After reading through these forums I've put together the following system:
case: Antec P182 Case
motherboard: Gigabyte GA-P35C-DS3R
processor: Core 2 Duo E6750
memory: Corsair Dominator TWIN2X2048-8500C5D
harddisk: 500 GB WD Caviar SE16
graphics: XFX GeForce 8800 GTS 320 580M XXX
PSU: Corsair VX450W
Heatsink: Thermalright Ultra-120
On the positive side it is very quiet (compared to my old machine it's inaudible) - so thanks for the tips!
The frustrating part is that it isn't stable. I experience random crashes, memtest86 fails and Prime95 gives rounding errors. Sometimes it runs for hours playing BioShock just fine, the next time I can't even boot.
I have already replaced the ram - no effect.
I have loaded BIOS failsafe defaults - no effect.
I have shuffled the ram modules around, used them in single channel mode, used only one of them - no effect.
I have updated the BIOS to the latest F8 version - no effect.
I have tried to call GigaByte tech support but so far could only listen to their "please hold the line" jingle for two hours straight.
After some more googling I have discovered that the memory voltage settings are the most likely culprit for all of this. The board defaults to around 1.9 volts, while the memory is designed for 2.2V. However I can't get the board to deliver that amount. I change the BIOS setting to +0.1 or +0.2 with the effect that the machine won't even POST until I clear the cmos (via shorting the jumper). Changing the voltage setting to +0.3 (which would be the correct value) doesn't have any effect at all - it doesn't change the measured voltage one iota.
The board is only minimally loaded (CPU, graphics, memory, single SATA drive) and all temperatures are well within tolerable range. I never tried any overclocking and don't intend to.
So - what am I to do? How can I identify/fix the problem? I'm getting desperate here...
best regards,
Sören
I'm sooo frustrated with my new build...
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
NeilBlanchard
- Moderator
- Posts: 7681
- Joined: Mon Dec 09, 2002 7:11 pm
- Location: Maynard, MA, Eaarth
- Contact:
Thanks for the suggestions.
I will try to find replacement RAM to test with, however, that might take a while and I don't know anyone offhand who'd rip out RAM just for my tests.
I think it's unlikely that all my RAM is bad, since this already is the replacement. I sent back the first shipment because I thought the RAM was to blame for my troubles. I think it's unlikely that I should've gotton bad RAM twice in a row.
Using cheapo RAM in the first place would have been a good idea. I didn't know I was buying "OC" RAM - I was actually paying a premium thinking it would spare me exactly this kind of trouble.
Ah well, guess everyone has to learn something on his first system...
Any other tips what I could try? Why won't the motherboard let me set +0.3V? It just completely ignores the setting.
best regards,
Sören
BTW: GigaByte Germany must have the worst tech support ever! I didn't manage to reach a single person on the phone despite trying for hours.
I will try to find replacement RAM to test with, however, that might take a while and I don't know anyone offhand who'd rip out RAM just for my tests.
I think it's unlikely that all my RAM is bad, since this already is the replacement. I sent back the first shipment because I thought the RAM was to blame for my troubles. I think it's unlikely that I should've gotton bad RAM twice in a row.
Using cheapo RAM in the first place would have been a good idea. I didn't know I was buying "OC" RAM - I was actually paying a premium thinking it would spare me exactly this kind of trouble.
Ah well, guess everyone has to learn something on his first system...
Any other tips what I could try? Why won't the motherboard let me set +0.3V? It just completely ignores the setting.
best regards,
Sören
BTW: GigaByte Germany must have the worst tech support ever! I didn't manage to reach a single person on the phone despite trying for hours.
-
protellect
- Posts: 312
- Joined: Tue Jul 24, 2007 3:57 pm
- Location: Minnesota
Looking at google, you're not alone - and actually one guy has got faulty RAM more than once...
http://www.asktheramguy.com/forums/show ... hp?t=64684
There are lot of more problems (and RMAs) with these RAMs.
But generally seems that they should work on your motherboard - well, some people do run them at 2.1V, some at 2.2V, thereby your motherboard should allow to set such voltage too.
http://www.asktheramguy.com/forums/show ... hp?t=64684
There are lot of more problems (and RMAs) with these RAMs.
