SF3D OC Gathering Vol. 1 -Powered by AMD

[MTP04]

Awesome stuff guys!

This is getting me excited about AMD again.

[Vivi]

wicked stuff

[SeLecT]

Awesome results!

[FlanK3r]

nice scores

[carpo93]

congrats

[Beleno]

great job guys, seems you had fun

[massman]

good to see Gigabyte kicking some ass there as well with their boards

For the 7GHz validation, we used the new DFI board with digital PWM, because the first UD5 wasn't doing -200°C all too well.

In total, I think there were almost 10 different motherboards used in the overclocking event and it seems that almost ALL motherboards are capable of maxing out a CPU. The difference between boards is really, really small. Which is a good think, of course

These are ORB W.R.s. Mikeguava has the current 05 record over there.

i dont think they will put these up on HWBOT being super cherries and all but i could be wrong

After having spent testing all three variants of the 955 (retail - TWKR - high-leakage), I have to say that the word 'super-cherry' is a bit misleading. The results you use to found your 'cherry'-argument are in fact the LHe results, which you can't compare to any of the LN2 results out there. In fact, these Phenoms scale so incredibly well with temperature that basicly every degree matters. To give an example: the whole team was working very hard to get the lowest temperatures possible when aiming for 7G stability. Waiting several minutes to push the 'apply'-button to have just that 0.1°C lower to make it stable. Hell, we even overclocked the LHe dewar (increasing pressure not guaranteed by manuf.) to get that 7G run.

The difference between TWKR and High-leakage is probably less than 200MHz suicide and around 100MHz in 3D. GiorgioPrimo and me were pushing a sample on the event and hit 6625MHz CPU-Z, 4950MHz NB and 6.25G 06 cpu test stable.

The difference between TWKR and retail is a bit bigger of course, especially when you try to compare the average overclocking results. However, there are some reasons for this (which I mentioned on Hwbot as well). Looking at the results of the retail sample of Loopy83 (link), I'd say even retail can get close to the TWKR. Of course, there's the difference of expectancy: with high-leaked samples, a high overclock is almost guaranteed whereas with retails you just don't know.

I still have some results on my laptop. I'll post them here in a couple of minutes.

[TraX]

Congratz on all those great results & some WR's guys!

Looked like you all had alot of fun

[SF3D]

Massman said it really well. TWKR did show some potential during the weekend.

Me and macci did not have too much luck, cause in these 955 ES CPU's there was few, which did not scale under -200c tempratures. The fisrt we tested did not boot at -230C anymore. Luckily Massman and Giorgioprimo was testing very nice chip on LN2 at the same time and we just carried the setup to the table where Lhe pipe was and started to go lower on tempratures. That CPU did scale well and we "overclocked" the helium tank by shutting down the safety valves + feed the tank with some gas. I think only overclockers are brave enough to do that with LHe dewar We just aimed to 7GHz and when we got it time was running out. If I would have known in that point, that we will have more bad luck with CPU's I would have done the 3D runs directly after that.

On sunday we tested two motherboard and one CPU under helium again and the CPU had a cold bug. It did 6750 validation easily under LN2, but when temperature was getting up to -240c the CPU did only 6.4GHz anyomre.
I think we all can accept, if CPU does have a cold bug at that temperature

Well.. luckily elmor was running his Gigabyte UD5 in the second room and we just swapped the setup. At that point we put the pressure to the max and closed all the safety valves
Temperature was under -250C all the time in some runs and that did help a lot with that CPU which was being used. No cold bug, just scaling all the time.

Well, now we have a lot of experience of runnign LHe and how to use these dewars + pressure controlling. I think we might have to do some more runs in the future.

Nice weekend for sure and this event was a great success, with sick scores. Thank you guys! You are the best and you all know it too

Pictures etc. coming a bit later.

[dinos22]

thats interesting about -250C after pressure adjustments
good stuff

[Solarfall]

hahahah helium tank overclocking

[kiwi]

Massman, what are the costs of LHe in your area, or in Finland for example? Looking at the price 30x more expensive than LN2 makes it not really practical despite great scalability with temps

[Oliver]

SICK , thats what i call ... ,

keep pushing it

[massman]

Luckily Massman and Giorgioprimo was testing very nice chip on LN2 at the same time and we just carried the setup to the table where Lhe pipe was and started to go lower on tempratures.

