Prescott in the bat cave computer

[batboy]

Just got my new OEM 2.8 Prescott. It's running at default speed right now making sure it works ok. Here is the info printed on the chip (note that it's a 2003 week 50):

2.80GHZ/1M/800
SL79K Philippines
7350A620

I will confirm that it does run warmer than the Northwoods. I have it installed in my Swiftech watercooled case. I have modded cooling on the northbridge of this IC7 mobo. Case cooling consists of two 120mm intake fans and two 120mm exhaust/radiator fans.

Idle temps of the Prescott at default speed are very similar to both my 2.4C @ 3.4 gig and 2.8C @ 3.4 gig. Here are the load temps after 10 runs of Sandra CPU burn-in (which is really only moderately loading the CPU).

Prescott 2.8 load temps at default speed:
CPU=49.5, system=19, PWM=35.5

Northwood 2.8 @ 3.4 gig load temps:
CPU=43.5, system=19, PWM=32

Northwood 2.4 M0 @ 3.4 gig load temps:
CPU=40.5, system=19, PWM=30

Ok, time to O/C... be back in a couple minutes.

 

[TC]

Mine is on the FedEx truck - hope to add my observations to the collective later today or tomorrow.

 

[batboy]

Ok. here is load temps for the Prescott 2.8 overclocked to 3.4 gig.

CPU=52, system=19, PWM=38.5

Vcore is still at default of 1.3875v listed in the BIOS and measured at 1.36v in HW Doc.

It wasn't stable at 3.6 gig with default vcore... but I just remembered I still had the PC4000 RAM in using the 1:1 ratio (I couldn't run those setting with the Northwood 2.8C either). So I'll go back to 3.4 gig and run all the benchmarks before swapping in the Mushkin PC3500 level 2 RAM at the 5:4 ratio.

 

[Edward2]

Wow, 9C warmer than the [email protected] just running at default with water cooling. And 12C warmer running at 3.4 GHz. Good luck and keep us updated.

 

[Ross]

2.80GHZ/1M/800
SL79K Philippines
7350A620

Mine is an '03 week 50 also....sorry, I forgot to mention it previously. Good luck batboy!

 

[batboy]

Everyone knows those 2.4C M0 steppings run cooler than other Northwoods. It's probably better and more fair to compare the Northwood 2.8C temps to the Prescott 2.8. So, with both 2.8's running at 3.4 gig... then the Prescott is 8.5 degrees warmer load CPU temp. But look at the PWM temps... very interesting. That sensor is near the power mosfets if you didn't know. The Prescott obviously sucks up lots of power

 

[batboy]

COMPARISON OF 2.4C, 2.8C and 2.8E OVERCLOCKED TO 3.4 GIG

Sandra CPU ALU:
2.4C/3.4=10722, 2.8C/3.4=10363, 2.8E/3.4=9998

Sandra CPU FPU:
2.4C/3.4=2941/6413, 2.8C/3.4=2947/6469, 2.8E/3.4=2783/5864

Sandra memory bandwidth:
2.4C/3.4=6166/6160, 2.8C/3.4=5876/5791, 2.8E/3.4=5919/5919

PCmark2002 CPU
2.4C/3.4=8436, 2.8C/3.4=8437, 2.8E/3.4=7190

PCmark2002 memory
2.4C/3.4=11450, 2.8C/3.4=10955, 2.8E/3.4=12549

PCmark2002 HDD
2.4C/3.4=1379, 2.8C/3.4=1299, 2.8E/3.4=1279

PCmark2004
2.4C/3.4=5323, 2.8C/3.4=5259, 2.8E/3.4=5222

3Dmark2001 (Radeon 9700 Pro at default speed and settings)
2.4C/3.4=17056, 2.8C/3.4=17138, 2.8E/3.4=17129

 

[batboy]

I'ts interesting to look at the 2.4C/3.4 vs. 2.8C/3.4 to see what difference FSB makes. But, we probably should look closer at the 2.8C Northwood @ 3.4 gig vs. the 2.8E Prescott @ 3.4 gig if we want to see a true the comparison between the two cores, because the 2.4C had different RAM (Mushkin PC3500) and different memory ratio (5:4) then the two 2.8's (Corsair PC4000 Pro at 1:1).

1) The Northwoods beat the Prescott in the Sandra CPU tests.

2) The Prescott's cache allowed it to squeak ahead of the 2.8C in Sandra memory bandwidth. The higher FSB and tighter RAM timings of the 2.4C shined through for it.

3) The Prescott did poorly in the PCmark2002 CPU test, but kicked butt in the memory test.