But generally seems that they should work on your motherboard - well, some people do run them at 2.1V, some at 2.2V, thereby your motherboard should allow to set such voltage too.
Either the memory is defective (again?) or the GigaByte board. I have found lots of threads with people complaining that the BIOS voltage settings are incorrect, silently overvolting the memory. I noticed that my settings aren't very constant but fluctuate somewhere between 1.84 and 1.92. Setting the voltage to +.1, +.2 simply crashes the machine and +.3 apparently doesn't have any effect at all.
For the GigaByte voltage issues see for example here:
http://forums.tweaktown.com/showthread.php?t=25275
I have been a spoiled ASUS customer for years - from good experiences with them I just assumed mainboards by now would simply work and that stability was a non-issue. I'll never buy a GigaByte board again. Even if it turns out to be a memory problem, GigaByte's tech support is the worst imaginable.
Anyways, I'll investigate further and report back...
regards,
Sören
For the GigaByte voltage issues see for example here:
http://forums.tweaktown.com/showthread.php?t=25275
I have been a spoiled ASUS customer for years - from good experiences with them I just assumed mainboards by now would simply work and that stability was a non-issue. I'll never buy a GigaByte board again. Even if it turns out to be a memory problem, GigaByte's tech support is the worst imaginable.
Anyways, I'll investigate further and report back...
regards,
Sören
If so there are allot of issues with them not being able to reach there rated speeds. Try 400MHz (DDR2800) and boot with that.Thanks Mythy, will try that after I let the machine run the Prime95 torture test again over night. I seem to have reached a semi-stable state by shuffling around the memory modules again and playing with the BIOS settings. Here's what I did and my findings:
- removed one of the memory modules completely (I'm only running with one now)
- changed the remaining one into slot 2
This got me a booting system.
I then toyed with voltage and timing settings and checked with SpeedFan and in the BIOS.
I set the memory to it's specified 5-5-5-15 timings - no problems.
I tried the following voltage settings:
+0.0 => 1.95 volts
+0.1 => 1.79 volts (_less_ than +0.0!)
+0.2 => 1.84 volts (again, _less_ than +0.0!)
+0.3 => 1.90 volts
+0.4 => 1.95 volts
I didn't dare going any higher. Notice that it didn't seem possible to go past 2 volts.
Running Prime95 torture test blend for half an hour didn't report any problems.
My conclusions so far would be:
GigaByte's voltage settings are completely fucked up and don't bear any resemblance to reality whatsoever. Either that or the readout is broken. Unfortunately I don't have the equipment to manually measure the voltages.
Since one memory stick seems to run stable I could either assume that the other one is defective or that the corrupt voltage settings only affect memory operation in dual channel mode. If the stick survives prime tonight I'll test the other one in the same slot with the same settings. If both pass I'll have to assume the board is broken.
regards,
Sören
- removed one of the memory modules completely (I'm only running with one now)
- changed the remaining one into slot 2
This got me a booting system.
I then toyed with voltage and timing settings and checked with SpeedFan and in the BIOS.
I set the memory to it's specified 5-5-5-15 timings - no problems.
I tried the following voltage settings:
+0.0 => 1.95 volts
+0.1 => 1.79 volts (_less_ than +0.0!)
+0.2 => 1.84 volts (again, _less_ than +0.0!)
+0.3 => 1.90 volts
+0.4 => 1.95 volts
I didn't dare going any higher. Notice that it didn't seem possible to go past 2 volts.
Running Prime95 torture test blend for half an hour didn't report any problems.
My conclusions so far would be:
GigaByte's voltage settings are completely fucked up and don't bear any resemblance to reality whatsoever. Either that or the readout is broken. Unfortunately I don't have the equipment to manually measure the voltages.
Since one memory stick seems to run stable I could either assume that the other one is defective or that the corrupt voltage settings only affect memory operation in dual channel mode. If the stick survives prime tonight I'll test the other one in the same slot with the same settings. If both pass I'll have to assume the board is broken.
regards,
Sören
offtopic
What I mean - you can never judge about particular brand by just one bad sample.
