Corsair Performance 3 128GB.

[Old Hippie]

Can you link me to where yo bought that LSI card, please?

I purchased from Buy.com.

It's got the break-out cable included.

 

[Aristotelian]

Those prices seem quite bad if the reported pricing of the 400's turns out to be true. However, the C400 isn't around yet so I guess they have a captive market.

How is it a captive market if the speeds are reported to be significantly lower than the cheaper Crucial c300 series?

 

[smakme7757]

I wouldn't mind see an ATTO benchmark. On my Force 160gb i get really poor results with AS SSD, but the ATTO benchmark confirms the speeds of my SSD. From what i've read ATTO is a more consistant benchmark for Sandforce based SSDs.

Now i'm not an SSD guy, but my Force gives me on average 270mb reads and writes, but on these other drives it seems you get awesome read speeds, but poorer write speeds? WHy is that and why wouldn't people want to buy an SSD which give decent performance in both categories. The C300 128gb gives 355/140 and the Force 120gb gives 285/275. Why loose 135mb write speed for an extra 80mb read?

 

[Burner27]

I wouldn't mind see an ATTO benchmark. On my Force 160gb i get really poor results with AS SSD, but the ATTO benchmark confirms the speeds of my SSD. From what i've read ATTO is a more consistant benchmark for Sandforce based SSDs.

Now i'm not an SSD guy, but my Force gives me on average 270mb reads and writes, but on these other drives it seems you get awesome read speeds, but poorer write speeds? WHy is that and why wouldn't people want to buy an SSD which give decent performance in both categories. The C300 128gb gives 355/140 and the Force 120gb gives 285/275. Why loose 135mb write speed for an extra 80mb read?

ATTO is not a true benchmarking tool for SSDs (So I have been told)

 

[smakme7757]

ATTO is not a true benchmarking tool for SSDs (So I have been told)

Fair enough, but then with AS SSD favouring other chipsets over the Sandforce chipset it seems we have a problem. There needs to be a benchmark that sets a standard and isn't biased towards certain chipsets.

 

[Yellowbeard]

ATTO is not a true benchmarking tool for SSDs (So I have been told)

By who?

 

[Yellowbeard]

Fair enough, but then with AS SSD favouring other chipsets over the Sandforce chipset it seems we have a problem. There needs to be a benchmark that sets a standard and isn't biased towards certain chipsets.

It's a simple matter of compression. ATTO uses none, the other benchmarks do. Since the SandForce controllers have their own compression, their speeds drop to more "mortal" levels when dealing with certain types of compressed files.

P3 Series SSDs (which are not SandForce based ) are validated using ATTO like pretty much any SSD you'll see anywhere. All of ours back to our first SSD are validated with ATTO. It seems to be a defacto industry standard.

 

[McLovin]

...

P3 Series SSDs (which are not SandForce based ) are validated using ATTO like pretty much any SSD you'll see anywhere. All of ours back to our first SSD are validated with ATTO. It seems to be a defacto industry standard.

If I may ask, why switch from SandForce? It seems like their technology is the up-and-comer with a lot of success, especially with you drives.

 

[Aristotelian]

How is it that in this thread only Old Hippie and I are addressing the drastic difference in performance (and price) between these drives and the Crucial c300 lineup (which appears to be faster and cheaper). If we are mistaken about these benchmarks, could someone correct this?

 

[Burner27]

It's a simple matter of compression. ATTO uses none, the other benchmarks do. Since the SandForce controllers have their own compression, their speeds drop to more "mortal" levels when dealing with certain types of compressed files.

P3 Series SSDs (which are not SandForce based ) are validated using ATTO like pretty much any SSD you'll see anywhere. All of ours back to our first SSD are validated with ATTO. It seems to be a defacto industry standard.

And wouldn't that be a true test of any SSD? How it handles compressed and uncompressed data in a random sequence?

 

[Dadofamunky]

Hmmm... I get basically the same readings as SolMeister does, about 450 synthetic benchmark for my C300.

I get a reading of 350MB throughput on sequential reads. My RAID 0 set hits 260MB/s on sequential reads. HDs are not that much slower than SSDs on sequential reads (although the RAID 0 factor might skew things a little), based on my recent experience with a VMware storage startup company these results bear that out.

What I do find is that SSDs do dominate on random access - C300 around 52 MB/s random writes, RAID 0 set at a brutal 2.69 MB/s! SSDs are 15x faster. And watching it crawl through a 4K random read (Lesen) in that AS SSD benchmark makes me want to take ol' Bessie out behind the barn and shoot her for mercy. Try 0.61 MB/s 4K random reads vs. 28 MB/s in SSDland. Ouch! Spindles really are a limiting technology performancewise.

