Whats your 24/7 Core i7 920 and temps under water?

[T_Flight]

I'm not...IMC, QPI, L3. That's what I'm trying to isolate and see. The HT test was a good one, becasue that is a piece of evidence that is confirming my suspicions, but hasn't quite confirmed it completely yet. The temps are still high though, and I really don't like seeing 70's for 24 hour use. I'm talking a machine folding 24/7 at near 100% load on all cores. To do that at relatively high OC's I need to get the Temps down and to do that would help a whole lot of people here. Maybe these things might handle 70's for years, btu I don;t wanna risk it with a CPU that costs 1000 dollars. My stuff has to last. I'll clock the heck out of it, but it's gonna run cool if I can make it run cool.

We might even have to wait for Block redesigns. I really hope that's not the caseIt's really gonna be bummer to play the waiting game after waiting so long again. Even before I was a member here I was waiting on this CPU. It's been almost 2 years now since the time I knew that these were gonna become a reality and were not just a rumor. I'm more determined than ever to figure them out, and will do whatever it takes. It's gonna have to be water though. Phase is out of the question.

[T_Flight]

OK did some tests. I can't post screens very easy because my system has no security yet, and is not hooked up to my router yet either, but these things do run cooler with HT disabled and with speedstep enabled. That's just on stock clocks. Since I just got my Video Card RMA back today I haven't been able to do any OC'ing, but I do see a measureable difference using CoreTemp at load and my idles are slightly lower also. Idles are already very cool in the 35-38 range on stock air with it running at stock speeds. It's the vCore with an OC that'll bring those up. I suspect it may have to do with the IMC and QPI being onboard which is something we haven't seen until now.

I agree with Riptide though that Water is better than Stock air, but I still would like to keep these temps down and maintain HT because of the work I'll be doing with my machine. SpeedStep I'm gonna leave enabled, because it looks like it may perform better with SpeedStep than without it.

[X2beach]

I did some more test. the results speak for themself

This is with 68F Room Temp.. inside case air intake 70F

[MrBean]

water temp don't differ too much. its air in air out that's what important.

I disagree. The delta water temp will tell you how efficiently you are removing heat from the water - assuming it goes into the water in the first place. I am talking purely from a primary cooling perspective, not considering the secondary effect airflow across the motherboard will have.

What a few guys here are inferring is that a (decent) 240x120 rad is overcome by a single i920 - and what I am saying is, measure ambient water temp, start your system, run it at full load for a few hours, and measure differential in water in/out.

If it's fairly close, within couple degrees C, and close to ambient, it means:

i) CPU is not dumping as much heat as we thought into the water, either because it's not that hot, or, because we can't transfer the "real" heat away from the cpu.

ii) If there's a big delta between in/out, it means cpu block is doing a good job at removing heat from the processor, and the rad is efficient, and within it's design parameters.

[rge]

20C Difference in load between E8600 12 hrs prime vs i7 940 12 hrs prime, using exact same cooling and setup, only motherboard, ram, cpu changed.

cpu only loop is PA120.3, 3 yates at 1300 rpms, ddc 3.1, dtek fuzion v1 with 5.5 nozzle. (I made a bracket to use for 1366 for dtek).

E8600, 1.33 vcore 4.2ghz max load 54C, have several runs, none were ever over 55C all at 26C ambients (pic1)

Core i7 940, 1.31 vcore 4/4.2 turbo max load 74C, have several runs max was 75C, all at 26C ambients (pic2)

I now have a GTZ installed with 1366 bracket and backplate, my load temps are ~3C cooler at 71C on prime 12 hrs at same ambient...so it was not my homemade bracket causing problem...cpu is just hot.

[zlojack]

Not surprising that a quad core with the MCH on-die is hotter than a dual core.