4) The Prescott and the 2.8C basically have another tie in PCmark2004. Once again, the 2.4C probably does better due to higher FSB and tighter RAM timings.

5) The 3Dmark2001 was essentially a draw between the Northwood 2.8C and Prescott 2.8E (both of which edged out the 2.4C.

 

[micamica1217]

COMPARISON OF 2.4C, 2.8C and 2.8E OVERCLOCKED TO 3.4 GIG

thanks for all the benchies...

after reviewing your numbers, and looking at my own, I'm rethinking about going with the prescott.
I may try a 2.8c just for the heck of it, as most hit 3.5ghz and some reach 3.6ghz.

I see no reason at this time to even try the prescott, not for the heat the CPU does, but because of the heat it causes the mobo to reach...you are not the only one to site higher mobo/PWM temps.
(plus, I'm getting faster results, in PCMARK and 3DMARK with my 2.4b, as compared to the prescott at 3.4ghz)
(the only thing the prescott beats me in is sandra mem and pcmark mem tests....gota love my 2.4b )

thanks again batboy, your the best.

mica

 

[tom10167]

If I were you I'd hold out until the 3.2 release, there's no real rush, and by then the chips may have matured some.

 

[TC]

Just got mine - will try to get it hooked up to the koolance unit in a couple of hours.

 

[batboy]

I'm running at 3.8 gig right now. Really had to start bumping up the voltage above 3.5 gig even though it ran perfect at default vcore when I was benching 3.4 gig. I had to use 1.6v to benchmark at 3.8 gig and 3Dmark2001 set off my audible temp alarm that I had set for 60 degrees, so I aborted the test.

This tells me the yield for this core stepping is less than we hoped. For Intel to release a speed above 3.4 gig, they will need significant improvements. I suspect the rumors about Intel speed binning the 2.8E is probably also correct. My advice is to wait for the next stepping and get one of the higher speed Prescotts if you are going for 4 gig.

 

[Edward2]

Very interesting results Batboy, and thank you for taking the time to run all of those tests as a comparison. I know that I appreciate your effort in doing this.

 

[batboy]

Oh yeah, here are the Prescott benchies for 3.8 gig.

Sandra CPU ALU: 11193

Sandra CPU FPU: 3121/6557

Sandra memory bandwidth: 6062/6044

PCmark2002 CPU: 8041

PCmark2002 memory: 13413

PCmark2002 HDD: 1329

The Prescott is starting to come alive more and more as the speed scales up. Too bad I can't get this bad boy completely stable at 3.8 gig (at a reasonable vcore and temp).

 

[Bald Shadow]

My advice is to wait for the next stepping and get one of the higher speed Prescotts if you are going for 4 gig.

That is exactly what I am doing. By the way thank you for all of the test result, very imformative comparison.

 

[Vio1]

it seems that from the PCmark2002 benchies that a 3.8E is beaten by a 3.4C (8041 vs 8437 CPU MARK)

It looks a lot like how the celeron can get high clocks, but is still beaten by a lower clocked p4.

 

[Big Nuttz]

Hopefully I'll receive my 2.8e tomorrow. I'm taking my MachI in on Friday to regass it with R507 so that should be enough to keep the temps down low enough for a decent stable overclock. I'll post tomorrow or Friay my results also. I'm curious to see how it handle Seti WU's with that 1MB of cache.

 

[hUMANbEATbOX]

the only problem i see is the heat. the scores look fine, provided intel can tame the heat and get some new steppings out.

i'd like to see that 3dmark score at 3.8, open your window batboy! put your whole rig in the fridge! i don't care!

thanks for the benches.

 

[TC]

Well I tried to fire up my new AI7 with the prescott and all it does is immediately start blaring an alarm sound and the post code indicator stays on AF. I hope I don't need a bios update just to get it to post because I'll have to buy another chip and wait on it. Anybody heard anything on the AI7 not supporting the prescott?

 

[#18]

Just like I thought.

PCmark, Sandra, each show a single fact.

Prescott can do less work per MHZ than Northwood, which has show in PCmark02 CPU score. The extra cache really improved memory performance, show in PCmark02 memory score.

Sandra CPU test is rather testing the total processing power than pure CPU speed. Memory bandwidth is also another one of the many facts which is really depended on FSB.

Only 3dmark2k1SE here is a benchmark with mixed facts.

The resault show Prescott isn't that bad.

I am sure at 3.8GHZ Prescott will outperform a 3.4GHZ Northwood by a little bit. Not bad at all. At least Intel is making progress.