LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited December 2007
Are you sure your DRAM has enough power? Is the memory overclocked? Your SMP Folding problem is a symptom of one of the following:
1) bad memory
2) memory under volted
3) OC too high
4) not enough power for the OC
5) fluctuating voltage - enough to damage F@H ('dirty' power from the wall outlet or poor PSU voltage output quality)
Here are the BIOS settings that I have for my P35-DS4 Rev. 2, Q6600 B3 @3.3GHz (keep in mind that B3s require more vCore at default and especially at OC:
Of course, as you know, voltage on the high side is better than too little. Don't worry, the DS4 is very well made board with excellent passive cooling. You won't hurt it by jacking up the voltages. I assume you have good CPU cooling? What are you maximum core temperatures - the highest core out of the four?
I am disinclined to think the problem is your motherboard. I'm thinking undervolting or OC above your CPU's capabilities.
The thing I'm concerned the most about is the power. No matter what I jack the Vcore to, it never seems to get above 1.44V at load (and I'm talking about setting the BIOS to 1.5125 or such); even going a notch or two above the listed 1.5v in BIOS doesn't seem to get me any extra juice in Windows. Hence why I'm thinking perhaps the PSU is getting stretched too thin.
The memory's correctly handled; it's overvolted to around 2.15v and it's only at 900MHz. CPU cooling is fantastic; with Prime95, I never hit 64, with F@H SMP I only hit 60 when my room gets hotter than normal. Regular usage is somewhere around 58 for the hot cores and 53-54 for the cool ones.
And as I said before, the rig was solid at 3.6 before I tried to push it to 4GHz for fun. All the voltages stayed reasonable in that attempt, imo, but maybe it just got something too stressed. Thanks for your insight, Leo. Any other ideas?
For the record, settings that failed when they had previously been good:
1.49375v core voltage
8x450 for 3600 MHz
100MHz for the PCI-E
+.4v on the memory to ~2.15v at 900MHz, 4-4-4-15 timings
+.3v on the FSB
+.05v on the NB
And all the power savers turned off, including the one for temp that lowers voltage automatically, so it doesn't seem to be that that's giving me the problem.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited December 2007
What PSU is in the computer? That's a strange problem you are describing.
It's a Thermaltake 750W (Toughpower, I think). Modular. So yeah. From my research, I'd heard the TTs were just fine, and 750 is surely enough to power my setup, isn't it?
Q66 on the P35-DS4
8800 GTS, single
2 optical drives
6 HDDs
Misc fans (6? 5 120mm and 1 200mm)
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited December 2007
The Toughpower PSUs are considered above average. 750 is certainly enough power. I'm running a Q6600 (G0) at 3.4GHz on a 430W Antec.
Even returning it to no overclock, I was BSODing before I even got into Windows. 0x7Es, to be precise. I looked around on the interwebs a bit and figured I should check my RAM. Various configurations later, and I'm able to boot at completely stock everything. 9x266, no voltage changes, nothing. One of the other sticks in by itself screws the POST completely, and attempting to up the voltages on the good sticks does the same, in exactly the same manner. It starts by failing to detect one of my HDDs and then fails to detect the RAID. It seems clear that some of my RAM is already bad, and the other isn't responding to prodding well. I'll do some more testing on it, but that's what I've got so far. I'm not sure I tried raising the RAM voltage by itself or if I tried the entire overclock at once, so I'll go back and make sure it's not the NB or FSB voltage that's screwing with HDD detection...
Still not sure if this is looking more like a RAM or a mobo problem at the moment, so I'll keep trying. My Folding output is suffering greatly
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
Right, run Memtest86 and make sure which stick(s) of ram are the problem. I've had bad ram myself and it can give you all sorts of problems that don't seem directly related to bad ram.
Ran through my memory sticks. Had a hell of a time. Only once did a pair of them manage to boot into memtest, and none of the sticks successfully booted in a second comp.
Bought some new Ballistix to try while the RMA's going on. Ballistix did the same thing. First pair booted ok, then none ever booted again. Still bad in the second comp. Now this really hurts, because it makes me think I'm killing the sticks. I must have really hurt the board when I tried to push it to 4GHz, and now it's destroying memory.
