in 280 Folders and got about 55MB/s average on the transfer.

BEFORE
--------------------------------------------------
CrystalDiskMark 2.2 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 428.340 MB/s
Sequential Write : 96.821 MB/s
Random Read 512KB : 273.375 MB/s
Random Write 512KB : 68.290 MB/s
Random Read 4KB : 18.223 MB/s
Random Write 4KB : 44.069 MB/s

Test Size : 100 MB
Date : 2009/04/14 15:19:48


AFTER
--------------------------------------------------
CrystalDiskMark 2.2 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 559.376 MB/s
Sequential Write : 139.647 MB/s
Random Read 512KB : 282.363 MB/s
Random Write 512KB : 111.899 MB/s
Random Read 4KB : 18.395 MB/s
Random Write 4KB : 73.922 MB/s

Test Size : 100 MB
Date : 2009/04/14 16:41:58

 

[alcoholbob]

Hmm this is odd, does this mean there's something wrong with Anand's numbers? He was reporting 50MB/s random reads.

 

[Glenn]

I don't think anand was using crystaldiskmark to get those numbers.

Originally posted by: Astrallite
Hmm this is odd, does this mean there's something wrong with Anand's numbers? He was reporting 50MB/s random reads.

 

[alcoholbob]

Could it really be that far off with a different program? The Vertex's CrystalDiskMark performance is identical to what Anand was getting, but the X25-M numbers is 1/3 of Anand's.

 

[Glenn]

I just did a random read test using Everest's disk bench and got 440.4 MBs so I guess there is quite a difference between different test suites

 

[alcoholbob]

http://img16.imageshack.us/img...492/attobenchmarks.png

I got 30MB/s for random read while benchmarking from Windows 7 64bit, SATA channel in AHCI mode.

I got 15MB/s for random read while benchmarking from Windows Vista Ultimate (32bit), SATA channel in IDE mode.

So I don't know, either it has something to do with Native Command Queuing, or Windows 7 is somehow messing up bench numbers.



Edit: disregard the file name...

And I'll post my X25-E numbers tomorrow, after I reinstall everything thanks to the new Vertex firmware...

 

[IntelUser2000]

Originally posted by: Astrallite
That's amazing.

Well My Vertex gets 30MB/s, which is pretty close to what Anand got (32MB/s).

The X25-M is 55MB/s, and the X25-E is 57MB/s. The Velociraptor is just under 1 MB/s.

I understand everything is relative here, but 55 times faster than a Raptor and 60% faster than the second fastest competition is "generally" considered fast.

Are you perchance, running something like this at home?

http://www.youtube.com/watch?v=96dWOEa4Djs&fmt=22

If that's the case I would absolutely agree with you--if I was in your shoes, all of this is not only ordinary, but trifling.

It does not matter. The JMicron controller based SSDs get random read numbers that achieves 16.2MB/s or "ONLY" more than 16x faster than a Raptor. Reads are nothing exceptional its the writes that are.

 

[vjbelle]

I have two X25-M disks and one of them would not flash. The only solution is to have the drive replaced by Intel. They were very interested in the failed drive and overnighted me a replacement which had the old firmware but flashed properly.

 

[alcoholbob]

Originally posted by: IntelUser2000

Originally posted by: Astrallite
That's amazing.

Well My Vertex gets 30MB/s, which is pretty close to what Anand got (32MB/s).

The X25-M is 55MB/s, and the X25-E is 57MB/s. The Velociraptor is just under 1 MB/s.

I understand everything is relative here, but 55 times faster than a Raptor and 60% faster than the second fastest competition is "generally" considered fast.

Are you perchance, running something like this at home?

http://www.youtube.com/watch?v=96dWOEa4Djs&fmt=22

If that's the case I would absolutely agree with you--if I was in your shoes, all of this is not only ordinary, but trifling.

It does not matter. The JMicron controller based SSDs get random read numbers that achieves 16.2MB/s or "ONLY" more than 16x faster than a Raptor. Reads are nothing exceptional its the writes that are.

As The Rock used to say "It DOESN'T MATTER WHAT YOUR NAME IS!"

Would you agree that the difference between an 8800GT and a GTX280 are neglible? They are both significantly faster than a Geforce 6.

 

[shabby]

Here's a review of the new firmware http://pcper.com/article.php?aid=691 amazing to say the least.

 

[taltamir]

interesting that pcper concludes that the x25 is capable of much greater speeds that 80MB/s, but simply has a hard cap in the firmware... hacked firmware anyone?

 

[coolVariable]

On an intel x60t (SATA2 speed is capped to SATA1):

CrystalDiskMark

Before:
Read
Sequential 131.9MB
512k 123.5MB
4k 17.21MB

Write
Sequential 49.28MB
512k 45.04MB
4k 24.28MB

After
Read
Sequential 132.8MB
512k 124.1MB
4k 15.14MB

Write
Sequential 65.61MB
512k 67.95MB
4k 31.55MB

 

[aceO07]

Originally posted by: coolVariable
On an intel x60t (SATA2 speed is capped to SATA1):

CrystalDiskMark

Before:
Read
Sequential 131.9MB
512k 123.5MB
4k 17.21MB

Write
Sequential 49.28MB
512k 45.04MB
4k 24.28MB

After
Read
Sequential 132.8MB
512k 124.1MB
4k 15.14MB

Write
Sequential 65.61MB
512k 67.95MB
4k 31.55MB

Yea. Sucks. I have a x61t and the crippled SATA-II screwed x/t 60/61 thinkpad owners for the new SSDs.. I think it's limited to less than even SATA-I. The thinkpads forum had more details.