[doox00]

Just got all my water cooling stuff installed and running. So far at 3.8ghz 1.33 vcore I am hitting 74C with HT on, room temp 68F. I tried 4ghz but kept locking up.. gunna have to play with voltages some to get the oc up. Each core temp 74, 70, 71, 67 using real temp 2.89.5

This is with mc-tdx cpu block, 2 mcw60 gpu blocks, mrc320-QP rad with 3 s-flex 1600rpm fans and a ddc 3.2 pump.

[Vapor]

I hope no one is expecting a solid improvement in clocks over air....

Just look at benchmark stables speeds--people are hitting 4.4-4.5GHz on air and the same chip is barely cracking 5GHz (if that) at -110C. Heck, FUGGER got 5GHz on air and 5.7 on cascade with a chip DrWho? brought for him to play with.

Taking 15-20C off will maybe get 50-100MHz for most people (if that).

EDIT: long story short, these chips were designed to be indifferent to temperature. I recommend closing Core Temp/RealTemp, tuning your OC to max clocks that are stable, then tuning your cooling to reduce noise as much as possible without sacrificing stability--that's the way to make the most of a WC system on i7 (reducing noise).

[dinos22]

I hope no one is expecting a solid improvement in clocks over air....

Just look at benchmark stables speeds--people are hitting 4.4-4.5GHz on air and the same chip is barely cracking 5GHz (if that) at -110C. Heck, FUGGER got 5GHz on air and 5.7 on cascade with a chip DrWho? brought for him to play with.

Taking 15-20C off will maybe get 50-100MHz for most people (if that).

EDIT: long story short, these chips were designed to be indifferent to temperature. I recommend closing Core Temp/RealTemp, tuning your OC to max clocks that are stable, then tuning your cooling to reduce noise as much as possible--that's the way to make the most of a WC system on i7 (reducing noise).

we dont know what the issues are with current CPUs

maybe they are all heavily cold bugged man you just dont know

i think ppl will get great gains from water first of all because deltas will be far lower than air cooling so while you could prime at close to water temps on air it would not be sustainable due to higher temps

LOL Eric how can you say that these CPUs are indifferent to temperature lol

[DarthBeavis]

Whats your 24/7 Core i7 920 and temps under water?

You will eventually get shorts if you run your rig under water . . .

[Freaky Freezer]

Wait, I just looked on Intel's site, what the hell is the Intel rated Max temperature for these chips to begin with?

[doox00]

Wait, I just looked on Intel's site, what the hell is the Intel rated Max temperature for these chips to begin with?

I was not able to find out that info last time I looked.

[Vapor]

Also a good point....65C is still a huge 35C away from assumed TjMax of 100C.

Anyway, undervolt + underclock and observe idle temps, if they're close to ambient, then it's not far off I suppose

[T_Flight]

I did some more test. the results speak for themself

This is with 68F Room Temp.. inside case air intake 70F
[/IMG]

i920 at 4GHz and 1.36vCore with 60C AVG is a sweetheart chip. I could definitely live with that. That pretty nice with a 100% Prime load.

[twwen2]

You gotta remember that the 920 pulls almost 200watts at load, stock speeds. The 965 (3.2GHz stock) pulls ~250w. Add another 800MHz and you guys may well be using around 270watts under load; that's a heck of a lot more than penryn/wolfdale.

As far as power consumption goes we're back in the Q6x00 days, in the 2xx watt region. It was common back then to see 50-60 degC under load.

[jasonelmore]

Wow i did'nt realize my i7 920 at 4ghz was pulling that much power.. that leaves hardly nothing left for my GTX 280.. There's no way i could add a second 280 on a Corsair TX-750W is there?

[Vapor]

You gotta remember that the 920 pulls almost 200watts at load, stock speeds. The 965 (3.2GHz stock) pulls ~250w.

WTF? Source....multiple please.

That's insanely ridiculous for a chip conservatively rated at 130W TDP.

The design of this chip is meant to reduce power consumption--that was the whole idea...1% power increase for minimum 2% performance required vs. Yorkfield--and there's not that much performance increase.