Is that even possible? Is there some other ailment that matches my symptoms that anybody can think of? Now I've got 8 gigs of RAM in for RMA and it's really cramping my style. This blows, and I hate thinking/knowing it's my fault. Anybody? Thanks.
Sounds like you agree that replacing/RMAing the mobo is the only course of action? Thanks for your thoughts.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2008
One other thing to try: Check the MOSFET/PWM heatsink. Is there anything obstructing airflow in that area? Is the sink flush with the MOSFET chips? Those MOSFETS can really heat up fast.
But yeah, based on this troubleshooting thread, it would appear your board is not performing up to specification. MJ, I'm running a B3 Q6600 at 3.3 with dual WinSMP, monitoring other computers the VNC, and other multi-tasking. Absolutely rock solid. The DS4 is highly renowned for stability under stress, so I think you got a lemon.
... but to cheer myself up, I just ordered a GA-X38-DQ6.
I have the X38-DS4 here (same PCB/bios according to Gigabyte) You will like it. Needs some bios work yet, but was very easy to get 2GB of ram running over 1200MHz. 4GB on the other hand (4 x 1GB that is, not tested 2 x 2GB yet)...so far nothing stable over 1066.
That's interesting. I wasn't terribly interested in OCing RAM, but I'll let you know what I find out of it. I've got some OCZ Plat Rev2 and Ballistix that I can compare.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2008
I just ordered a GA-X38-DQ6
Oh, oh nice!!
If an RMA doesn't come through with the DS4 (I have no knowledge of Gigabyte after-sale service) you are welcomed to give it to me. I'd love to play with it.
I'll keep you in mind there, bucko. It'd be a good repayment for all the input you've given me.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2008
I have the X38-DS4 here (same PCB/bios according to Gigabyte)
Now why did Gigabyte do that - take superb, high quality performance hardware and mate it with boy-wonder-wannabe overclocker BIOS. The DS4 and X38 are great boards, but the BIOS is goofy city.
It makes me wonder whether something in the ram voltage circuitry fried on your mobo somehow and is overvolting seriously or spiking it and killing the ram. But that's too many bad sticks of ram from different vendors for this to be just bad ram. I would lean toward a mobo problem myself in your shoes as it is the constant in the problem equation.
EDIT: Boy I must be tired. I didn't even notice there was another page to this thread until I posted this. Glad to hear you decided to pick up an X38-DQ6, since I have one waiting for me to get in from work in another week or so for upgrading my main machine. I also have a G0 Q6600 and I'm running some G Skill HZ ram (Micron D9GMH chips, same as the Ballistix). One thing that's nice about the board is that they made it easy to remove that crazy cool heatsink from under the mobo socket if you plan to use some good aftermarket cooling that has a bracket to fit behind the mobo. Gigabyte included 2 little screws just for that task.
^ That's nice to know. Luckily (or unluckily, in some eyes), the Infinity only uses four posts that go through the normal mounting holes. Might induce a bit of board flex, but it doesn't require any extra hardware.
I just want to get back up to 3.6 safely, haha. Board's coming in the 9th, hopefully the RAM gets back close to the same time. I have nothing but high expectations out of this setup. Can't wait!
Given that the only thing that's changing is P35->X38, I'm hoping everything picks up right where it left off.
Heh, I opened the box today and saw the two little screws you mentioned, but of course didn't remember hearing about them... so I just sat there and wondered to myself, "now what in the hell..." But I didn't need 'em, so all's well that ends well!
Board is up and running with the 4 gigs of Ballistix and with everything stock right now. Vista picked up right where it left off, and with care, I even managed to save my RAID0 array perfectly. Drivers installed and everything's going all right, temps are around 30, so I'll give it a few days for the AS5 to cure and I'll be bumping it back up again!
Interesting to note, at first try, the X38-DQ6 put the Q6600 at 8x333 instead of 9x266, which is actually faster than the stock speeds. Possibly the BIOS caused that, but who knows. I've downloaded the F7 BIOS from gigabyte.com.tw, which is interesting; the US Gigabyte site only shows me an F6b, so I'm not sure which I should really go with - the US beta, or the later, but non-US full release? I can't imagine why they wouldn't have put it up for one and not the other. Anybody have any ideas on that?
