GD80/UD6 and 750 at 5.2G

[Massman]

Time to start using the Hwbot forums more often; it's too quiet in here!!

 

I've been testing the GD80 two months and a couple of days, but only recently I was able to lay my hands on a retail 750 (thank you, Leona from MSI). Got the chip in Beijing, went home, was bored, so threw it under some LN2. The whole session lasted about 5h maximum and wasn't particularly focussed on anything besides getting familiar with the retail i5. Anyway, I'm done talking, let's give some numbers.

 

Chip:

 

 

10 sheets of paper towel - no condensation whatsoever. What's all this BS about eraser being the best

 

 

Results

 

 

 

 

 

 

For those who are interested, I also did a quick Everest benchmark. This was after I finished the 32M run.

 

 

Everything can also be found at the bot.

 

I didn't push VTT higher than 1.55V as the board had some instability issues. Forwarded everything to MSI, so let's hope for improvement. It does seem, though, that all 750's have inferior silicon quality compared to the 800 series as almost everyone maxes out at around 240-246BCLK.

[Massman]

Coldbug and coldboot are a bit of an issue; not only with this motherboard, but probably with the chips. I did all these runs at temps between -60°C and -70°C; -80°C was still pretty much stable, but anything lower than this results in complete instability

 

Coldboot wasn't a problem as long as you don't mind resetting the CMOS. Everytime I had issues booting I cleared the CMOS and the board could boot up again.

 

Btw, this combo was very vcore/vtt sensitive. You need to find a balance between vcore/vtt to boot, preferably setting vtt higher and vcore lower. With the DIP switches you can adjust the volts quite easily and that's how I did it: low vcore/high vtt boot, +0.1V vcore via the dip switches just before I enter Windows.

 

Oh, right: issues above 1.6V VTT!

[bazx]

mass i have a bios here A7581IMS.154 any good to you

[Massman]

I used the 1.5B2, but I reckon 1.5B4 is as good as that one. None of the previous biosses (except the very first beta releases) have ever let me down.

[bazx]

do you think the cold problem is in the chip or the m/b

 

i will be trying some sub 0 this weekend i will let you no my findings

[Massman]

I want to say it's the chip, because a lot of people are pointing in that direction, but I can't help but thinking the board is maybe a bit to blaim. I will test my 750 on the UD6 this weekend, so I'll know for sure the board was holding me back or not.

 

Let me know what you think; i'll forward the stuff to MSI HQ anyway, would be cool to know if I'm the only one having issues or not . Btw, don't try high PCI-e clock, it doesn't really work. 102-103 maximum ... with earlier biosses I could get it running at 110MHz, but it seems they've "fixed" it, hehe. Doesn't matter anyway, it helps maybe 3-4MHz BCLK.

[Massman]

Tested the more more intensively last night; a couple of remarks:

 

- Had a lot of issues with memory multipliers: couldn't boot over 1500CLx and higher than 1800CLx was quite unstable. Voltage/timing change/B2B didn't change this.

- Before I used the 1.5B2 bios and this one had an issue with coldbug on the PCI-e. You know it's this issue when the boot locks at "2E"; just change VGA card from pci-e lane and you are good to go again. In bios 1.5B4 this issue has been fixed *hooray*

- On single stage I'm limited to 237BCLK

- 1.5B4 still doesn't allow me to use more than 1.6VTT

 

The one score:

 

 

I just installed the UD6, let's see what that brings

[bazx]

mass do you have the x-ratio function in cell menu

 

i can only set multi 22x is there one extra?

 

baz

[Massman]

That's the forced turbo mode function.

 

Ok, for 750, you have the following turbo modes:

 

4 cores = 21x

3 cores = 22x

2 cores = 23x

1 core = 24x

 

To open up the forced turbo modes, set EIST to AUTO. Then, if you want 23x or 24x, set number of cores to 2 or 1. You can't force all 4 cores to work at 24x.

[bazx]

thanks for that

 

helps quite a bit

[bazx]

hehe just set it to 1 core

 

and the next multi is 27

[Massman]

Don't think it'll help much in LN2 situations: the upper core ratios aren't that wonderful, even the odd multipliers.

[Massman]

I've been playing with the P55-UD6 on air and single stage these last couple of days and I found a whole new appreciation for the GD80 board .

 

As for the UD6:

 

- Memory bandwidth performance is excellent; very easy to reach high numbers without actually pushing the memory.

- Bios is absolutely fantastic in terms of lay-out

- Board feels smooth overall (<-> gd80), but that's a very subjective opinion. Don't know what it is, just feels smooth

- Hits 150MHz PCI-e with small increase in voltage - too bad not much performance increase

- On air it clocks pretty much the same as the GD80

 

- On phase I can run 230x20 as stable as on the GD80, but with the GD80 I can push to 233MHz 01 stable whereas it's nearly impossible with the UD6. Reason underneath.

- A lot of issues with vcore/vtt setting. I boot 1.50/1.50 or 1.518/1.52V, but anything over that just doesn't work ... on the MSI I can't go over 1.54V VTT either, but I can push a little bit harder on the Vcore on single stage, hence why I can bench a tad higher.

 

In any case, thanks to the better memory and uncore performance I've been able to increase my 01 score with 1200 points using less mhz.

 

[Massman]

- Hits 150MHz PCI-e with small increase in voltage - too bad not much performance increase

 

Don't go high in PCI-e when subzero; although 150MHz was stable on air cooling, with the CPU at -30°C 115MHz already gives problems.

 

On phase I can run 230x20 as stable as on the GD80, but with the GD80 I can push to 233MHz 01 stable whereas it's nearly impossible with the UD6. Reason underneath.

