prescott 3,4ghz benches

[LTC8K6]

The P4 has always had an on die thermal diode. Why are folks here, who should know this, trying to say it doesn't? If you don't even know that much about the P4, then why are you so down on it?

The P3 had an on die thermal diode, too. Sheesh!

 

[THUGSROOK]

Originally posted by: LTC8K6
The P4 has always had an on die thermal diode. Why are folks here, who should know this, trying to say it doesn't? If you don't even know that much about the P4, then why are you so down on it?

The P3 had an on die thermal diode, too. Sheesh!

you my friend dont know squat about P4.

i on the other hand (along with Duvie and a few others still on this forum) literally pioneered P4 OCing. dont believe me? ask around buddy!

quoting Toms hardware? Tom is a schmuck ~ you are quoting a schmuck.
why dont you sit back and learn something from the forum elders before we put a dunce cap on your head, ok?

 

[LTC8K6]

I'm not your friend. I know that Intel chips have had on die temp sensors for a long time now.

I am the only one who actually bothered to ask the OP if his P4 was throttling. I suppose I could have jumped to the conclusion that it was throttling and not bothered to ask....

I referred to Tom's because he had an article earlier about how the P4 3.6 could throttle at the drop of a hat. Lots of folks ridiculed the Prescott over that article. Later, during the stress test with the same Prescott, Tom's can't seem to get it to throttle at all.

Same thing with the Sudian article. A lot of ridiculing, but very little attempt at verification by any reviewers. Lots of reviews have occurred of the high end Prescott P4's haven't they? Are the reviewers just not noticing the throttling?

I have already said, in far earlier threads on the subject, that if your Prescott throttles with the stock cooler and a reasonably well vented case, you should complain to Intel and ask for your money back. I remain of that opinion.

I still think that if that's all you lose to Intel's throttling, then it's amazing.

 

[THUGSROOK]

Originally posted by: LTC8K6
Are the reviewers just not noticing the throttling?

you will learn very quickly that most reviewers are morons, plain and simple.

 

[LTC8K6]

you will learn very quickly that most reviewers are morons, plain and simple.

I can't agree with you there. Some are, certainly.

 

[Zebo]

I referred to Tom's because he had an article earlier about how the P4 3.6 could throttle at the drop of a hat. Lots of folks ridiculed the Prescott over that article. Later, during the stress test with the same Prescott, Tom's can't seem to get it to throttle at all

And you're still refering to toms after that obviuos inconsistancy?

I get it, if his data agree with your presshot love then it's valid, if it don't he must be mistaken. Tell me I'm wrong, you've gone to great lengths to explain (one liner wise) there is no problem with Prescott. OK prove it, prove prescott is'nt/does'nt idle under long term stress at varied freq. It's not that hard, but it does take time and scientific method (which toms does'nt employ) but you should be done in a day with testing. You tell me what chip you have and I'll even develop proper method for you to rule out mem, vcard etc.

 

[Zebo]

Tstep, a64 does'nt idle period, unless you request it to visa vi C&Q which is off in most of our systems. You will go into thermal protection ~110C or crash before CPU idles. Not only that, they (winch's) use 35-55W underload vs. 110W-150W for you know whom. Trust me, things will get really embarressing for chipzilla when competing dualcores are compared (if these review sites actually house them in a case)

 

[LTC8K6]

I never said anything about Prescott not having any problem.

There is no consistency to the claims of Prescott throttling at the drop of a hat. That's the point.

The idea that they throttle at idle is a new one to me. If that's what you meant with the quote below.

It's a bit silly to ask anyone to prove a negative, btw.

OK prove it, prove prescott is'nt/does'nt idle under long term stress at varied freq.

 

[THUGSROOK]

ok friend, buddy, pal, its time to stop your pointless bickering cause you are very alone with your arguement.

actually, WTH is your arguement???

 

[dev0lution]

There is an on-die thermal sensor. From Intel -

"There are two independent thermal sensing devices in the Pentium 4 processor on 90nm process. One is the on-die thermal diode and the other is in the temperature sensor used for the Thermal Monitor and for THERMTRIP#"

 

[LTC8K6]

I don't recall arguing with anyone about anything.......