(Well, I'm somewhat contradicting myself - I'll never buy another Deskstar named HDD again
)
Of course support is absolutely another thing. I've no experince - we here [in post-soviet area] either make things work or return them
Same for Corsair - I've used numerous Corsairs, no problems with standard ones. I've seen problems with one expensive "OC" stick on one specific motherboard, copying memory timings from another mobo made everything stable.
--
back to topic
About memory voltage - what exactly is configurable in BIOS? I remeber that in one of their P35 BIOS initially were activated various automatic voltage settings, which kicked badly in while overclocking (CPU got hot without reason) - after setting manual mode many things became adjustable and they didn't drift away.
But probably you've tried all possible settings anyway.
I'm waiting for test of another RAM stick - do they behave differently or not.
(According to Corsair documentation, these RAM modules have wired JEDEC SPD information for 5-5-5-18 at 400MHz/DDR800 - IMHO thereby they should run at these timings without any overvolting.)
I don't know about support, but my last experience with one Asus board is negative (DOA) and all 3 Gigabyte P35 boards, seen last month, are working and overclocking perfectly.BuschnicK wrote:I have been a spoiled ASUS customer for years - from good experiences with them I just assumed mainboards by now would simply work and that stability was a non-issue. I'll never buy a GigaByte board again. Even if it turns out to be a memory problem, GigaByte's tech support is the worst imaginable.
What I mean - you can never judge about particular brand by just one bad sample.
(Well, I'm somewhat contradicting myself - I'll never buy another Deskstar named HDD again
)Of course support is absolutely another thing. I've no experince - we here [in post-soviet area] either make things work or return them
Same for Corsair - I've used numerous Corsairs, no problems with standard ones. I've seen problems with one expensive "OC" stick on one specific motherboard, copying memory timings from another mobo made everything stable.
--
back to topic
About memory voltage - what exactly is configurable in BIOS? I remeber that in one of their P35 BIOS initially were activated various automatic voltage settings, which kicked badly in while overclocking (CPU got hot without reason) - after setting manual mode many things became adjustable and they didn't drift away.
But probably you've tried all possible settings anyway.
I'm waiting for test of another RAM stick - do they behave differently or not.
(According to Corsair documentation, these RAM modules have wired JEDEC SPD information for 5-5-5-18 at 400MHz/DDR800 - IMHO thereby they should run at these timings without any overvolting.)
The first stick of RAM checked out OK tonight (running 10 hours of Prime95 blend torture test). The second stick, used in the exact same slot with the exact same BIOS settings died during the first iteration.
Conclusion: I've got another shipment of bad RAM.
Worries: It might still have something to do with the motherboard's strange voltage settings/readings. Maybe the mobo fried the RAM in the first place or one of the sticks just happens to be more robust to undervolting/overvolting while both would work just fine with the correct voltage...
I'm gonna try replacing the RAM one more time. If that fails I'll try RAM from another manufacturer.
thanks for all the suggestions so far,
Sören
Conclusion: I've got another shipment of bad RAM.
Worries: It might still have something to do with the motherboard's strange voltage settings/readings. Maybe the mobo fried the RAM in the first place or one of the sticks just happens to be more robust to undervolting/overvolting while both would work just fine with the correct voltage...
I'm gonna try replacing the RAM one more time. If that fails I'll try RAM from another manufacturer.
thanks for all the suggestions so far,
Sören
Very unlikely, although not impossible.BuschnicK wrote:Maybe the mobo fried the RAM in the first place or one of the sticks just happens to be more robust to undervolting/overvolting while both would work just fine with the correct voltage...
I've seen statical electricity killing RAM (nice blue spark occured while one co-worker give RAM stick to another; we have some synthetic floors in our office
), I've heard of RAM killing RAM slot on motherboard (and vice versa) when attempted to insert backwards while PC was running (I hope you didn't make such experiments 
), but ruining motherboard by just bu using almost working RAM doesn't seem logical.First, run it at 667MHz, 5-5-5-18, 1.8v. Once in the BIOS, hit CTRL+F1, you should see it flicker, then you will have more RAM options in that section of the BIOS, and be able to set that (leave all but the CAS, tRCD, tRP, and tRAS alone*). See if it works like this. Try slots 1,3 if 0,2 does not work.