- A lot of issues with vcore/vtt setting. I boot 1.50/1.50 or 1.518/1.52V, but anything over that just doesn't work ... on the MSI I can't go over 1.54V VTT either, but I can push a little bit harder on the Vcore on single stage, hence why I can bench a tad higher.

 

Switched to bios F5c which fixes the vcore/vtt issues I had. I'm now running 1.57/1.58V without any issue!

 

Note that the maximum BCLK hasn't improved; actually: I can still only run upto 231,5MHz stable, whereas the MSI did 233 before. Clock per volt the GD80 still is a tad better, but the UD6 has the advantage of the bandwidth. I'll have to re-run the GD80 with another bios, though, as my 89190 01 score does suck because of the 1400CL6 memory configuration. I'm hoping it was just a quirk of the 1.5Bx bioses I used ... can't be anything else really: I've been running 2000CL9 for my reviews

 

Score is coming.

[Massman]

Forgot to update with my best run for the day:

 

[Turrican]

good work

[Massman]

I'm hoping it was just a quirk of the 1.5Bx bioses I used ... can't be anything else really: I've been running 2000CL9 for my reviews

 

Quickly retested this morning: 1.5B4 equals issues; 1.5B2 equals 2000CL9 perfectly bootable.

 

[bazx]

interesting

 

what was 1.5b4 supposed to fix

[Massman]

I think 1.5B4 fixed the PCI-e coldboot error as I described in post 7.

[bazx]

always the way

 

one bug out

 

one new one in

[Monstru]

Nice work Pieter. Weird stuff the vcore issue though, I used F4E a lot and I could run 1.63v on LN2 without any problems. The werid problem I had with UD6 is the multi dropping when I adjust BKCL in Windows. Also, if OC fails in Windows under LN2, I have to reset the BIOS in order to boot (a problem that I also have with GD80, but not with Maximus III). Other then that, UD6 does indeed perform very well. From the 3 (UD6, GD80 and Maximus III), under LN2, the GD 80 impressed me the most untill now. It just feels right

[Massman]

Nice work Pieter. Weird stuff the vcore issue though, I used F4E a lot and I could run 1.63v on LN2 without any problems.

 

With the F5c bios I have no bootup issues anymore: 1.7V vcore and 1.7V vtt with -65°C is perfectly bootable. It's very good Gigabyte fixed this already.

 

The werid problem I had with UD6 is the multi dropping when I adjust BKCL in Windows.

 

Not an issue of the board, but of the version of CPU-Z you're using. For 64-bit OS versions you should switch to the 64-bit version of CPU-Z. I've had the multi-drop issue before, but switching to the latest version fixed it for me.

 

Also, if OC fails in Windows under LN2, I have to reset the BIOS in order to boot (a problem that I also have with GD80, but not with Maximus III). Other then that, UD6 does indeed perform very well. From the 3 (UD6, GD80 and Maximus III), under LN2, the GD 80 impressed me the most untill now. It just feels right

 

Okay, here's what you do with the UD6 if you have the 88 debug code:

 

- Switch of power supply and wait untill all leds of the board go out

- Boot the board; it will go: C1 -> 88 -> shutdown

- Wait for the board to restart

- Debug will go a couple of C1's and then boot IF the frequency/voltage/temperature allows it

 

I've done this a couple of times and it always works for me; never had to reset the CMOS during my last session.

 

As for the GD80: same issues here, but no easy solution for this. Imho, the easiest method is to reset the CMOS and use the OC profile to get started. Doesn't take that long anyway so who cares, hehe.

 

Btw, you're right about the GD80: it's definitly not a bad board. It can run the same temps as my UD6 and is in fact a little better in the clock per volt department. Only issue with the board is that the vcore/vtt problem still exists, which makes it a challenge to boot with the right voltages. As I've mentioned earlier, this is where the v-switch comes in handy big time.

 

I'll update this thread with more info tomorrow when the valid.canardpc.com website is up and running again. I've got some nice validation on my 750

[Massman]

I retested my 750 on LN2 yesterday evening with some surprising results. First of all, I must say that the Gigabyte UD6 is a very, very solid board for extreme overclocking: no voltage issues whatsoever, which makes you go that extra percentage faster.

 

http://valid.canardpc.com/show_oc.php?id=714433

 

Also ran 1M 8.25s and 17.17s pifast. Those are on the bot, didn't bother to upload them here.

 

Temperature:

 

-boot: -65°C

-benchmark: -75°C to -78°C

-validation: to -83°C

 

The lower I go, the more unstable the system gets so it's really a matter of finding the right temperate for your benchmarks. It's also noticeable in the benchmarks itself: at a certain point I had to re-try the pifast at 240.2MHz BCLK 5-6 times and 241MHz was okay from the first try.

[Splave]

any luck getting the rated fsb over 4 ghz? 248 seems to be close to my ud4p max albeit lower multi. 5.2ghz is nasty

[Monstru]

Not an issue of the board, but of the version of CPU-Z you're using. For 64-bit OS versions you should switch to the 64-bit version of CPU-Z. I've had the multi-drop issue before, but switching to the latest version fixed it for me.

 

Sorry man, but I have checjed the real frequency with some special tools, and I also checked it with WPrime, and the multi drop is real. And I am using the 64bit CPU-Z for x64

 

 

Anyway, I am glad to hear that the new bios improves the booting issue, I will give it a try for sure.

 

 

Only issue with the board is that the vcore/vtt problem still exists, which makes it a challenge to boot with the right voltages. As I've mentioned earlier, this is where the v-switch comes in handy big time.

 

I am sorry, I didn't see where you have explained this problem. It needs a special ratio between the two, or what is the problem exactly?