Comments on: You asked for it: RealSSD C300 random IOPs

By: ecards

ecards — Mon, 08 Feb 2010 20:14:21 +0000

I’ve been visiting every few days to catch the release of the C300 and now I see the release is being described as in a few weeks (as of 2/4/10) instead of just “February”.

Technically that’s still February but I was hopeful .

However the important point would be when can we get testing results from AnandTech, which would be a great validation of the numbers shown above.

By: chris

chris — Thu, 10 Dec 2009 13:46:41 +0000

Wow, very nice numbers, thanks for sharing. Would you care to share some 4kb random read and write numbers via crystaldiskmark 2.2.

The one in your benchmarks before are all sequential (ATTO).

By: MRFS

MRFS — Wed, 09 Dec 2009 22:08:22 +0000

@ Yorick

Forgive me for double posting, but
my last message failed to appear here.

On very close examination, we see
an SFF-8087 cable connected to the
back edge of his HBA. That cable
terminates at 4 x SATA connectors.

You can also see some of the other
3 tucked away above the motherboard:
just follow the black bundle from
the HBA, to the one he uses for his
6G test.

The Rocket 620 only has 2 ports,
and that HBA is much smaller.

Hope this helps.

MRFS

By: GullLars

GullLars — Tue, 08 Dec 2009 20:36:04 +0000

Great random numbers. These are the reverse read/write of competitor x’s SLC drives. In IOmeter benchmarks done on a forum i frequent, we got about 40K random read 4KB IOPS and just above 50K random write 4KB IOPS @QD 32 on the 32GB SLC drive just running from ICH10R. There will also be quite a few new competitors launching SSDs with simelar throughput and IOPS numbers in Q1 2010.

One thing missing that would be of great help is the average accesstime on an IOmeter test pattern: 66% read, 100% random, 4KB, @QD 32. Many earlier SSDs have shown max responsetime spikes of above 50-100ms in such a load, including “competitor Z” in your graphs. I don’t know if such events may have been caused by buggy chipset drivers or firmware, but a test of C300′s max responsetimes under such a load would bring peace to mind for some of us storage geeks.

By: Yorick

Yorick — Tue, 08 Dec 2009 16:44:02 +0000

@MRFS: Chances are it’s a Rocket620/622, which uses the Marvell chip and is limited to 500MB/s by the PCI-E 1x design. I doubt you’ll see that impact random read/write much, if at all. On sequential read, the Dec 2 benchmark shows 371MB/s top read speed. Taking SATA overhead into account, that could be a limit imposed by the PCI-E 1x bus, in which case you may look at a 400MB/s to 450MB/s sequential read speed achievable at true SATA 3.0 speed.

Time will tell.

By: Yorick

Yorick — Tue, 08 Dec 2009 16:27:20 +0000

Those IOPS are crazy talk. That’s higher than the Enterprise-class SLC SSD just announced by one of your competitors!

I’m hoping those numbers hold up in independent tests after retail release. My hat’s off to you – with the performance you’ve shown so far, this is one engineering marvel.

By: MRFS

MRFS — Mon, 07 Dec 2009 20:10:51 +0000

Hello Todd,

You mentioned the Marvell HBA in your latest video (“HBA” = host bus adapter).

Can you be more specific about that model etc.?

Some of these “6G” HBAs suffer from a bandwidth
ceiling of only 500 MB/second (e.g. PCI-E x1 Gen2).

I for one would like to know if you were
using an on-board controller, e.g. ASUS
or Gigabyte P55, versus an add-on controller.

We’d very much like to know if a RAID
controller with full 6G support, like
Intel’s RS2BL040 and RS2BL080,
will produce a noticeable difference.

Thanks!

By: Mike

Mike — Mon, 07 Dec 2009 19:10:30 +0000

Wow! Those numbers look great. I can’t wait to get my hands on a C300. Is there any chance that you could run the AS SSD Benchmark on the C300 and post the results? The benchmark is available from http://forums.legitreviews.com/about21729.html.

Thanks,
Mike