MSI Elite OC Summit - Three results // Wprime record

For those who do not know, I was at the MSI Elite OC Summit in Taipei this week to test and provide feedback to the MSI team. The actual overclocking was on the first day of the summit (on monday) from 11AM to 18PM. I was mostly testing out the Clarkdale overclockability on the Trinergy and then switched to LN2-ing the 5870 with 1h30 left to go. Anyway, here are the results ... sorry, no system pictures. I'm sure that many people will post a lot of pictures in the next couple of days.

I was amazed that it was actually possible to run Wprime so close to 6G with this chip as initial LN2 tests lead to believe me that this chip was hardly capable of 6G. With only 1.7V and ~105°C this should actually be a pretty decent chip.

I was always booting in windows at this frequency ... ran the 1024M just to complete the benchmark list.

Had about 1h30 minutes when starting to prepare the system. Card went really high without doing any effort ... no tweaks, didn't check any scaling. The ref design suffers from serious OCP, tho, which kinda sucks. This was ran at 1.47V and ~ -100°C.

Conclusive thoughts: I ran this exact combination at home a couple of weeks ago. Due to serious vcore/memory issues, that session lasted exactly 10 minutes, so I'm sure you're not surprised when I say that I was a bit skeptical when booting up the first time. Messing around in the bios a bit seemed to show the real power of the Trinergy and I'm liking it more and more. For reference, here are the three main issues when dealing with this mainboard:

1) You will have coldbug and coldbootbug! I had them at -90°C and -110°C. Clarkdale is scaling insanely with temperature, so MSI will have to figure this one out if it wants to see high frequencies on a lot of chips.

2) The memory dividers are, not surprisingly, a drama. On air cooling, the BCLK oc difference between 2:6 and 2:8 was 30MHz(!): with 2:6 I could boot into windows at 220MHz no problem, with 2:8 only at 190MHz. Maybe consult Gigabyte to fix this one?

3) For this specific bios, the upper vcore limitation was 1.7V. Above 1.7V, it looked like there were no actual bios settings (although bios said there was) and setting a higher vcore resulted in a non-bootable system.

FYI: I was running bios 1.2B3. Also tested bios 1.2B4, but this one was worse, so quickly returned to the B3.

Oh, nice chip! Quite accidentally I did my first i5-661 tests @ air today and tested the Trinergy at some weeks ago @ air, too. So it`s good to see what the hardware is capable of under subzero conditions.

...but 03 screen resolution sucks.

Hm, seems like photobucket is resizing too much ... here's the link to the score: http://hwbot.org/community/submission/947139_massman_3dmark_2003_radeon_hd_5870_102674_marks

As far as I can see, the coldbug is still present on all mainboards apart from the Evga. I've tested several boards with the same CPU and coldbug/coldbootbug are all within 5°C range.

Last quick test for the day. Vcore ramps up to whatever I want and coldbug can be controlled very easily. For Wprime, this board is no doubt extremely solid ... I might even be able to push the chip harder than I did now as I'm getting BCLK limited. There's some issue with BCLK/MEM which is making the system seriously unstable. Can't run SPI 1M at wprime 1024M speeds LOL. Anyway, good for today ... have to wait a couple of weeks before I can run again; don't want to hit hwbot top-20

http://valid.canardpc.com/show_oc.php?id=967816

Another update for this board.

http://valid.canardpc.com/show_oc.php?id=970217

Went upto 5175 but didn't bother to take validation. QPI's in the right area

Boards's incredibly strong in multi-threaded stress applications, but fails bigtime for the other 2D apps at the moment. I think I say enough when I state that SuperPI 1M isn't even stable at 230MHz.

On the right path, but still work left to do

Some pics for those who like to see the workspace: