Changed my CPU from C to an E
Missileman
Orlando, Florida Icrontian
I just changed from a 3.0 Ghz Northwood to a 3.0 Ghz Prescott. (Did a swap out with my buddy for 20 bucks). Here's the differences that I have seen:
Idle temp went up from 38C to 46C.
Full load temp went up from 48C to 59C.
My UPS displays current draw for system. It went from 308 watts at idle to 379 Watts. Current is at 384 watss now running 2 Prime95's.
I lapped my aero4 heatsink while I had it off and used AS5 instead of AS3 when I reinstalled. Temps may drop a bit after the "cure" process runs it's course.
Not scientific data, but may prove useful to somebody.
Idle temp went up from 38C to 46C.
Full load temp went up from 48C to 59C.
My UPS displays current draw for system. It went from 308 watts at idle to 379 Watts. Current is at 384 watss now running 2 Prime95's.
I lapped my aero4 heatsink while I had it off and used AS5 instead of AS3 when I reinstalled. Temps may drop a bit after the "cure" process runs it's course.
Not scientific data, but may prove useful to somebody.
0
Comments
Won't a 3.0 prescott perform below a 3.0 C due to the longer pipeline? I didn't read up that much on the performance charts.
His motherboard won't support prescott so He swapped out with mine and I just had to pay the difference to him. My C was only 6 months or so old. He just picked the wrong 3.0 at Newegg and rather than return it we did the exchange.
It seems a little snappier in use. Course it's only been running 3 hours. Some benches so far are a little lower while others are higher. I'll know more in a few days. gonna do a fresh install since I'm overdue and then I'll hit the performance hard.
what UPS are you using that displays the amount of watts ur currently using?
TIA
My UPS is an APC Back-ups XS 1000 line conditioning UPS. Catches High, low, and electrical noise as well as surges and black-outs. Only runs about 100 bucks or so at Comp=USA when it is on sale. It has a USB connection to the computer so it sends all kinds of info back and forth. One thing I like is you can tell it to set all alarms off. No way to do that from the control panel
Wouldn't run 2 instances of Prime95 without one of them bombing out.
Went to Intel and check the specs on that batch and it said operating Voltage was 1.25 to 1.525. My board only goes as low as 1.378 so I steped it up from there all the way to 1.52 and it always failed. Tried upping the memory voltage still failed. Went back to system defaults and started stepping down frequency. It runs fine at 2.85, but fails at 3.0. I guess it was a ragged edge which would probably run apps fine. Just not Dual Prime
Went back to my C and reran dual prime just for grins and it runs just fine. Was the processor. Stamped Costa Rica. Maybe the rumors are true. Processors from Costa Rica are trouble. Gonna try one from ZipZoomFly so I know I'll get a different batch. Either that or I'll wait a couple of weeks so NewEgg stock gets changed out.
/me fetches popcorn..
I remember sitting at the Intel press event in January when they revealed the thermal and power requirements for the prescott. A large swarm of "whoah" and shocked expressions
All Voltages are at default. All the readings so far are identical to the last processor. Idles at 48.5, runs dual prime at 59 - 59.5 at most. System is drawing 384 watts while at full load. The fancy ductwork is showing 45C. Sys temp is showing 38 C and that is with side of case off. Dang 9800Pro really puts out a lot of heat to rise up.
Gonna throw a SETI WU in on top of the 2 Primes and let it run the rest of the night. Try to start OC tomorrow a bit. I want to see how much the temp rises when you add some clock cycles.
With regards to SETI, the P4E crunches SETI units somewhat faster than a comperable P4C, mainly due to the added dedicated Multiplication ALU, the additional 512 KB of L2 and additional ALU improvements.
Case and point:
A P4 2.8E OC'd to 3.5 GHz, i875 chipset (Asus P4C800-E) & Single-Channel DDR400 is crunching 2 SETI units every 2 hours and 5 minutes.
A P4 2.8C OC'd to 3.8 GHz, i875 chipset (Asus P4C800-E) & Dual-Channel DDR400 is crunching 2 SETI units every 2 hours and 30 minutes.
Whatever Intel added to the Prescott sure makes SETI happy
SimGuy - Yes the processor is marked from Costa Rica also. It is from a different batch though. It has ran all night with 2 prime95 and SETI running so I would say it is stable. My voltage rails are all stable (5.1, 12.1, 3.30), I am a little concerned about the fall off on the Vcore even though it stays within spec. It drops from 1.38 to 1.30 under full load. I know Abits drop during load with the P4s but this seems a bit much. Spec says 1.25 to 1.525 for an E, but I may turn up the voltage a notch and see what the fall off is with more current available.
Temperatures have stayed about the same. Record shows it hit 60.5 for a bit during the night. I don't really believe the temps because the heatsink is not even warm to the touch. It is actually quite cool. Readings must be caused by the smaller core and having the heatsource in a more compact area. Temps do rise and drop faster than the C models which of course are faster than the other P4 cores. You don't hear much about the Coolermaster Aero line, but it is a good heatsink/fan combo. I was impressed with the Aero 7+ on my Athlon so when I tried the P4s I tried the Aero 4+ and it was just as impressive. I have lapped both of them and got an extra degree or 2, but from what I hear, holding an E model P4 below 60C at full load is an accomplishment in itself. Let alone to do it on an Abit Max3 which is known to read VERY high on it's temps.
I'm gonna start pushing up the FSB speed as soon as I do a backup. i will report the results later.
