Gigabyte P67 SATA performance bug?

Discussion about Intel CPU Motherboards
Post Reply
User avatar
egloeckle
Mr Awesome
Mr Awesome
Posts: 254
Joined: Wed Sep 28, 2011 11:48 pm

Gigabyte P67 SATA performance bug?

Post by egloeckle »

Found an odd one that is fully repeatable, wondering if anyone else has a similar board that can test this as well.

Board: Gigabyte GA-P67A-UD4-B3 (try with any of the GA-P67A-UD* boards, I am curious if this is just this version or mine specifically)
SSD: Samsung 830 512GB (you need a newer SSD to test this as it looks like the bottleneck is in the ~260MB/s write range)
Ran the tests with both the F6 BIOS revision then upgraded to F8 (latest) revision, exact same results.

With BIOS set to defaults (c3/c6 enabled) run CDM/AS SSD. Take note of the write speed.
In BIOS disable c3/c6 under MIT-> (second option down, dont remember what its called) -> Advanced CPU settings -> C3/C6 (near the bottom) and run benchmarks again.

My results (c3/c6 enabled): ~260MB/s seq write
C3/C6 disabled: ~400MB/s seq write
Other tests showed similar speed increases.

Caught this by running benchmarks on a fresh install with only basic drivers installed (didnt install video driver yet, ran after video driver also, same results) and thought 260MB/s was way too slow for this drive. Checked drivers, cabling and other obvious things with no luck. Found a random post somewhere else on a different model board saying to try disabling c3/c6 so figured what the hell, wouldnt hurt and was surprised. Checked with Dan and it surprised him as he hasnt noticed it on any of the boards he has reviewed (he reviewed this board long ago, but it was a much older/slower ssd so the issue probably would not have come up).

Anyone have any thoughts (besides leave c3/c6 disabled)? Anyone else have a close enough board (or any other board really) that can reproduce this?
User avatar
XstollieX
Staff Writer
Staff Writer
Posts: 1641
Joined: Mon Mar 24, 2008 10:51 am
Contact:

Re: Gigabyte P67 SATA performance bug?

Post by XstollieX »

May try and swing by one night with the Corsair Force GT we now use for testing and see if we can replicate with that drive. That may tell us if it is the Samsung controller or the P67 chipset
User avatar
Major_A
Legit Extremist
Legit Extremist
Posts: 3793
Joined: Tue May 15, 2007 2:11 pm
Location: Houston, TX

Re: Gigabyte P67 SATA performance bug?

Post by Major_A »

What AHCI drivers are you running? I think Intel pulled the latest RST drivers from the website and reposted the ones what are close to a year old. Seems like MS AHCI driver is the way to go if you don't need any of Intel's software.
User avatar
egloeckle
Mr Awesome
Mr Awesome
Posts: 254
Joined: Wed Sep 28, 2011 11:48 pm

Re: Gigabyte P67 SATA performance bug?

Post by egloeckle »

I was running the Intel driver. Switching to the MS ACHI driver gave me the following results

C3/C6 States disabled: 512MB/s read / 400 MB/s write
C3/C6 States enabled: 505MB/s read / 300 MB/s write

So changing the driver from the Intel one to the MS one helps (~40MB/s faster seq write w/ c3/c6 states enabled).

Also, on the gigabyte bios I am enabling/disabling C3/C6 state under the following (I did not have the exacts in the OP):

M.I.T. -> Advanced Freq Settings -> Advanced CPU Core Features -> C3/C6 State Support (near bottom)

Setting to Auto or enabled gives the same slow result. Disable this and I get the major increase.

Doing some google-fu I found some others reporting on all sorts of benchmark increases by disabling c3/c6 states, including the potential for a 3dmark11 increase (http://www.evga.com/forums/tm.aspx?m=92 ... print=true). I havent had a chance to test this on my board yet (dont have 3dmark11 installed yet) but maybe Dan wants to give it a shot on whatever he has on his testbench right now. To my knowledge, disabling c3/c6 states disables Intel Turbo Boost which to me isnt that big of a deal (I can manually overclock if I cared enough), but I still find it odd that this setting can potentially cause such a large performance dropoff. Though power savings and performance are fairly opposite goals, I still find it odd how badly it is affecting a seemingly unrelated component.
User avatar
Major_A
Legit Extremist
Legit Extremist
Posts: 3793
Joined: Tue May 15, 2007 2:11 pm
Location: Houston, TX

Re: Gigabyte P67 SATA performance bug?

Post by Major_A »

Definitely sounds like a bug. And from what I can tell it sounds like a P67 bug. Have you tried reporting your findings to Gigabyte or Intel?
http://www.intel.com/support/feedback.htm
http://www.gigabyte.us/support-download ... pport.aspx
User avatar
egloeckle
Mr Awesome
Mr Awesome
Posts: 254
Joined: Wed Sep 28, 2011 11:48 pm

Re: Gigabyte P67 SATA performance bug?

Post by egloeckle »

Sort of why I am asking. Trying to see if anyone else has the same issue with other boards (P67 or otherwise) and even if it is a P67 issue only if it is specific to Gigabyte or not.

For all I know it could just be an issue with my board specifically and not something common to P67 or even Gigabyte P67 boards.

Edit: It could also be some combination of the P67, Gigabyte, and Samsung 830 drive also. Until I finish getting everything off my 'old' Kingston HyperX 120GB SSD I dont have another SSD free to test with (unless Dan stops by with a few at some point). The only other SSD's I have are tied up on machines that I cant currently bring down to test.
User avatar
Major_A
Legit Extremist
Legit Extremist
Posts: 3793
Joined: Tue May 15, 2007 2:11 pm
Location: Houston, TX

Re: Gigabyte P67 SATA performance bug?

Post by Major_A »

Have you searched Intel's forums?
http://communities.intel.com/community/tech/desktop

I try to keep the power saving settings turned on and the only new board I have is a Z68. What app are you using to record drive speeds? Maybe I can see if my board has the same issue.
User avatar
egloeckle
Mr Awesome
Mr Awesome
Posts: 254
Joined: Wed Sep 28, 2011 11:48 pm

Re: Gigabyte P67 SATA performance bug?

Post by egloeckle »

CDM and AS SSD both with similar results.

I only did a bit of searching but did find some reports of this that other people are having. Havent found enough info to tell where the fault is exactly, figured a few people here would have P67 boards and a newish SSD and wouldnt mind doing a quick benchmark or two.
Post Reply