Just gonna let it run stable for a couple days and shut 'er down at night to give it the heating cycles AS5 wants and see what happens.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2008
I'll give it a few days for the AS5 to cure and I'll be bumping it back up again!
Slightly off topic, so please excuse me: Gentlemen, I've posted a few places at Icrontic mentioning Tuniq TX-2 Thermal paste. NO CURING time and its heat transfer properties are just as good or better than AS5. I'm using it on all my quad rigs now. In all systems, CPU temperatures have been the same with AS5 or better.
I thought about picking up the MX-2, which caused me to completely forget about the TX-2, and the MX-2 at high temps is supposedly less solid than the AS5. I'm not terribly confident in my application this time around, so I may look into it if I have to reseat it anyhow.
0
LeonardoWake up and smell the glaciersEagle River, AlaskaIcrontian
edited January 2008
The main reason why I switched to TX-2 was for the flexibility - no more waiting for curing times and the thermal cycles.
Welp, after a lot of testing, I finally moved from Prime95 to OCCT to try a different test, because nothing was coming up stable; OCCT's CPU-only test passes with flying colors, but the RAM or CPU+RAM tests crash pretty quick. It's interesting, because the RAM's at stock speeds and a nice high voltage, so maybe it's just not at a sweet spot or something. That said, Folding runs perfectly, no EUEs or failures of any kind, so it can't be that unstable.
For the record, it's 9x400, and I've tried 2.00A, 2.00B, and 2.00D strappings to no avail. Might have to check it out at 2.5 and 2.66 and see if it'll handle well there. Upped it to 2.2v as well, so it's not underpowered. Memtest passed 21 times with no failures the other night, so I dunno.
My overall voltages are pretty high right now because I was testing with Prime95 (1.488v idle despite being set at 1.506v in BIOS, haha, +.225 MCH, +.35 FSB, +.05 PCIe), so I need to find some time to back those down again and test with OCCT and see if that makes a difference. That said, still only hitting 55 on the hot cores at load, so I'm fairly happy. Just need to optimize.
This is all on the F7 BIOS, which isn't listed on the US website, so I'll try and see if that RAM OC issue still happens here.
Comments
1) bad memory
2) memory under volted
3) OC too high
4) not enough power for the OC
5) fluctuating voltage - enough to damage F@H ('dirty' power from the wall outlet or poor PSU voltage output quality)
Here are the BIOS settings that I have for my P35-DS4 Rev. 2, Q6600 B3 @3.3GHz (keep in mind that B3s require more vCore at default and especially at OC:
8X413 = 3304MHz
FSB: 1652MHz
PCI Express: 100MHz
Memory: 826MHz
DRAM: + 0.20v
PCI-e: + 0.20v
FSB: + 0.30v
NB: + 0.200v
vCore: 1.41875v (CPUz/SpeedFan = 1.380v)
Of course, as you know, voltage on the high side is better than too little. Don't worry, the DS4 is very well made board with excellent passive cooling. You won't hurt it by jacking up the voltages. I assume you have good CPU cooling? What are you maximum core temperatures - the highest core out of the four?
I am disinclined to think the problem is your motherboard. I'm thinking undervolting or OC above your CPU's capabilities.
The memory's correctly handled; it's overvolted to around 2.15v and it's only at 900MHz. CPU cooling is fantastic; with Prime95, I never hit 64, with F@H SMP I only hit 60 when my room gets hotter than normal. Regular usage is somewhere around 58 for the hot cores and 53-54 for the cool ones.
And as I said before, the rig was solid at 3.6 before I tried to push it to 4GHz for fun. All the voltages stayed reasonable in that attempt, imo, but maybe it just got something too stressed. Thanks for your insight, Leo. Any other ideas?
For the record, settings that failed when they had previously been good:
1.49375v core voltage
8x450 for 3600 MHz
100MHz for the PCI-E
+.4v on the memory to ~2.15v at 900MHz, 4-4-4-15 timings
+.3v on the FSB
+.05v on the NB
And all the power savers turned off, including the one for temp that lowers voltage automatically, so it doesn't seem to be that that's giving me the problem.