 

[Yellowbeard]

If I may ask, why switch from SandForce? It seems like their technology is the up-and-comer with a lot of success, especially with you drives.

We are not abandoning SF.

And wouldn't that be a true test of any SSD? How it handles compressed and uncompressed data in a random sequence?

It could be one of many but, no more or less than a standard HDD. I did not say ATTO was a perfect test. However, it is the test used to document the specifications by Corsair and most other SSD makers I have seen. Comparing ATTO scores to other tests that use compressed data is useless. But, this does not invalidate the results of either type of testing.

How is it that in this thread only Old Hippie and I are addressing the drastic difference in performance (and price) between these drives and the Crucial c300 lineup (which appears to be faster and cheaper). If we are mistaken about these benchmarks, could someone correct this?

Are you quoting spex from a specific review or using published spex? I've only glanced at the thread on XS but I would not recommend basing your opinion from 1 person of unknown qualifications. Read the thead. There are several factors there that could have skewed the results.

Example:

Not sf-2000. This is an overclocked marvell controller used on the micron c300 series. Was this drive used just for benching, or did you bench on this drive as the os drive?

Yeah I had to install the OS on it unfortunately

ah, that's why performance numbers looks so low. Not bad for 50% filled. Most reviews only test fresh. Thanks a lot!

I didnt realise it affected it so much, I should've tested it blank when I got it but I was in a rush to put my old SSD in my netbook

 

[bargetrav]

I can confirm (i have a separate thread too) that I'm seeing the same low random 4k/read writes on my P128 Series 3 drive, it does hit 400mbps read / mid 190s write, so it technically performs well based on throughput and my system feels very fast, but i am unsure why the random read performance is so low, now i am more worried as i have duplicated the results of another user.

 

[Aristotelian]

Are you quoting spex from a specific review or using published spex? I've only glanced at the thread on XS but I would not recommend basing your opinion from 1 person of unknown qualifications. Read the thead. There are several factors there that could have skewed the results.

Well, let's put it this way.

A) There are no reviews to be found (which already has me wondering how an item has made it to market without a single review, at all - were drives sent out for review?)
B) Users on multiple forums have run the benchmark (with different specs) and produce consistently lower results using the corsair 3 128GB drive than the crucial c300 128GB.

It isn't really about 'opinion' here, as I'm looking to buy 2 256GB drives any day now, but considering the pc I'm building it feels a bit odd to buy 1 year old tech (although they are the fastest drives available from the measurements I've seen) when other companies are releasing (Corsair) or about to release (Crucial c400) their next generation of drives.

I don't have a solid opinion here yet, I'm simply trying to get to the bottom of this. Perhaps you could let us know why we can't find any reviews of any model of the Corsair Performance 3 series anywhere? If I've missed one somewhere I apologize, but a number of users on this forum are looking for benchmarks (we have a few from users now) and/or reviews (none) in order to compare the different drives.

 

[Yellowbeard]

The review drives just went out VERY recently. Remember, we just announced the drives at CES on or around JAN 6. I am looking for published reviews myself but have not found any yet. They'll be popping up soon.

B) Users on multiple forums have run the benchmark (with different specs) and produce consistently lower results using the corsair 3 128GB drive than the crucial c300 128GB.

If you have any links, I'd like to look into them.

Thanks for considering Corsair SSDs.

 

[Yellowbeard]

I can confirm (i have a separate thread too) that I'm seeing the same low random 4k/read writes on my P128 Series 3 drive, it does hit 400mbps read / mid 190s write, so it technically performs well based on throughput and my system feels very fast, but i am unsure why the random read performance is so low, now i am more worried as i have duplicated the results of another user.

Can you link it please?

1. Is the OS installed on this drive?

2. How much data is on the drive. % wise, how much space is used and how much is empty?

3. What benchmark(s) are you using?

 

[Old Hippie]

I can confirm (i have a separate thread too) that I'm seeing the same low random 4k/read writes on my P128 Series 3 drive, it does hit 400mbps read / mid 190s write, so it technically performs well based on throughput and my system feels very fast, but i am unsure why the random read performance is so low, now i am more worried as i have duplicated the results of another user.

Link?

 

[bargetrav]

Can you link it please?

1. Is the OS installed on this drive?

2. How much data is on the drive. % wise, how much space is used and how much is empty?

3. What benchmark(s) are you using?

PLEASE PM ME!!!! I HAVE ONE OF YOUR P3-128 SERIES 3 DRIVES.

