Vertex LE vs Crucial C300

[SteveRo]

only 102? - odd.
That was the main reason I bot the UD7 was so that I could go to high pcie oc.
It was reported that you needed to oc the pcie on GGBt boards to get ultra high bclocks (210+)
I didn't do anything special to enable high pcie - that said i have only used the 5870 - no other vc.
Do you have a different vc? Maybe give that a try?
I have had no problem all the way to pcie 119 - both with ich10 and with the 1231.
Seems llike the 9211 was more picky - had to drop to 110 or 115 I think.
I would first suspect the vc - is your vc oc'd?

edit - I did have an ioemeter lockup problem with the 4x x25v on the 1231 - but that problem was even with pcie at 100.

[Anvil]

No OC on the VC , unless it's OC'd by default.
The VC is a Vapor-X by Sapphire.

I've got an nVidia 250? somewhere, I'll give it a try.

Did you leave the PCIE Voltages at auto?

[SteveRo]

In the bios screen where you increase the pcie speed - make sure the pcie v is set to 900mv
The big pcie voltage - I increased one click but not sure that was needed

[Anvil]

Thanks Steve

You're referring to this setting Link?

It might also be the BIOS, I think I'm using F6, I have downloaded F7h today and will give it a try as well if changing the VC doesn't help.

[SteveRo]

Thanks Steve

You're referring to this setting Link?

It might also be the BIOS, I think I'm using F6, I have downloaded F7h today and will give it a try as well if changing the VC doesn't help.

Yes - the link is correct - pcie clock drive @ 900mv is what i use.
Actually i am all the way back on F4 - been no reason to change (yet).

[jalyst]

I'm so jealous of your guys buying 4 or more of these top-end SSD's, i can't afford that right now.
I'm wasting my time even following this thread any more I think

[alfaunits]

The funniest thing is bulk of the people posting in this thread don't need anything close to that - well, no single person does really
But it's FUN!

[Anvil]

Yes - the link is correct - pcie clock drive @ 900mv is what i use.
Actually i am all the way back on F4 - been no reason to change (yet).

OK,
I checked and it was set to 700mV, I'll just swap my boot drive and do a few tests.

Meanwhile, here's the scores for the 4R0 C300+LE thingy
Stripe size 16KB

wbc Off - wbc On
as-ssd-bench Volume_0000 01.05.2010 21-45-46_wbcOff.png as-ssd-bench Volume_0000 01.05.2010 20-20-11.png

CDM3_wbcOn.PNG

iometer 4KB random read
rr_4kb_wbc.PNG

1st and only run, the drives aren't clean
vantage_1st_run.jpg

[GullLars]

Most people wouldn't notice anything above 2R0 x25-M/LE/C300 on ICH10R, if you have a powerfull setup (OC'd Quad or better) and do heavy stuff, you may notice a slight benefit of 3R0 from ICH10R, or using a controller with cache.
RAIDs generating 1GB/s+ is overkill for most people, but can be nice for people running production or server testing setups, like Anvil, or going for benchmarking VRs, like steve, mike, Nizzen and mbreslin.

EDIT: for clarification Anvil, did you use WBC on that PMV HDD score?

[Anvil]

Yes
WBC On, should have been higher but...

Here's the iometer result for that config. (using wbc as well)
512B_256KB_RR_C300LE_4R0_4KBALIGNED_WBCON.zip

I'll try again when I get the new firmware for the C300, supposedly tomorrow.
(according to tweaktown)

edit:
Did you read the new article at Anand about over-provisioning the SF drives?

[GullLars]

Hadn't seen it, checking it now

EDIT: it was interresting, but roughly what i expected.

[SteveRo]

Anvil - very nice indeed!
It appears that the LE's and C300's play nice in R0!
Did you notice anything unusual otherwise? - the benches certainly look great.
Am I seeing that right? -pcmv HDD test over 100k? and this on ich10 right?

