Gigabyte GA-P35-DS4 motherboard with DPC spikes - RMA?

[Timo]

Regards: Gigabyte GA-P35-DS4 deferred procedure call spikes.

System specs:
* Gigabyte GA-P35-DS4 revision 2.1, BIOS version F12
* Intel Q9450 2.66GHz quad w/ Zalman CNPS9700 HS & fan
* 2x1GB Corsair DDR2 PC6400 CAS4 RAM
* 512MB MSI NX8800GT graphics
* 2x500GB WD Caviar RE2 SATA-II HDs
* Mitusmi FDD & 7-in-1 card reader
* Samsung SH-S203P DVD+/-RW
* Corsair 520W HX PSU

I bought the above mobo, RAM, hard-drives, CPU HS and fan, DVD-rewriter, and FDD from Scan last week. (The graphics card, PSU, and CPU were purchased elsewhere).

I'm a musician/engineer attempting to use the above computer configuration for high performance low-latency music studio recording, but I'm suffering from unduly large and regular Deferred Procedure Call spikes on the machine.

I am using the free DPC monitor utility from Thesycon to monitor the system, which anyone can download here: DPC Latency Checker

Below is an average screengrab of the system merely idling, which clearly displays the spikes. The DPC latencies oscillate wildly between 5 microseconds up to 1500 microseconds.



In comparison here's an associate's P5K motherboard made by ASUS, which shows what it should be like:



I have attempted to narrow the suspect down by disabling all non-critical devices on the motherboard such as ethernet/LAN, CD/DVD and Floppy Disk components from within Device Manager, and every non-critical Service, as well as even disabling System.ini and Win.ini at startup, to no avail. I've even gone as far as disabling all USB and IEEE1394 (firewire) devices, and using a PS/2 mouse and keyboard instead, again to no avail.

It occurs even right after a completely clean installation of Windows, before even installing chipset drivers, graphics, etc.

Looking at other public forums, such as Tom's Hardware and the Cakewalk Sonar music sequencer forums it appears I'm not alone with the Gigabyte motherboard issue:

Audio latency problem with P35-DS3R

and...

Gigabyte P35 motherboards and DPC

... shows many others with the identical problem (even with the same off-beat, 'staircase' type pattern in the DPC monitor window) with BIOS versions F6 and thereafter.

Apparently BIOS versions before F6 worked well, without any DPC spikes. However, my Intel Q9450 Yorkfield CPU requires BIOS version F9 or above, so downgrading to BIOS version F5 or below is NOT an option for me.

I received the motherboard last thursday, so the 7-working-day RMA period is almost up (this Friday). I have contacted Gigabyte but haven't received anything back from them yet.

Not sure what to do. I hope Gigabyte acknowledge the problem and work to fix the BIOS issue that's creating the DPC spikes, but at the same time I have only 2 days left of the 7-day RMA period.

The main reason I picked the Gigabyte board is because, unlike the ASUS boards, it has native onboard Texas Instruments firewire chipset (for maximum compatibility with my Motu Traveler 20-in 22-out firewire soundcard), great clocking features, excellent connectivity (12 USB, 3 firewire), and good expansion options (PCI-E slots, Sata channels, etc.).

Thanks,

Timo

My original Sound-On-Sound forum post

[Timo]

Hi, how are RMAs handled, please?

My 7-working-day RMA period ends tomorrow (Friday), but I have not received any acknowledgment back from Scan Technical Support or Scan RMA department regards the motherboard problem. I sent the query to Technical Support on Tuesday, and the RMA department yesterday. I'm still awaiting support from Gigabyte regards the matter, too. Otherwise I will have to exchange it for another board (like Asus PK5 or Abit IP35 Pro).

[rabbid]

Am I correct in thinking you want to return this under DSR as the product seems not fit for the purpose you require?
If in doubt raise an RMA for return under DSR within the 7 days to protect yourself (has to be in writing), you then will have some reasonable time to return it or if you/they find a solution you can cancel the RMA but it keeps your options open as its obviously a major factor in your requirements to have a board that wont do this.

