[Team.AU] Core i7 965xe, Gigabyte UD5/Extreme, ADATA 1600+ 3Gb

19-11-2008, 07:13 PM #1 (**permalink**)

The last few weeks i've had the pleasure of playing with Nehalem based systems in various configurations (and in various locations).

Sampling Gigabyte's GA-EX58 UD5 and Extreme board line up has been my main priority, but at the same time ive managed to squeeze out some decent bench results in time for launch.

Thanks to our various supporters for supplying great hardware to test:

Intel - Core i7 965xe ES, 80Gb 25-M SSD, TRUE Extreme
Gigabyte - UD5 and Extreme mobos, Core i7 965xe's ES, 4870x2
ADATA - Tri-channel 1600+ 8-8-8-24 (3 x 1 Gb kit)

My experience thus far on air is great, with a TRUE strapped to the chip you can easily push out 4GHz plus multithreaded tests, and at that speed and the right tweaking you can have yourself a blazing fast monster PC.

On subzero thats where things get a little more shaky, with various chips ive tested so far having coldbug/boot problems at temperature much warmer than extreme benchers are used to (e.g. -60C). The effect of vcore is also slightly different, with chips having a preferred maximum voltage above which their behaviour becomes erratic.

Tweaking for performance is much more satifying than with Core 2 for me, quite reminiscent of the s939 days where you can play with multiplier, base clock, uncore clock, and QPI clock combinations to get the most from your CPU, memory, and PCI-E lanes. Add to that the usual miriad of votlage options available on the Gigabyte boards and you have hours of fun tweaking ahead.

Hicookies notes prior to some board testing:

The Gigabyte boards are performing really strong for me, no show stopping bugs (ala P45T

and robust stability for even the highest loads (e.g. 5GHz+). The BIOS feels really good, with only some minor details for extreme overclockers to be resolved such as PCI-E frequency adjustment problems and a vcore 'bug' above 1.6V causing an extra 0.04V on top of your BIOS set value. LLC is also not working properly yet, but that can easily be resolved with a pencil/solder/paint vdroop mod as below.

The ADATA tri-channel kit is going really nice, with max benchable speeds so far of ~2050MHz @ 8-9-8-24 1T @ 2V.
I like to run them more around the 1900-2000 range which allows me to tighten things up a little.
20K+ in everest is really simple to achieve, and SuperPI 32M definitely loves it that way

So far ive managed to run QPI up to around 7400+, uncore up around 4500+ which brings about some nice efficiency gains in 3D and 2D respectively.

Here's some links to results ive achieved so far using my better 3831Axxx ES CPU:

5267MHz CPU-Z: http://www.hwbot.org/result.do?resultId=791460

7m 31.950s SuperPI 32M: http://www.hwbot.org/result.do?resultId=791900

18167 PCMark'05 (no iRAM): http://www.hwbot.org/result.do?resultId=792285

2m 46.250s wPrime 1024M: http://www.hwbot.org/result.do?resultId=792600

And finally, here's one of me overclocking @ 5GHz at the Core i7 launch party last night in Singapore:

19-11-2008, 07:38 PM #2 (**permalink**)

Nice start Tim, so the first set of photos/rig setup were at Gigabyte HQ ?

19-11-2008, 10:04 PM #3 (**permalink**)

Very nice mate, top job indeed!

The X29 multi seems to be doing the trick I gather?

20-11-2008, 01:20 PM #4 (**permalink**)

x29 is more than high enough to get max clocks at low enough baseclock
Yeah the photos on the desk are in the office in Taiwan

20-11-2008, 02:12 PM #5 (**permalink**)

Added to frontpage news http://i4memory.com/wp/article/505

20-11-2008, 11:08 PM #6 (**permalink**)

hope you dont mind i posted your thread here
http://forums.guru3d.com/showthread.php?t=279413

21-11-2008, 03:35 PM #7 (**permalink**)

Nice results Tim, you're popping up all over the net with this.

Quick question, (for anyone) what do you reckon would be a good psu (wattage) to run this type of setup with SLI 260's?

Cheers

21-11-2008, 10:12 PM #8 (**permalink**)

I reckon a decent 800 will do you fine