Haha .

That was really a great moment during the event. I think it was Macci who ran upstairs and asked Giorgio and me if we'd be willing to shutdown the system at -196°C, unplug everything, run downstairs (setup still at -190°C) and go LHe with the setup. A fine example of teamwork: people unpluggin everything, people ready downstairs to get the system running ... really awesome!

Massman, what are the costs of LHe in your area, or in Finland for example? Looking at the price 30x more expensive than LN2 makes it not really practical despite great scalability with temps

I think around €10~15 per liter in average. I asked my local dealer here and he confirmed this price and for this event I think it was around that as well. It depends from dealership to dealership, I reckon, as Petri told me that the first place they asked for a price, it was much more expensive.

But yes, the scaling is fantastic. A difference of 0.2°C can make the difference in stability or no stability. I'd very much like to see retail 955's (from the shelf) under LHe ... just to see the difference with the high-leaked ES samples.

[cowie]

WOW
going by the pics you guys left your game faces home and had a good time....the scores just fell into place after that

[NeoForce]

I wanna say many thanks to everyone on this event.
It was really awesome. Teamwork, organization - everything on the highest level!

[massman]

For those who wonder, this is what the TWKR chip could do under LN2.


http://valid.canardpc.com/show_oc.php?id=599028

CPU-Z 1.52 reads out TWKR correctly, but no validation possible with that version.

NB clock was nice also:



4.7G with 940MHz 7-6-5 1T was 3D stable, by the way.

[DeDaL]

Just now i sit at NeoForce's workspace. Tomorrow i will upload some fun pics
I just want to say many thanks to AMD for this great event. It is big honnor for me to bench with that kind of Top-guys.

[@rne]

4.7G with 940MHz 7-6-5 1T was 3D stable, by the way.

Veery strong imc

[LardArse]

how about pulling a huge vaccuum with a strong compressor to lower BP of ln2 / LHE for phenom benchin?

[Leora]

Liquid helium???? wow!

[Sampsa]

First of all, what a great event it was. Everything was very well organized and prepared by SF3D and macci. We did not only focus on overclocking but had lots of fun, did some drinking, sauna, swimming, eating, night club, etc

I think we will see pictures and results from everyone who attended the event so I'll just post the pics of liquid helium tests with Phenom II X4 955 Black Edition ES, Gigabyte 790FX-UD5P and ATI Radeon HD 4870 X2 CrossFire X me & elmor did on Sunday evening. Since SF3D & macci were mainly focusing on liquid helium, me and elmor established a Finnish/Swedish team and started benching together on Sunday. Everything worked out really well and first we tested normal TWKR CPU with LN2 and just two hours before we had to leave the place, we got opportunity to bench our well prepared system with liquid helium.

Since we only had very short time with LHe we started benching 3DMark05 straight away at 6,497 GHz and scored 49187 points. Next we raised the CPU to 6603 MHz and scored 49568 points. The best one was at 6678 MHz and score was 50017 points. We tried pushing 05 many more times but after scores like 49999 we just ran out of time

Then we did 3DMark06 and since we were in a hurry we started @ 6541 MHz which already gave us 38139 points. We could run CPU-tests successfully @ 6541 MHz but 6614 MHz crashed at the end of CPU-test 2. Unfortunately full run @ 6541 MHz crashed in CPU-tests. Here are the pics of system etc:



400 liters of liquid helium



macci & SF3D preparing the helium





Macci adjusting liquid helium flow



Tank's safety valve "overclocked"



Happy LHe-overclockers elmor & Sampsa



Kingpin F1 EE pot's temperature showed below -250 in idle and between -230 and -240 under 3D load. These might not be very accurate numbers, but we did pretty good work preparing F1 EE for LHe and results speak for themselves









System after

Thanks to everyone who participated and especially AMD powering the event. It was nice to see some old and new friends and everyone did some really nice scores

[[XC] Oj101]

Loving the CPU ID

Very nice work guys

[Oskaliber]

Haha, system after looks cool. Did something die?

[0cool]

Your pics are very nice !

Congratulations for all scores !
It have been very fun for you