Official GIGABYTE X38-DQ6/X38T-DQ6 Info/BIOS Thread

[freqy]

"I have been using speedfan to monitor"

speedfan generally reports lower temperatures than actual figures
almost everyone on here uses coretemp. a free temp prog.

http://www.alcpu.com/CoreTemp/

dont worry , you temps seem low enough even for speedfan.
but i highly recomend coretemp as will most people on here.

hope this helps.

peace

[Frostybyte]

I'm still having problems. The computer reset, not froze, not long after I posted that.

Then, while adjusting settings, I couldn't get the machine to post without restarting.

I only have the G.skill in atm. One of the Crucial sticks died as well so I'll be RMA'ing it.

Right now I'm priming again with memory at 400mhz and 4.5.5.15. Really frustrated to be honest.

[ChrissTi]

Frostybyte your cpu is undervolted. For 3.6 Ghz on quad Q6600, I think you should have at least the voltage I have on my dual core, meaning 1.65V ... (You have two more core power hungry ). Rise to 1.65 and watch the temps in Everest, while testing with two orthos at once (CPU only stress). And leave the mem at 900 Mhz (2.0B multiplier). DDR2 voltage +0.45. Timings 5.5.5.15.
If you got CPU stable then play with mem timings.

[ChrissTi]

Frostybyte forgot to tell MCH voltage +0.100

[~aoe~]

@Frostbyte: If you suspect the memory, then I would suggest running Memtest86+ to check it. Prime95/Orthos is not suitable for confirming memory stability.

Problems like this are best solved by process of elimination. If your memory stability is ok in Memtest, then you can consider ChrissTi's suggestion on the vCore, although I'm running a Q6600 @ 3.6GHz with 1.5v (1.472 in Windows) without problems.

As suggested, the vMCH may be something to look at. I haven't played around much with that on this board, but it can help stability.

[2DIFF]

Happy new year everybody!
Someone once wrote in this thread that the PSU quality has something to do with the fact that you could lower the core voltage all by getting the same OC results. So I changed from my Thermaltake ToughPower 700W back to my Seasonic SS-500HT, just by curiosity. Now, with an OC of 75% and 2 HD3870 in CF, 500 watts should be insufficient. The thing is that I'm using a little gadget called "Energy Check" (see image), which tells me the power consumption of my rig. It's 262 watts under full load, by which I mean that Orthos is doing a blend stresstest, the HD3780's are both @ 840/1316 in CF and are having fun rendering some globes in rthdribl (the Masaki Kawase tool) with 16x Multisample. Did this for an hour (remember: one hour of vacuuming heats our planet much more ). With the TT 700W PSU (about which I wrote here before) I had to push the cpu voltage to 1.66 for 3.68 Ghz, with the SS-500HT I could lower it to 1.500, below that the system became instable. The next thing I'm going to do is buy myself a 300W PSU (costs nothing) just for the fun, because it should be sufficient for 262W max power consumption . Now I'm just wondering why people (me too) buy 700 to 1KW PSUs...
Here's the Brennenstuhl PM 230:
Attachment 69931

[ChrissTi]

~aoe~ if the cpu gives errors then automatically the memory gives errors . So first thing to do is checking cpu stability. Orthos has a cpu only checking, and it works. When I did cpu only checking it gave me errors till I've reached the right CPU voltage. After that I played only with memory tests. And to test memory the best thing is to test in multitasking, memory tests programs gave me no errors. When I runed a few programs, it turned that the memory was not running well. So now my mem is at 900 for running stable. Right now I'm testing to see what are the best timings for the modules I have. I'll wait and see maybe the price will drop on 2x2GB ddr2-9600 low timing modules.

[Rob94hawk]

@Frostbyte: If you suspect the memory, then I would suggest running Memtest86+ to check it. Prime95/Orthos is not suitable for confirming memory stability.

Thanx for the link!

[~aoe~]

@2DIFF: Interesting info there. However, I wouldn't be too optimistic about running that system on a 300W PSU. The overall wattage of the PSU isn't what's important, more how that power is distributed across the various rails. Specifically, the power available on the 12V rails tends to be most important.

There's a very good thread here on PSUs with lots of links to good further info.

[ChrissTi]

2DIFF just waiting to see how it worked with the 300W PSU By the way there may be spikes that the gadget is doesn't show because of low time response.

[~aoe~]

~aoe~ if the cpu gives errors then automatically the memory gives errors . So first thing to do is checking cpu stability. Orthos has a cpu only checking, and it works. When I did cpu only checking it gave me errors till I've reached the right CPU voltage. After that I played only with memory tests. And to test memory the best thing is to test in multitasking, memory tests programs gave me no errors. When I runed a few programs, it turned that the memory was not running well. So now my mem is at 900 for running stable. Tight now I'm testing to see what are the best timings for the modules I have. I'll wait and see maybe the price will drop on 2x2GB ddr2-9600 low timing modules.

It is true that any system instability can manifest itself as memory errors. However, this does not necessarily mean that the memory itself is unstable.

By testing the memory with memtest, you are confirming whether the memory alone is capable of running as set in bios (with virtually no CPU load). Once you have confirmed this, you have eliminated one possible problem. Stress testing with Prime95/Orthos then tests overall system stability.

Again it's possible that an unstable system can be made stable by simply relaxing memory timings. However, memory instability under load when the memory is stable under memtest is generally caused by the CPU/motherboard (unable to communicate reliably with the memory at the set frequency/timings).