 

[Idontcare]

Originally posted by: taltamir
interesting that pcper concludes that the x25 is capable of much greater speeds that 80MB/s, but simply has a hard cap in the firmware... hacked firmware anyone?

Intel sandbagging/handicapping the performance of their products to force an artificially larger product differentiation gap to go along with the commensurately large ASP gap? Gasp! Say it ain't so! They've never done this before, ever. Have they?

 

[alcoholbob]

Reminds me of pre-laser etching when lower SKU'd video cards (ATI and nvidia both) had fully functional, but BIOS locked processors.

 

[taltamir]

this... this really intrigues me. The amount of leeway suggested by the results indicates that it is SEVERELY handicapped by the firmware... I really wanna see what it is like unlocked!

 

[n7]

We need some modded firmware now

 

[IntelUser2000]

Originally posted by: Idontcare

Originally posted by: taltamir
interesting that pcper concludes that the x25 is capable of much greater speeds that 80MB/s, but simply has a hard cap in the firmware... hacked firmware anyone?

Intel sandbagging/handicapping the performance of their products to force an artificially larger product differentiation gap to go along with the commensurately large ASP gap? Gasp! Say it ain't so! They've never done this before, ever. Have they?

Yea and they compare it to X25-E. The truth is its not going to do much if you just basis it on X25-E having faster writes. X25-E is SLC. SLCs are naturally faster.

And it still stutters slightly. You can see on the graph that the fluctuations are great. On the previous so-called "slow" firmware, the difference between the drop and the fastest weren't too big. When it does 80/10/80/10/80/10 you'll notice it.

The difference this firmware makes is to have sure catastrophic slowdowns don't occur. You will no longer see 15MB/s sequential writes.

 

[Ourasi]

Originally posted by: IntelUser2000
The strength of the X25 drives are not random reads AND writes its just random writes. Random 4K reads of the drives are nothing exceptional.

Originally posted by: Astrallite
So you applied firmware without reinstalling?

And whats with your 4k random read performance? It looks terrible. You should be getting >50MB/s, not 16-18. My used Vertex runs at 30 which is right there with Anand's benchmarks (32). The 4K random reads and writes DEFINE the X25 series. Even my USB flash drive is getting 11MB/s.

CrystalDiskBench, uses random 4KB reads/writes with a queue of 1, which is not at all probable in a windows environment, it is an extremely linear load and can at best serve as a test measuring average random access times. So using Crystal as a measurement for random read/writes 4kb serves no purpose at all.. A depth of 3-4 on the other hand, can represent the most rudimentary of operations such as, say, loading up Windows calculator, and spikes of 30-50 I/Os when loading applications are very common these days, and are categorized as light to moderate workload.

This is the reason why different benchmarks show totally different scores on 4KB Random Read/Write and why people get all confused by the different results. Queue of 1 does not tell the real story as it is not representative of any real kind of access. A queue of 3-4 will atleast show a real workload, even if it's a very light one. Anand's IOMeter 4kb random read uses Queue Depth of 3, and look at the gap that emerges from a real test, in an application load situation with QD@~64 the Intel is 2-300% faster.... The higher the Queue depth, the bigger the gap between X25-M and Vertex becomes, the truth is out there, you just have to know where to look...

 

[coolVariable]

Why, oh why, are you reviving an old thread?

 

[alcoholbob]

You are right about queue depth.

With 1 thread, 4K random reads:
60G OCZ Vertex EX - 35MB/s
64GB Intel - 22MB/s

64 Threads, 4K random reads:
60GB OCZ Vertex EX - 85MB/s
64GB Intel - 150MB/s

 

[shabby]

A new ssd benchmark program popped up recently that tests 64 threads in random 4k read/writes http://www.hardwareluxx.de/com...howthread.php?t=617226
My x25-m is only pulling 20mb/sec in it, probably due to not having ahci enabled?

 

[Ourasi]

Yes, not having AHCI enabled will severly limit the performance in real QD#3-256 scenarious, as seen on many Intel drives in the same configuration..

 

[alcoholbob]

http://img38.imageshack.us/img38/2430/slccompare.png

When I ran the Vertex EX and Intel X25-E in IDE mode, the 64 thread 4K reads on the Vertex EX remained at 80MB/s, while the Intel dived to about 40MB/s. With AHCI mode, the Intel is up to150MB/s.

This tells me that the Intel SSD relies quite a bit on NCQ for performance.

 

[shabby]

Originally posted by: Astrallite
This tells me that the Intel SSD relies quite a bit on NCQ for performance.

Here you can see just how much it relies on ncq, other ssd's see no gains at all.