[rabbid]

FYI I ran your tool on my GA965P-DS3r3.3 (onboard audio) and from watching the current latency numbers its typically 8-26us on system idle.

However I tried to provoke it a bit and so opened some webpages and noticed spiking as you are experiencing.

Edit: Ok I read you already tried disabling LAN, nvm

[Timo]

Ideally I would have liked to have heard what Gigabyte say about the problem, as it appears to be a common problem affecting many users after reading the aforementioned other internet sources, but as the DSR regulations end tomorrow (Friday 18th) and I haven't heard from Gigabyte regards the problem I can't wait any longer to find out, so I've had request the RMA under the DSR regulations.

As mentioned, I contacted Scan Technical Support on Tuesday and Scan's RMA department yesterday via online forms on Scan's site, but haven't heard back from either.

Given that I've had no technical support from Gigabyte or Scan, and that I don't have time to wait under the DSR regulations, I've filed for an RMA just a few minutes ago. I just wish there was some automatic acknowledgment that they've received it.

PS > Yes, as mentioned, the motherboard works fine for BIOS versions F5 or lower.
BIOS versions F6 and above causes the spikes. And, because I have one of the new Intel 1333MHz Yorkfield CPUs, the earliest BIOS version that supports this chip is F9, so reverting back to F5 is not an option for me.

PSS > The spiking occurs at idle. You don't even need to provoke the system to get the spikes. They are there right after a completely clean Windows install.

[rabbid]

Is this evident just with the Q9 series? Or people with Q6600 using older BIOS than F6 wont notice/report it?

What version of xp pro are you using, and is it updated?

[Timo]

Windows XP Home (32-bit) with SP2, fully updated.

If BIOS versions F6 or above are installed, everyone with Q6xxx or Q9xxx CPUs suffers the problem regardlessly.

Users with Q6xxx CPUs with motherboard BIOS version F5 or less seem to be OK.

It's definately a BIOS issue. I just don't have time to wait for Gigabyte to fix it as I don't want to be left with a dead duck if they don't. Gigabyte were made aware of the problem a month ago from other users.

[rabbid]

I dont remember if it gives an OLQ reference number when you file, I would assume it does so as long as you have that its fine. An RMA number will no doubt be issued in the next opening hours after someone has reviewed it, as far as I remember its not automatic.

[rabbid]

So until you guys can lobby Gigabyte to update/patch the BIOS, if you want this board you have to use older CPU with older BIOS?

Do you know of boards that are unaffected, or is it more of a GA range issue?

[Timo]

Yes, I have the OLQ reference, which I've kept safe. Oddly the reference number for both my Technical Support and both RMA queries is the same number.

Regards Gigabyte patching/fixing the BIOS, it doesn't look good.

From another fellow musician user that raised the issue with Gigabyte:

"Quick update regarding the P35-DS4 motherboard and DPC spikes relating to BIOS versions >F5. I had a reply from Gigabyte who say that I should stick with the F5 BIOS and that they aren't planning on fixing DPC spikes in later BIOS releases because they are now just adhering to Intel's specifications!

If this were really so then surely many other boards would suffer the same issues? Also unless this is a specific hardware design issue (quite possible) then what does this say about the suitability of Gigabyte P35 motherboards for DAW use? Here are the responses from Gigabyte:

"We discovered this status is caused by south bridge settings and those settings are followed by Intel ICH9R specification to modify. It means that BIOS before F5 version had not been added Intel’s spec and we just followed Intel. Stay with F5 BIOS is the best solution."

Q: Are you likely to "fix" the issue in a later BIOS release?

"Sorry but we will not to modify a special version BIOS for this issue since we had followed Intel’s specification already."

(DAW = Digital Audio Workstation).

However, there are many more of us with the problem now, so Gigabyte may be forced to issue a fix.

