I'm using the Acard in Raid0 configuration using the P55 chipset. Should I do a run without Raid0?

Just installed the RE5 benchmark, gonna do a run on DX10 with the current 9.7 and new 9.12's

WTF!

Just ran the DX10 benchmark:

Turbo disabled: 73,9
Turbo enabled: 64,7

Single Acard (no raid)

TURBO DISABLED



TURBO ENABLED

With the new Catalyst 9.12 it seems fixed/gone:

Turbo disabled: 100
Turbo enabled: 102

Vantage High preset: issue still present:

Turbo disabled: 6991
Turbo enabled: 6871

interesting RE5 numbers; so driver issue?
do you have other HDD around beside the ACARD? don't even need to install anything, plug it in , quick format

The HDD I've installed Windows on?

Not sure about the graphics part: Vantage gives me incorrect figures again.

TESTED WITH A NORMAL PLATTERED HDD

TURBO DISABLED



TURBO ENABLED



=> Still a performance drop when enabling turbo mode, however not as spectacular as with the Acard. I guess I'm not that far off if I blame the HDD bottleneck :-p

so for max storage performance with SSDs in RAID (as in: pushing SATA 300 specs) disable turbo mode

Even without raid (post #43)

yes, correct, reason I mention raid is because it's easier to hit SATA 300 limits with it;)
got a technical contact at Intel you can relay this info to?

Sure. I prefer to release the article first ... then I can write "they didn't listen" :D

I'd prefer the other way around ;)

Intel should now be aware

thanks!

Asus/MSI/Gigabyte contacted.

I can lend you the vertex drives MM

lend them why?

just to prove you wrong :p

prove me wrong how? seriously not following here?

I wrote this in my first Core i5 article:

:)

if you disable C-state, turbo mode doesn't work; and SSD performs at max
if you enable C-state, you get turbo mode, but SSD speed suffers

+/- correct?

I've read something about power state and SSD performance in the past
edit here: http://www.tomshardware.com/reviews/...ower,2170.html



also on AMD

Re-did testing with C1E and C-states disabled; EIST enabled to get turbo mode enabled.

TURBO DISABLED:



TURBO ENABLED



Major issues are solved by disabling power saving mode. The 4k still isn't performing as it supposed to do ... but oh well.

If you don't change any bios setting, I assume that C states and EIST are enable by default?

For turbo mode, you only need to enable EIST.

ok; C-state is enabled by default I think yes
you could try to do a BIOS page reset without saving the changes to see what value it is set at; but I think C-state is enabled by default, it cuts down idle power usage a lot

Hm. Not sure how I should classify these results then. On the one hand, the performance problem is easily fixed by disabling major power saving features, but on the other hand, if you disable these power saving features, you can as well, the force the CPU multiplier to the highest value.

Retesting last HDD config (c-states disabled turbo mode) as we speak ... results are off.

a mixxed bag indeed; has more to do with underlying SATA interface, than with the motherboard; hopefully SATA 6G will solve this

Findings are in line with THG - good enough, I won't rant in the article :)

TURBO DISABLED



TURBO ENABLED



Only 4k is a problem with speedtech enabled.

so now only C-state disabled in BIOS?
does in impact power usage at idle, do you have a power meter,

Only have a broken one :(.

Coming back to the HDD issue, this is what THG said:

Assuming that this hypothesis is correct, we should see this kind of behavior when manually decreasing the CPU multiplier. Assuming that the PCMark HDD test results are accurate, we can see that:

- 3.80GHz = 84000
- 2.67GHz = 80000

The result when enabling the power-saving settings, the result is around 52000. In theory, these power-saving settings decrease the CPU multiplier to the lowest possible value, in this case "12x133" = 1600MHz.

So, there are two possibilities:

1) The hypothesis is correct and the HDD transfer rate is affected mostly because of the reduces amount of CPU cycles.
2) The hypothesis is incorrect and the HDD transfer rate is affected by another cause, possibly in interaction with the reduced CPU cycles.

There's an easy test to test the hypothesis: if the reduced CPU cycles are indeed the cause of the sudden drop in HDD performance, we should see the same behavior when manually decreasing the CPU multiplier while disabling all power saving features. If the hypothesis is correct, we should find a logarithmic relation between HDD tranfer rate and CPU cycle: the lower the clock frequency, the faster decrease in HDD performance. If we find a different relation, eg linear, we'll see that the performance result when enabling power-saving technology differs from the performance result obtained when manually decreasing the CPU multiplier.

Sadly enough, I don't have the tools to measure the performance right now as the Acard is at a colleagues place.

the C-states disable features on the CPU to reduce power usage at idle, it's not completely identical to manually lower the multiplier, that is just speedstep in action

Hm.

I assume that we can trick the CPU not to go to the C-states by putting a bit of load on the cpu's using an application that load all four cores.

Intel TAT can put a 10% workload on all cores easily :)

Ok.

Is there a tool that reports in what power saving state a CPU is in?

is this any good? http://intel-r-processor-id-utility..../3.9/download/

More info on power managment:
- http://www.intel.com/technology/itj/...management.htm
- http://www.hardwaresecrets.com/article/611 (C States)