QX9650 Heatload

[jinu117]

Well, Denny came to pick up his single stage, (or is it Andres?)
I requested him to bring QX9650 with him so we can do some quick load testing session He has another CPU he can test out once he gets back at home but I am not sure if he will have time to do so to match and figure out what the trend of load is.
Basically the unit is 210-220w tuned unit (it can go to 230w around 19c-20c ambient but I typically give myself a little room due to different ambient condition possible with units I make). Since last post regarding 2 evaporator comparison under Q6600 G0, I've changed the unit so it won't be reading as cold of temperature under less load
Basically skipping all steps, we went to meat of it... started around 1.6v and than decided to kick i up to max volt under Asus P5K-E since it wasn't stressing cooler much at all. Bios setup 1.7v reading from CPU-Z around 1.664v.
I've recorded the temperature of evap under load tester last night in preparation to know exactly what load I am subjecting it to (well I think it was about 1-2c cooler in the place today) and here it is.


Basiaclly, based on what I wrote down last night and comparing, this is only about 175w+/-10w!!! Unlike the Q6600 G0 which at same voltage and about 400mhz less dumping about 220w+/-10w.
I was actually hoping to see closer to 5ghz number but that wasn't the case... -__- (45nm and being unlocked multi chip and all...) This is pretty damned good news as I can see quad retail product probably going another 200-400mhz on default speed as product matures and overclock might stop around 200w-210w range which most of my product will cover quite easily (other than really early one tuned optimal around 150w and up to 180w..... even this might have some fun with reasonable overclock :P)
keep in mind, that voltage alone is not what I would use for daily to begin with and also makes me wonder why there was no major default v-core drop on this 45nm transistion from 65nm... This makes me wonder if Intel is having not so good yield rate quite up yet and artificially inflating default voltage to compensate for it for time being.
Anyways, it seems to be quite good news to me as cost of tuning for 250w was getting too much (component cost, tuning involvements, increase in noise level, general loss on lower wattage temperature, etc). I guess I am back down to 220-230w tuning from this point on which will work quite well with current quad g0 as well as some room for still unreleased 45nm quad
Of course, R&D won't stop for more load as I should prepare my self for 8 core as well...

[n00b 0f l337]

Good seems all the guesses were pretty close, but only 175+-10? Thats great news indeed.

[kayl]

nice work jinu117.
The little time i have had to play with QX9650 it seems to be almost as hot as the old dual cores.
No matter what volts you give these cpu they seem to do around 4.8ghz on a single stage.
If you use dry ice to cool these cpus you can almost forget about the system while benching it uses that little dice unlike the QX6700's who eat the stuff .

jinu117 did you have trouble booting into windows over 400FSB.
Once in windows the fsb tools can take you close to 500fsb but these 45ns cpus just don’t like to boot high FSB from what I have seen.

[eva2000]

nice sounds about right for my Kayl Black FrozenSS 1/4HP unit still

[PhilippF]

Very nice to hear the power is going down a little instead of rising steadily over time. My only great fear for the future is that also Intel processors get a coldbug like AMD.

[tim-]

I told you it was about 170w glad to se someone could confirm my guess.

nice clock too!

regards
Tim

[boshuter]

Glad to see these results I've been telling everyone for months that it's not worth tuning for the current quads @ 250-300w because they would be obsolete as soon as the new cpu's arrived.

The current quads are just not good cpu's, they output way to much heat for the limited gain. I see them as the "Preslers" of a year or so ago, the presler was the same way, too hot with too little gain. I've been telling all my friends to stay as far away from the quads as they can, you won't be able to give one away when the 45nm cpu's arrive I had a half dozen preslers (only because they were free es cpu's) and there was no way a single stage was going to cool one that was oc'd. I got a 950 to 5.6ghz on a 1/4hp single stage, but temps were climbing fast and I was lucky to get a 1M pi run in before it had to be shut down, the same presler running at 6.1ghz and 1.85v was eating dry ice in a solid copper mousepot so fast I couldn't keep the tube filled

[AndreYang]

Great work. I will enjoy this single stage.

[Fhqwhgads6680]

That is really good to hear! Although the first thing I thought was.... 8 cores = 350w..... That is going to be some fun tuning....

[TopherTony]

So in recent history, what would you assume is the hottest chip overall? B3 Qx6700? If these are dropping the heatload again, then any unit that can handle the heat from that chip shouldnt falter from these

[tim-]

tophertony:
yes the qx6700 is the hottest chip so far.

regards
Tim

[xgman]

Jin just revamped my Vapochill LS unit and did a super job. Only took a week to turn it around and get it back across the country. On a P5K dlx setup I am getting 4800 mhz on a QX9650 at 1.55v (I'd say 95% stable) It's hard to get an accurate temp reading since nothing really reads negative temps on this short of a probe. It posted into windows above 4900 mhz on a p5E3, but I had a bad mount and the board ended up dieing, so I'm not sure about that one. Anyway, Jin did a great job on my LS mod!!

[xgman]

nice work jinu117.
jinu117 did you have trouble booting into windows over 400FSB.
Once in windows the fsb tools can take you close to 500fsb but these 45ns cpus just don’t like to boot high FSB from what I have seen.

Interesting you should mention this. I had a much harder time with this than on the QX6850. I am able to boot well over 400, but not in the 480 range I was on the 6850. I think right now I have it set at 436 with a divider that runs the mem at 515. I'm really not sure what the optimum vcore is yet for this. I think I need at least around 1.5v, but I'm not sure if more will get me any higher really. What do you think is relatively safe for these on single phase for 24.7 use? I had been running my 6850 at 1.61, but I assumed that the 45's needed to be proprtionately less?

[Pete]

The OC style of all 45NM is down to BIOS right now, i've been kicking mine all day as it's picky as hell.

So far 490FSB into Windows with full range of 6 - 10 on water so far, need to get ot colder for higher multi

[jinu117]

Uhmmm not sure We didn't have problem booting in with 400fsb but haven't tried beyond that :P Unlocked multi :P

[sacha35]

Nice work and post jinu, very informative and only 175watts to 185 watts with this vcore and clock, was it possible to go higher with the chip changing the multy or was this the max clock that it would do on the S/S unit?

[jinu117]

Sacha, I would think that was max stable oc for that chip unless higher voltage was applied or colder temperature was subjected. It seems 5ghz quad stable chip under SS will take VERY good CPU indeed to reach. (Maybe 9770) - there could have been a little more as originally we tried 4800mhz quad stable and backed down a little (5fsb?) just to get 100% stable and never bothered going back up to see where in between would it have been stable. (it was getting late)
We haven't heard much on QX9650 to know how general trend of the chip is yet so all is subjected to some ??? at this point

[HoriWanderFuLL]

nice job Jinu long suction line is good idea..free up bench space

[jinu117]

Guess we are getting a little off topic. Suction line length is really decided up on customers In general though, longer it is, better it is for few reasons as long as it doesn't become cumbersome

[kitfit1]

That, is really good news jinu117. Still on Dual core at the moment because of the heatload issuse with Quads. But if the new ones are only pushing up to 200w, then my present SS should cope with that nicely.