I don't see why people are putting so much emphasis on temperatures especially compared to Yorkfield/Wolfdale temps--maybe Intel put the DTS sensor in a warmer part of the chip this time, maybe the sensor is actually correctly functioning this time, maybe the process just runs hotter (but lower power--YES that is possible).

Only compare Bloomfield temps to other Bloomfield temps--forget everything you've gotten accustomed to, this is a new chip.

[flopper]

40-45c idle, 60-65c load 4.2ghz.
1.344vcore under load.
made 7hours prime 4 threads.

[T_Flight]

WTF? Source....multiple please.

That's insanely ridiculous for a chip conservatively rated at 130W TDP.

The design of this chip is meant to reduce power consumption--that was the whole idea...1% power increase for minimum 2% performance required vs. Yorkfield--and there's not that much performance increase.

I don't see why people are putting so much emphasis on temperatures especially compared to Yorkfield/Wolfdale temps--maybe Intel put the DTS sensor in a warmer part of the chip this time, maybe the sensor is actually correctly functioning this time, maybe the process just runs hotter (but lower power--YES that is possible).

Only compare Bloomfield temps to other Bloomfield temps--forget everything you've gotten accustomed to, this is a new chip.

Heat and voltage degrades electronics over time. How much? We don't know and that's the point. I don't wanna be the guinea pig who finds out the hard way. I'd rather just keep the CPU as cool as possible, or cooler by getting new stuff. Call it insurance I guess.

Until Intel will come forth and say "This CPU will run at these clocks for this many years" all we can do is cool the coolest and hope for the best. I've traditionally stayed under 60, and that's what I've had good luck with but higher would be getting into new territory. I'd be willing to test it over time if I had Intel backing, but I'm not gonna gamble with a chip that cost a grand on my own. It's just too expensive. I don't think one can hurt one of these CPU's by running it cooler, and that's what I'm trying to do, but running one hotter sure will. The cooler the better for me.

[beast200]

Rge

hi do you think its was / is worth changing from fuzion v1 to swiftech gtz
i am using my fuzion v1 with temp plate and waiting for D-tek to release a mounting system for the 1366
but still like the look of the gtz but is it worth £60 for that


cheers

[jasonelmore]

Heat and voltage degrades electronics over time. How much? We don't know and that's the point. I don't wanna be the guinea pig who finds out the hard way. I'd rather just keep the CPU as cool as possible, or cooler by getting new stuff. Call it insurance I guess.

Until Intel will come forth and say "This CPU will run at these clocks for this many years" all we can do is cool the coolest and hope for the best. I've traditionally stayed under 60, and that's what I've had good luck with but higher would be getting into new territory. I'd be willing to test it over time if I had Intel backing, but I'm not gonna gamble with a chip that cost a grand on my own. It's just too expensive. I don't think one can hurt one of these CPU's by running it cooler, and that's what I'm trying to do, but running one hotter sure will. The cooler the better for me.

You just need to get you a cheap 920 to play around with. I'm at 4ghz on mine and loving it. idling around 40C and loading at 73C. my voltage is at 1.420 and i'm not worried a bit about it frying because it's only $300. Go over to the thread saaya started about the VTT Graveyard. People are posting the chips capabilities right now. 95-100C is the new norm of what not to go over

[rge]

Wait, I just looked on Intel's site, what the hell is the Intel rated Max temperature for these chips to begin with?

If looking for Tcasemax, it is 67.9C at 130W...not that there is a tcase sensor. Intel states in many docs that cpus are individually calibrated so that tcase max is reached when tjmax (100C on i7 read directly from cpu msr) is reached, to avoid unnecessary throttling below tcasemax. (Though that relationship presumes a full load of 130W, intel cooling, stock settings, whatever loading program intel uses, ambients same as testing otherwise that gradient relationship is highly variable.)
http://download.intel.com/design/pro...hts/320834.pdf page 79