I'm in the same boat as the small group of others. I benchmarked my drive literally with no other program installs, right after a fresh win7ultimate 64 bit install, AHCI on, on a Marvell 9128 sata iii controller.

other users on different setups have reported the same terrible random 4k read/writes like me. any insight you have would be great.

when i benchmarked i had over 100gb out of the 119gb of free space
OS definitely installed on this drive
benchmarks were AS SSD and CrystalDisk

both showed the advertised throughput roughly...

both showed the piss poor random read/writes

 

[bargetrav]

Link?

http://forums.anandtech.com/showthread.php?t=2141510

 

[Old Hippie]

Not sf-2000. This is an overclocked marvell controller used on the micron c300 series. Was this drive used just for benching, or did you bench on this drive as the os drive?

Overclocked Marvell controller? Gimme a break.

Yeah I had to install the OS on it unfortunately

Everybody does. If you couldn't install an OS there'd be little need for SSDs.

ah, that's why performance numbers looks so low. Not bad for 50% filled. Most reviews only test fresh. Thanks a lot!

If that drive performs that bad at 50% filled it doesn't belong in the marketplace.

None of those comments/situations should have anything to do with the performance of that drive.

I find it hard to believe that Corsair would release a new drive, with those specs, at that price, and expect the enthusiast crowd to be enthusiastic.

 

[Yellowbeard]

http://forums.anandtech.com/showthread.php?t=2141510

Have you tested it on the ICH10R or an ICH7R, 8R, or 9R for that matter? The only Marvell testing I have seen so far, the Intel controller smokes it.

 

[jwilliams4200]

My guess on what is going on is that the firmware for some SSDs has been tuned for different priorities than other firmwares.

Specifically, Micron/Crucial seems to tune their firmware to give the best possible 4K/QD=1 and 4K/QD=high performance, while sacrificing some of the sequential performance, particularly sequential write speed.

It looks like Corsair has tuned for the highest sequential write speed, while sacrificing the 4K performance. I think Samsung 470 series SSD made this same tuning choice, since the sequential write numbers look very good, but the 4K numbers, and particularly high QD 4K, look relatively poor.

It kind of makes sense if you think about it as choosing the chunk (stripe) size in a RAID. Since SSDs are essentially operating in an internal RAID (with 4, 5, 8, 10, or 16 channels in most cases), the decision about how to stripe the data among the available channels will affect the performance. If you choose to optimize the striping for high sequential write speed (maybe with large chunks), you are not going to get the best performance for high QD 4KB random writes.

If that is correct, then personally, I think both Crucial and Corsair got it wrong. I would prefer something in between the two extremes. Say, a 128 GB SSD with 160 MB/s sequential writes and 25 MB/s 4K random reads.

 

[bargetrav]

Have you tested it on the ICH10R or an ICH7R, 8R, or 9R for that matter? The only Marvell testing I have seen so far, the Intel controller smokes it.

No, can you literally just unplug from my Marvell and plug into the ICH10R? I don't want to have to re-install everything just to do that, also the few others I've seen with the same results were on Intel controllers I believe.

Obviously the Intel controller would limit my drive in read speeds too, as it's a sata 2 controller

 

[Aristotelian]

I was looking on the forums for an explanation, and apparently OCZ's new drives (on 25nm) are slower than the older drives as well. One forum user argues:

"Using higher density, but less NAND chips will most certainly cause performance to go down. SSDs essentially scale linearly (disregarding any firmware effect) with the number of NAND chips (that's why of course larger SSDs are faster). With 25nm, manufacturers can now halve the number of NAND chips, reducing costs significantly. What OCZ has done is basically taken the performance of a 60GB model and sold it as a 120GB model. That is what's unfair to customers. And the reduced lifecycle thing (25nm is good for about 3000 cycles, compared to 5000 for 34nm)."

His post here: http://forums.anandtech.com/showpost.php?p=31223883&postcount=13

And the source for the relevant thread here: Source: http://www.guru3d.com/news/ocz-vertex-2-with-25nm-nand-flash-reported-slower/

Essentially: reduced costs on the 25nm process as you need less memory chips to meet the size as you did on earlier processes. Less chips seems to entail slower speeds. And frankly the reduced costs were certainly not being passed on.

 

[bargetrav]

That is very interesting to note, but why only slower on random 4k read/writes.... we need some professional reviews of the C400, Series 3 drives, ASAP.