Maybe that'll get the price down a bit.
http://www.ebuyer.com/customer/products/index.html?action=c2hvd19wcm9kdWN0X292ZXJ2aWV3&product_uid=459
Nothing is cheap over here, and importing isn't an option due to weight. Whoopee for the UK.
Here's what it looks like though:
http://www.overclockers.com/articles952/
I don't think the above needs any further explaining.
But it will keep the house warm when it is winter outside.
Bet my dang power bill goes up another 20 bucks a month now !!!
:smokin:
I just knew you'd flash "your" furnace once again.
Great PhotoChop there.
It got pretty hot in their badly designed reactor. And on top of that they had a mad boss ignoring all warnings and procedures.
Guess we could have a number of PresHot furnace meltdowns on our hands if that furnace is released to the public.
Raising clock speed raises temp a good amount.
As we all know, raising voltage increases temp a LOT.
Raising voltage and frequency quickly put the Prescott core into the "must watercool category".
I got the processor to run stable at 3.7Ghz and voltage of only 1.5 which is still below the 1.525 spec. Problem was that the temp on the cpu bridge was now showing 75C at load and the system current draw was now showing 505 Watts. I did notice that each increment in voltage caused an inordinate increase in amount of power required by the processor.
From reading the Intel Tech papers it's obvious that this is due to the "strained silicon" technology. I understand some of the reasoning to it's use. It appears that you can make this processor run as fast as you want with the main limiting factors being power available and more directly how much heat you can carry away from the cpu.
I have settled back to a mild O/C of FSB 215 @ 1.4V with GAT = turbo. Temp idles at 48.5C and tops out at 60C. Current stays about 385 Watts at full load. 3.2 isn't a big overclock, but it's safe and stable.
Benchmarks seem to be following most of the reviews. Aquamark is about the same. 3DMark03 went down a bit. PCMark04 went up a bit. Actual use in games though I believe has improved. Responsiveness has improved, especially in OpenGL games. Of course this is all just "a feeling".
I'll continue to push it up as it "burns" in a bit over the next week to see if anything changes.
One thing to factor into this is what HS compund you are using and how fast it cures in. Creamique will not reach full effectiveness until it has been exposed to running heat (IE processor running under load) for about 200-250 running hours. Second, my 2.8 GHz EE is running at about 3.0GHz in a room ambient that averages 80 degrees F and case is about 31-32 C most of time. CPU is running stable at about 65-66 C. Funny thing about this processor, is it is heat tolerant more than AMD chips are of those I have used-- it can and does stay stable at temps that would kill an Athlon, plus a bunch more temp than that.
Florida, at least my part, gives me a natural hot environment lab to test CPUs in. So, when I post here with what for others would be gawdawful high temps, take the environment into account. I am air cooling my computers-- they both get used heavily. And they STILL produced 391 points yesterday... Both are OC'd P4s, one is a Northwood OC'd to 3.2 GHz. The Prescott is a 2.8 and NOW is only at a bit over 3 GHz, as the HS compund should now be about cured in. Prescott runs two folding instances, and one console is at .965 efficiency and the other is at .973 effciency. Prescott is doing about 60% of my total output on an XP console pair of clients, as the Northwood is OC'd a bit higher and is running on Linux (not an older Linux, a current version), which has less overhead in terms of CPU time slices for the O\S itself than XP and actually wants to give the client more time than it is using.
I can run CorelDraw 12 at same time as I fold with both clients running, and most of my DTP uses vector text with gradient fills or full color pattern fills. So that also puts a load on XP. If I want to run full business accounting and DTP at same time, I drop to one client runnig on prescott, and get the work done fast enough that effectiveness\efficency does nto drop much at all, try less than one percentage point for same effectiveness on P4 WU for two hours of time of one client suspended.
The clients float back and forth as to effectiveness\efficency rating, versus each other, depending on exact WU running on each console instance last. This tells me that both clients are running well, if one were consistently less effective I would start looking much deeper into things.
When I have low production days I know what is happening, from the verbose logs. It is not the CPU, that is for sure.
One thing I will say, try dropping CPU voltage to 1.35 to 1.37 VDC or so if you can, my IC7-MAx3 motherboard allows this. Actually, effective voltage to CPU is about 1.33 VDC right now. 1.4 seems HIGH, possibly a tib of overkill.
Unless the E you are talking about is a different series than my Prescott CPU (it looks from voltages as if it is a Northwood big L2 cache version like mine-- Linux box), which is rated at 1.525 MAX voltage also, I think the actual needed voltage can be stepped in hundredths of a volt or thousandths to tune, not tenths. It is eating total wattage simply to run things at very high speed, that wattage is what is piped through the CPU mostly, and at 3GHz it is running billions of on-offs per switch per second, all of which have to be ground plane drained out of motherboard very fast. To do this my motherboard here is inverse polarity loading the motherboard ground planes, at least some of them. You might have PULLDOWN inverse polarity voltages present in some CPU circuitry also, would not surprise me AT ALL. And wattage charge gens heat Kilocalories, no matter polarity.
I don't have load temps, because I disconnected the "F" part on the HSF (:D), and the temperature did rise VERY quickly, up to about 38*C in maybe 30s. I then plugged the fan back in, but the BIOS froze for some reason. And now the board won't boot with the 2.8E. It also won't boot with the 2.4A out of my laptop. Sooo I have to RMA it before I can do any more testing.