If that works, and you're OK with it, just remember to not get RAM like that again. If it doesn't work, it could be a mobo issue. Maybe not, even so, but could be.
If that works, but you're not happy having spent the extra $ (and think you can sell it, or if you can still RMA for refund or credit), ditch that RAM, get some decent cheap stuff that is rated for 5-5-5-x and 1.8v. If you are not willing to take the time and risk involved in getting a high OC, don't get any DDR rated over 2.5v, DDR2 over 1.8v, or DDR3 over 1.5v.
Here's what's behind those >1.8v ratings:
Thus: you're not buying RAM that is made with chips that the chip manufacturer says are good for 1066MHz operation, but buying RAM that the DIMM manufacturer says overclocks that high pretty well.
To top that off, ever since the Intel 845P, SiS 746, and nForce 2, buying high performance RAM for anything other than overclocking is just not worth it. The performance gains are seen in benchmark scores, but not felt when using the PC.
Oh, and DDR2 800 will get your CPU to 3.2GHz without OCing the RAM...you know, if you want to try it
.
* Don't be impressed. I had to Google for the names
If that works, and you're OK with it, just remember to not get RAM like that again. If it doesn't work, it could be a mobo issue. Maybe not, even so, but could be.
If that works, but you're not happy having spent the extra $ (and think you can sell it, or if you can still RMA for refund or credit), ditch that RAM, get some decent cheap stuff that is rated for 5-5-5-x and 1.8v. If you are not willing to take the time and risk involved in getting a high OC, don't get any DDR rated over 2.5v, DDR2 over 1.8v, or DDR3 over 1.5v.
Here's what's behind those >1.8v ratings:
- A given chip can switch at speed X, and is rated to at voltage J (as in, add EDEC).
- All can switch faster with higher voltage.
- Some chips will get damaged by higher voltages, have circuits that get out of sync when switching too fast, or otherwise have some issue preventing the whole chip from correctly switching faster with more voltage (we're talking about millions of little solid-state switches, here).
- Some designs or batches will switch faster with more voltage applied, no problemo.
- Said designs or batches are still made to run at speed X, voltage J, but good ones can run at higher speeds, with higher voltages applied.
Thus: you're not buying RAM that is made with chips that the chip manufacturer says are good for 1066MHz operation, but buying RAM that the DIMM manufacturer says overclocks that high pretty well.
To top that off, ever since the Intel 845P, SiS 746, and nForce 2, buying high performance RAM for anything other than overclocking is just not worth it. The performance gains are seen in benchmark scores, but not felt when using the PC.
Oh, and DDR2 800 will get your CPU to 3.2GHz without OCing the RAM...you know, if you want to try it
.* Don't be impressed. I had to Google for the names
Thanks Cerb,
I have started memtest with both sticks on the board (the good one and the bad one) and the 667MHz 5-5-5-18 settings you suggested (everything else on default) and it survived the first iteration without any errors. I'll leave it running tonight.
If I have a stable system at the cost of some performance (how big is the difference anyways?) I'll gladly choose stability.
Still - I'm very disappointed about how all of this turned out. Paying extra for what I thought was brand memory and fighting problems like this. GigaByte's support and board voltage settings also completely suck.
Anyways, big thanks to this forum and all the helpful comments - you guys rock!
Sören
I have started memtest with both sticks on the board (the good one and the bad one) and the 667MHz 5-5-5-18 settings you suggested (everything else on default) and it survived the first iteration without any errors. I'll leave it running tonight.
If I have a stable system at the cost of some performance (how big is the difference anyways?) I'll gladly choose stability.
Still - I'm very disappointed about how all of this turned out. Paying extra for what I thought was brand memory and fighting problems like this. GigaByte's support and board voltage settings also completely suck.
Anyways, big thanks to this forum and all the helpful comments - you guys rock!
Sören