But, again, I don't have that time to gamble on whether they will or whether they wont.

A great shame, as it's a great motherboard with excellent connectivity (12 x USB, 3 x native Texas Instruments firewire, 8 x SATA channels, 3 x eSATA on PCI, 3 x PCI-Express x1 slots, S/PDIF in and out) and has great enthusiast clocking features.

It initially appears that it is GA ranged, with P35-DS4, P35C-DS3x, P35-DS3R, P35-DS3P, P35-DS3L, P35-S3G, P35-DQ6, G33-xxx, X38-DQ6 and X48-DQ6 motherboard users being affected. All show the same patterned DPC latency spikes. The problem is not with the Intel P35 chipsets, as other manufacturer boards using that chipset work great. It's local only to Gigabyte.

[rabbid]

A great shame indeed, I hope they will listen to you eventually. Good luck with the replacement.

[Spook Tooth]

Have you tried the PCI latency tool?

[Spook Tooth]

PCI latency tool might be able to help... ?

(Separate post since you need 5 posts to err post link.)

[Timo]

Update!: Gigabyte personally issued me a fix for my P35-DS4 board: BIOS version F13b. I understand the "b" is beta, and not an official version. Just to recap, I have the GA-P35-DS4 revision 2.1 with an Intel Q9450 2.66GHz Yorkfield CPU.

Their words:

"Attached special bios is modified to support Q9450 and also keeping bios codes from F5 bios to avoid DPC issue."

It appears to be similar to what some other people have also been given by them.

Benchmarks. This is my previous XP 32-bit BIOS version F12:-



New XP 32-bit BIOS version F13b:



Anything from 4 microseconds to 22 at idle. I get an occasion lone spike at 129 microseconds every 15 seconds or so. But the average toggles between 5 and 12 microseconds.

I used a hard-wired PS/2 mouse for the above screengrab. My Microsoft wireless optical mouse was taking the latencies anything up to 42 microseconds whenever I moved it, but switching to a hardwired PS/2 mouse cut these latencies by half.

Now on to Vista-64...

This is what I was getting using BIOS F12 in Vista64 after a clean install (with chipset and graphics drivers only. I haven't updated Vista-64 via Microsoft Update yet, because Vista wont "see" my ruddy USB Modem even though the modem has genuine 64-bit drivers! Gah!):



... basically a mirror image of what was seen in XP-32.

Opening a few windows sent it wappy, though:



Now Vista64 with new BIOS version F13b:-

When it's all settled down (there's absolutely loads of spikes when the OS has first loaded up), and without optimising anything other than disabling ethernet, I get average of 59 microseconds. It oscillates tightly between 57 and 60 microseconds, apart from random spikes (anything up to 5000 microseconds):



Disabling Aero barely affected the values. However I went the whole hog and disabled ALL fancy frills, setting processor scheduling to Background services, etc., but then...

.. Disabling my Samsung SH-S203P DVD-RW drive (via Device Manager) took the average down to 29 microseconds:



Still get spikes when changing windows and stuff, though:



... these can be anything up to 5000 microseconds, as mentioned, and can last a second or two.

But still pretty surprised Vista got down that far, on the whole. I haven't even updated it yet, or installed SP1, etc.

All my tests/screegrabs were when the systems (XP-32 and Vista-64) were IDLE. I don't have any projects to test it as yet, as up until now I've generally been more hardware based, with a little bit of software. I can now work to integrating all my hardware with my computer, and get my hands more dirty with the software side of things.

Thanks for all the help, anyhow. Have thankfully cancelled my RMA request.

[Spook Tooth]

Good news. Faulty motherboards and amongst the worst hardware problems to have, especially the hassle of disassembling everything to RMA.

Glad it's working for you. Incidently, what kind of music do you produce and which software do you favour?

[rabbid]

Thats really good news!
Approx how long overall was it from contacting GA to getting a beta bios?