The generally accepted 'correct' methodology for overclocking involves the following steps:

1. Confirm max CPU overclock using very relaxed memory timings - test with Prime95/Orthos
2. Confirm max memory overclock with CPU at stock or lower - test with memtest or equivalent
3. Try and find the best combination of CPU and memory overclock that is stable under load (Prime95/Orthos for testing).

Steps 1 and 2 are interchangeable.

Step 3 generally involves taking the highest CPU overclock that you are comfortable running and then trying to get the memory as high as possible (within the constraints we determined in step 2).

Step 4 could include trying to run the memory with tighter timings.

Whilst the above is the 'best' way, it can be time-consuming, and therefore many of us tend to take shortcuts (try and do both memory and cpu at the same time). However, when we hit problems, they can take much more time to identify and resolve as we have 2 (or more) possible causes to investigate.

I still suggest the following:

1. Test memory with memtest to check if it is stable with no CPU load to confirm the memory is capable of running as set in bios
2. Test CPU under load with very relaxed memory timings to confirm CPU overclock stability

If both the above tests are successful, then we can eliminate 2 possible causes for the problem and focus on stabilising the communication between the memory and cpu (either by relaxing memory frequency/timings or increasing voltages where possible).

[Hornet331]

~aoe~ if the cpu gives errors then automatically the memory gives errors . So first thing to do is checking cpu stability. Orthos has a cpu only checking, and it works. When I did cpu only checking it gave me errors till I've reached the right CPU voltage. After that I played only with memory tests. And to test memory the best thing is to test in multitasking, memory tests programs gave me no errors. When I runed a few programs, it turned that the memory was not running well. So now my mem is at 900 for running stable. Right now I'm testing to see what are the best timings for the modules I have. I'll wait and see maybe the price will drop on 2x2GB ddr2-9600 low timing modules.

thats not completely true.

orthos/prime allways errors if your ram is instable.

Thats why i prefer to test first the ram for its max frequency (with cpu multi 6) and then try the cpu.
But i have a bad feeling that i cant max out the ram, max FSB i reach till now was 470. And my ddr3-1800 sticks do this without sweating at stock ratings. (1,9V)

[stevke]

So, I've tested the 2.00B divider, I can boot with 400FSB into Vista, than I've tried to increase the FSB with SetFSB, I reached 435FSB until it crashed, should be the CPU since I've set the Vcore only to 1.425V.

Today I'll test higher Vcore and try to boot with a FSB above 400MHz.

Any suggestions to the other voltages?

[Hornet331]

herst stevke warum fragst nicht im oc.at forum?

anyways

well max "recommend" overvoltage for the Vmch is 1,6V (default is 1,25V) which means you can give up to 0,35+ but i wouldn't do that without a fan blowing on the NB.

I havent found a stock reference voltage for the Vfsb, i have tried it up to +0,3 right now and it gives a bit more headroom, but not much.

Also it seems Gigabyte boards like voltage on the Pcie bus, on my DS4 i had to give 0,2+.

[stevke]

@Hornet: Ich dachte mir, wenns hier schon einen Sammelthread gibt...

I'll give it a try with more juice this afternoon.

[ChrissTi]

Hornet331 Maybe , But my memory freq/timings were not ok and, when I did ORTHOS only CPU stress until I get CPU stable a certain voltage for a certain CPU frequency (example I got CPU stable on FSB 444MHz/Voltage 1.5875, and FSB 450/Voltage 1.65). Memory and CPU tested together with orthos gave me errors.
After that I played with memory timings and multipliers until now. Finaly I can run Orthos, play Crysis, and listen to music, all in the same time. I think is very stable, and I can even try to overclock a little more the cpu cause I think that 1.65 is a little too much for the cpu, and I think it will hold to oveclock a little bit more.

[Richard Dower]

Where can i download the latest build of MemSet?

[~aoe~]

version 3.4 beta 3 is available at here. Looks like you may need to register to download though.

I've recently downloaded it but can't remember where from

If I remember or can find it, I'll let you know.

[Richard Dower]

Can someone take a look at my memory timings?...are these normal/standard:

[Richard Dower]

version 3.4 beta 3 is available at here. Looks like you may need to register to download though.

I've recently downloaded it but can't remember where from

If I remember or can find it, I'll let you know.

I downloaded version 3.4...are the betas a more recent build?
I see a 3.4 beta 4 available.

[Richard Dower]

HOLY CRAP!!...i got a MASSIVE boost when i altered the subtimings using memtest:

With default subtimings i was always just under 8000MB/s

Thanks to CapedCrusader for the guidance:

[ChrissTi]

Richard Dower: the memory timings are subjective to any manufacturer of memoryes and mainboard, as I see, so there is not standard. There is only memory timings tested to work on many mobos and put on SPD mem. Look at your mem SPD with CPU-Z or EVEREST and see what are the close upper timings to your mem frequency.
Where did you get F7j bios ? Do you have a link or an upload with it, please ?

[Richard Dower]

The memory latency has dropped a huge amount, was around 66ns...look at it now!

[ChrissTi]

Richard Dower: I advise you not to modify mem timings, unless you know what you are doing, it will result in loss of data, or coruption and even to make the OS not to boot . I had that experience myself. Now on my mem I know what are the right timings, but on yours I don't know.

[Richard Dower]

Richard Dower: the memory timings are subjective to any manufacturer of memoryes and mainboard, as I see, so there is not standard. There is only memory timings tested to work on many mobos and put on SPD mem. Look at your mem SPD with CPU-Z or EVEREST and see what are the close upper timings to your mem frequency.
Where did you get F7j bios ? Do you have a link or an upload with it, please ?

Altered my sig to reflect correct bios.