I don't think I have even claimed anything at all about Prescott in this thread, other than the on die temp sensor, and that was about P4's in general.

 

[Zebo]

Originally posted by: LTC8K6


It's a bit silly to ask anyone to prove a negative, btw.

OK prove it, prove prescott is'nt/does'nt idle under long term stress at varied freq.

You can prove a negative in this case by proving performance scales with frequency. What's been shown so far is already proven, performance got worse as CPU frequecy increased due to prescotts idleing. See page 1 and see chinebench thread.

Once again obsfucation by you.

 

[stevty2889]

All I can say for sure is that the throttling is kicking in at 70c. At 69c it's fine, at 70 it starts to throttle. Higher the temp above 70c, the more it throttles. It does not appear to be adjusting the frequency at all, watching CPU-Z the frequency stays the same, even as the throttling goes as high as 66% throttling according to throttle watch. When running cinebench, the rendering would clearly slow down during the throttling, but when playing games, even if the throttling was going on, it doesn't show any noticable effect, so I never realized it was going on before. I lowerd my voltage, and the temp stopped going above 70c, and the throttling stopped. Benchmark scores started increasing again at higher clock speeds, as long as the throttling didn't kick in. All I can say is that unless you are watching for it, it won't likely be noticed, but if you are using a 3.6 or 3.8 P4 for rendering, which was the one small advantage the P4 still had in certain programs, the throttling will very well affect rendering times, and I highly doubt the stock Intel cooling will keep temps down low enough to keep it from happening. My CPU scores were higher on my 3.2es at 3.45 than my 3.4 at 3.8ghz due to the throttling kicking in. Once I got the temp to stop going above 70c, by lowering the voltage, the scores went up to where they would be expected. My system doesn't stay 100% stable at 3.8ghz without the higher voltage, so I have lowered down to 3.74ghz to keep it stable and prevent the throttling.

 

[jpetermann]

Hehehe...

I had one of the first Prescotts in Retail. At first, I thought wow! I can save on my heating bill... then I had to take out a special insurance policy just in case of a Pressie fire....

Actually, I treid a few pressies, went back to Northwood, and then made a complete switch to A64... not going back until there is a real reason.

 

[LTC8K6]

Well, I don't think one case is enough to go on here. There wasn't much performance lost, and Super Pi gained.

More testing is needed, imo. The longer Super Pi test, as suggested earlier perhaps?

Do Prescotts in general throttle too much, too early, all the time? I don't know, reports vary too much.

Are Prescotts in general too hard to keep cool? Some report yes, others report no.

I have certainly seen posters here whose Prescotts were indeed throttling when they certainly should not have been. No doubt about it. I have also seen posters here whose Prescotts weren't throttling at all.

Why were those Prescotts throttling when others weren't? I have no idea. Maybe it has to do with unfamiliarity with the LGA775 socket and it's HSF installation? Maybe the Prescotts were just too difficult to keep cool? Again, I don't know.

If you want to make a conclusion about the issue, that's fine with me.

I don't have enough information to decide yet.

 

[Zebo]

Bring on dothan for goodness sakes! It probably has higher IPC than A64 and runs cooler...only detering factor may be price which should come down with massive production.

Look, p4, while a marketing gimmick from the biginning, got real sweet around northwood C time, only your most ardent AMD fanboy could'nt see that. People like me were still detered by it's price/performance which kinda sucked compared to Bartons (mobiles for $90), tbreds (1700's of $44) etc. But it was king of performance and did'nt have this 3.2 ghz "some of the time effect." Prescott is a disaster. It's slower clock for clock than northy and your chip has a high propensity for not running it's rated speed under extreme bore conditions like gamming w/o exotic cooling.

 

[LTC8K6]

Heck, I'm still not sure if the OP's Prescott was throttling at stock speed and voltage. I don't see where he reports on stock speed via throttlewatch.

Not much of an effort was made to see if the problem could be solved, either.

 

[stevty2889]

Perhaps it is possible to disable thermal throttling if somebody know which pin is for the PROCHOT# signal..I have an engineering sample I would be willing to experiment with if anyone knows where to find the pinout for LGA775.

Found it, it is pin# AL2. Now just gotta find out if isolating it will disable thermal throttling...