Rge

hi do you think its was / is worth changing from fuzion v1 to swiftech gtz
i am using my fuzion v1 with temp plate and waiting for D-tek to release a mounting system for the 1366
but still like the look of the gtz but is it worth £60 for that


cheers

It was worth it for me as my homemade bracket pretty much sucked, and required the block to be slightly rotated to fit. I was going to buy the first EK supreme, GTZ, or DTek that was sold in US with a 1366 bracket...happened to be GTZ first out the door. But I have run several prime runs with Dtek fuzion v1 5.5nozzle with my homemade 1366 bracket and with GTZ with proper bracket and backplate, there was only 2-3C difference on load. On the other hand I do a lot of temp testing with realtemp, and reseating wb frequently, and the retention bracket is awesome on GTZ, the time and aggravation that will save me alone was worth it for me, irrespective of the temps. But based on temps alone, I would not expect much more than 2-3C (I think that is what many reviewers found as max difference anyways), and I dont know how much of that on mine was from wb design or the fact that my homemade bracket was not tight enough.

[T_Flight]

You just need to get you a cheap 920 to play around with. I'm at 4ghz on mine and loving it. idling around 40C and loading at 73C. my voltage is at 1.420 and i'm not worried a bit about it frying because it's only $300. Go over to the thread saaya started about the VTT Graveyard. People are posting the chips capabilities right now. 95-100C is the new norm of what not to go over

Right now, I don't wanna spend anymore on hardware unless I have to. I've gotta wait until after the Christmas Holiday season before I get my watercooling stuff. I'm not afraid to OC it or anything like that. Just wanna be real careful. I don't want my first Extreme Edition to go bye bye. I like the unlocked multi's and the extra options for memory tweaking with the Extreme. It also has the 6400 QPI. Since it's an Extreme I figured it would be pretty stout too since these have to pass QC's to become EE's.

I was testing my vid card lastnight and man I got one that has ALOT in it. I found the max clocks but it needs water. The temps run off. I backed off alot for best temp to clock ratio, and it runs happy now. That's kinda where I'm headed. Getting the best clock to temp ratio without getting too wild.

I did see Saaya's thread, and have been watching that memory thread verrrrry carefully. I have not went over 1.65v and may never go over it. There is so much bandwidth with this platform I don't think I'd ever be able to use it all for years.

Maybe a GTZ or a Dtek v2 or a Supreme may be in order. I've gotta call up Petra's today about some VGA cooling. I think I'm gonna go ahead and get that so I'll have it since it's not very expensive. The rest I'll get after Christmas and when the CC recovers from the damage I'm about to inflict on it. 3 weeks to go...Ho Ho Ho.

[[XC] riptide]

WTF? Source....multiple please.
That's insanely ridiculous for a chip conservatively rated at 130W TDP.

The design of this chip is meant to reduce power consumption--that was the whole idea...1% power increase for minimum 2% performance required vs. Yorkfield--and there's not that much performance increase.

I don't see why people are putting so much emphasis on temperatures especially compared to Yorkfield/Wolfdale temps--maybe Intel put the DTS sensor in a warmer part of the chip this time, maybe the sensor is actually correctly functioning this time, maybe the process just runs hotter (but lower power--YES that is possible).

Only compare Bloomfield temps to other Bloomfield temps--forget everything you've gotten accustomed to, this is a new chip.

There was some short review done over on [T]ard forums by Bennett that showed ridiculous power draw figures. Frankly I ignored it.

[Freaky Freezer]

If looking for Tcasemax, it is 67.9C at 130W...not that there is a tcase sensor. Intel states in many docs that cpus are individually calibrated so that tcase max is reached when tjmax (100C on i7 read directly from cpu msr) is reached, to avoid unnecessary throttling below tcasemax. (Though that relationship presumes a full load of 130W, intel cooling, stock settings, whatever loading program intel uses, ambients same as testing otherwise that gradient relationship is highly variable.)
http://download.intel.com/design/pro...hts/320834.pdf page 79

So basically max temp is 100c, say 90-95c just to be safe? Well then running at 70-75c @ full load is pretty damn good I would presume, no?