[Anvil]

Steve,

I didn't notice anything out of the ordinary, it's just fast

CDM could have been higher if I had used the 0x00 setting (compressible data) but I figured I'd try again later using the stripe size of choice from my other tests.
Yeah, the PCMV is alright, given that the Intels can do more I'm not that impressed but things might change at other stripe sizes.

I'm about to try the PCIE adjustments, hoping it's stable.

[SteveRo]

Back to the pcmv HDD results -
Notice the windows media center score - 3192MB/s - and you would have thought that test used - already compressed - a/v data - go figure.

[Anvil]

The media center score is due to wbc.

Currently at PCIE 115 working my way up, looking good so far.
I'll check how far I can go just for the reference.

About to run PCMV HDD test on the single Kingston boot drive.

[SteveRo]

About "media center score is due to wbc" the 1231 has 4GB cache and still only gets ~1k in that test (vs your 3+k) - I think it must be more than cache.

[Anvil]

Steve,

What else could it be?
Just try it for yourself, using the 3-4 X25-Vs you should be able to reach 80' or more using wbc, probably ~60 or more without the wbc.

A single X25-V is 28-30K at PCMV HDD. (without any wbc)

[SteveRo]

I have already broken up my frankenstien - a leg here - an arm there - all in different machines.

I did go back and look here - http://www.xtremesystems.org/forums/...&postcount=114
I was able to get 2K in that test with ich10/acards - I guess I just never noticed how well ich10 did in that subtest!

[SteveRo]

Anvil - I can't wait to see what you post for pcie 115+
oc the proc also if you can - that obviously makes for a big improvement as well.

[GullLars]

And anvil, plz run an IOmeter test with #workers = threads for 4KB random read QD 1-512, just for max IOPS reference. If you can both ICH10R WBC off and LSI 9260 WT/WB.
It could also be interresting to se the same results for CPU @ stock, and CPU @ OC. And if you also have the posibility, CPU OC only FSB and CPU OC only multiplier.
I suspect both PCIe clock and FSB will help, and maybe multiplier too.
And how about NB / QPI OC?

Maybe we could make a thread dedicated to comparing IOPS and bandwidth gains from OC?

[Anvil]

I'm still testing for stability at PCIE 119, still looking good.

I've done a few tests using a single X25-V (Kingston) and it's hard to tell if pcie alone improves performance.
I'll test some more before I crank up the processor speed, it's a late D0, it did ~4.4 on air in early tests without any issues.

My 4R0 X25-V/Kingston array is also dismembered. (installed in 2 different computers)
I've kept one as a "bench-boot" drive for the UD7.

[Anvil]

And anvil, plz run an IOmeter test with #workers = threads for 4KB random read QD 1-512, just for max IOPS reference. If you can both ICH10R WBC off and LSI 9260 WT/WB.
It could also be interresting to se the same results for CPU @ stock, and CPU @ OC. And if you also have the posibility, CPU OC only FSB and CPU OC only multiplier.
I suspect both PCIe clock and FSB will help, and maybe multiplier too.
And how about NB / QPI OC?

Maybe we could make a thread dedicated to comparing IOPS and bandwidth gains from OC?

Great idea GullLars
As soon as I establish what makes the improvements on my system I'm ready for the other tests.

[Anvil]

OC 4.2Ghz pcie 119

I'm missing a baseline but here's my first benchmarks

2R0 Vertex LE

CDM_OC42_PCIE119.JPG CDM_OC42_PCIE119_wbcOn.JPG

It does look like it's higher than my earlier tests.

AS SSD is also the best I've ever got using the LEs
as-ssd-bench Volume_0000 04.05.2010 00-16-48_pcie119_42_wbcOn.png

I'll do a few more tests and then I'll add the C300s to the array.

[SteveRo]

very nice, you should also do oc of the 2xC300's befor you combine into Frankenstien

[Anvil]

OC 4.3 pcie 119

CDM_OC43_PCIE119_wbcOn_ss64KB.JPG as-ssd-bench Volume_0000 04.05.2010 01-06-33_pci119_43.png

It's still a bit early to conclude, returning to stock should confirm the effect.