 

[THUGSROOK]

Originally posted by: dev0lution
There is an on-die thermal sensor. From Intel -

"There are two independent thermal sensing devices in the Pentium 4 processor on 90nm process. One is the on-die thermal diode and the other is in the temperature sensor used for the Thermal Monitor and for THERMTRIP#"

you dont have to point me to Intel white papers.
P4 has a thermaltrip for throttling.
there is no ondie cpu temp. or if there is, no motherboard maker uses it. NONE NADA ZIP.

even Abit plainly stated that its CPU temp is not real, its calculated using a formula based on the PWM temp.



is class over yet?
im getting tired of teaching "basic" P4 knowledge :disgust:

 

[Zebo]

Originally posted by: LTC8K6
Heck, I'm still not sure if the OP's Prescott was throttling at stock speed and voltage. I don't see where he reports on stock speed via throttlewatch.

Not much of an effort was made to see if the problem could be solved, either.

You don't?

First of all throw away SP 2M, it's not near enough load time to show it.

dipite lower mem speeds he runs HL faster at 3.2 than it's stock 3.4 speed.
3,2--------------------17x188-----640x480 98.84 fps
3,3---------1m31s---17x194-----640x480 97.58 fps
3,4ghz-----1m28s---17x200-----640x480 96.48 fps

Like ski said, perhaps lowering to 3.0Ghz would have netted him 100FPS LOL.

But this thread is still inconclusive on that point (underclocked vs. stock) until more testing done.

 

[Strawa]

Originally posted by: stevty2889
All I can say for sure is that the throttling is kicking in at 70c. At 69c it's fine, at 70 it starts to throttle. Higher the temp above 70c, the more it throttles. It does not appear to be adjusting the frequency at all, watching CPU-Z the frequency stays the same, even as the throttling goes as high as 66% throttling according to throttle watch. q]

Yep, that's true, my 3,4ghz starts throttling with 63'C (gigabyte software readings), and cpu frequency seems to be unchanged.
I still have to investigate game's performance drop when increasing frequency even if throttle watch didnt register any throttling.
Stock Intel cooler sucks big time anyway

 

[Zebo]

LTC8K6 or anyone..

Where are you at now, speed wise? Do you have stock HSF?

If you have time would you test some things?
I have some ideas..

 

[Zebo]

Originally posted by: Strawa

Originally posted by: stevty2889
All I can say for sure is that the throttling is kicking in at 70c. At 69c it's fine, at 70 it starts to throttle. Higher the temp above 70c, the more it throttles. It does not appear to be adjusting the frequency at all, watching CPU-Z the frequency stays the same, even as the throttling goes as high as 66% throttling according to throttle watch. q]

Yep, that's true, my 3,4ghz starts throttling with 63'C (gigabyte software readings), and cpu frequency seems to be unchanged.
I still have to investigate game's performance drop when increasing frequency even if throttle watch didnt register any throttling.
Stock Intel cooler sucks big time anyway

Don't forget to press f5 on throttle watch before you launch bench, then again after you exit. The log file will tell you how much you idle along with times of benchmark itself.

 

[THUGSROOK]

Originally posted by: Zebo
>snip<

dont bother Zebo, these guys arent willing to listen to experience.

i hate it when members "force" me to talk down to them.
id much rather chat with ppl on my level, or at least ppl that are willing to listen.

ive wasted enough of time in this thread.

 

[LTC8K6]

I just remembered that I got into a big email argument with an Intel rep over the 570J and throttling re that Sudian article. They basically brushed me off. I asked them about recourse if I bought the chip and it throttled all the time. They just kept saying they would replace a defective chip, whch means nothing.

They said throttling could happen with inadequate cooling and high case temps. I pointed out that the test was sans case with ambient temps of 22C. That's when they blew me off with the line about replacing defective processors.

This was my final response to the rep:

Alex,

Thank you for your time. I am still not reassured regarding the 570J
processor. I still suspect that the Sudian testing is flawed, but I guess
resolution of that will have to wait until independent replication is
attempted. I was hoping Intel could point out the flaw.

If these processors do throttle way too much in a typical case when they
start being used more by the public, I am sure we will hear about it on the
forums.

Bill

That was in early December. Kinda ironic.....

I'm at work on a Dell with a Northwood.