Q66 on the P35-DS4
8800 GTS, single
2 optical drives
6 HDDs
Misc fans (6? 5 120mm and 1 200mm)
Even returning it to no overclock, I was BSODing before I even got into Windows. 0x7Es, to be precise. I looked around on the interwebs a bit and figured I should check my RAM. Various configurations later, and I'm able to boot at completely stock everything. 9x266, no voltage changes, nothing. One of the other sticks in by itself screws the POST completely, and attempting to up the voltages on the good sticks does the same, in exactly the same manner. It starts by failing to detect one of my HDDs and then fails to detect the RAID. It seems clear that some of my RAM is already bad, and the other isn't responding to prodding well. I'll do some more testing on it, but that's what I've got so far. I'm not sure I tried raising the RAM voltage by itself or if I tried the entire overclock at once, so I'll go back and make sure it's not the NB or FSB voltage that's screwing with HDD detection...
Still not sure if this is looking more like a RAM or a mobo problem at the moment, so I'll keep trying. My Folding output is suffering greatly
Ran through my memory sticks. Had a hell of a time. Only once did a pair of them manage to boot into memtest, and none of the sticks successfully booted in a second comp.
Bought some new Ballistix to try while the RMA's going on. Ballistix did the same thing. First pair booted ok, then none ever booted again. Still bad in the second comp. Now this really hurts, because it makes me think I'm killing the sticks. I must have really hurt the board when I tried to push it to 4GHz, and now it's destroying memory.
Is that even possible? Is there some other ailment that matches my symptoms that anybody can think of? Now I've got 8 gigs of RAM in for RMA and it's really cramping my style. This blows, and I hate thinking/knowing it's my fault. Anybody? Thanks.
But yeah, based on this troubleshooting thread, it would appear your board is not performing up to specification. MJ, I'm running a B3 Q6600 at 3.3 with dual WinSMP, monitoring other computers the VNC, and other multi-tasking. Absolutely rock solid. The DS4 is highly renowned for stability under stress, so I think you got a lemon.
... but to cheer myself up, I just ordered a GA-X38-DQ6.
If an RMA doesn't come through with the DS4 (I have no knowledge of Gigabyte after-sale service) you are welcomed to give it to me. I'd love to play with it.
EDIT: Boy I must be tired. I didn't even notice there was another page to this thread until I posted this.
I just want to get back up to 3.6 safely, haha.
Given that the only thing that's changing is P35->X38, I'm hoping everything picks up right where it left off.
Board is up and running with the 4 gigs of Ballistix and with everything stock right now. Vista picked up right where it left off, and with care, I even managed to save my RAID0 array perfectly. Drivers installed and everything's going all right, temps are around 30, so I'll give it a few days for the AS5 to cure and I'll be bumping it back up again!
Interesting to note, at first try, the X38-DQ6 put the Q6600 at 8x333 instead of 9x266, which is actually faster than the stock speeds. Possibly the BIOS caused that, but who knows. I've downloaded the F7 BIOS from gigabyte.com.tw, which is interesting; the US Gigabyte site only shows me an F6b, so I'm not sure which I should really go with - the US beta, or the later, but non-US full release? I can't imagine why they wouldn't have put it up for one and not the other. Anybody have any ideas on that?
Just gonna let it run stable for a couple days and shut 'er down at night to give it the heating cycles AS5 wants and see what happens.
For the record, it's 9x400, and I've tried 2.00A, 2.00B, and 2.00D strappings to no avail. Might have to check it out at 2.5 and 2.66 and see if it'll handle well there. Upped it to 2.2v as well, so it's not underpowered. Memtest passed 21 times with no failures the other night, so I dunno.
My overall voltages are pretty high right now because I was testing with Prime95 (1.488v idle despite being set at 1.506v in BIOS, haha, +.225 MCH, +.35 FSB, +.05 PCIe), so I need to find some time to back those down again and test with OCCT and see if that makes a difference. That said, still only hitting 55 on the hot cores at load, so I'm fairly happy. Just need to optimize.
This is all on the F7 BIOS, which isn't listed on the US website, so I'll try and see if that RAM OC issue still happens here.