What do I need to get Pent D to 4Ghz...

[Puni Puni Poemy]

Sorry to resurrect this from the dead but Ii have a few more questions.

With a Pentium D, Am I able to lower the Multiplier if I want to (Or is it fully locked)

Also I was thinking about finally sucking it up and stepping into water cooling.

is 4333Mhz Possible with a Pentium D on water?

4333MHz (13x333)
1333FSB (333x4)
DDR2 @ 1000 (3:2)

Is this even Possible? Thanks

[Lithan]

Most people I see can only get 4.3-4.5ghz on Single stage, so I think 4-4.2 is about the max you'll see typically on water.

[Donnie27]

keep in mind that its a prescott that we're talking about which already push air cooling on single core. prescott + dual core generates upwards of 210 watts of heat, thats 3x what my p4 is rated for. now overclock it. this is the reason few motherboards will support the chip. the mofsets get mightly toasty.

First of all, you can get an i945P instead of an i955X. These sell for as little as $119. They're similar to buying a none SLI or Ultra nForce4 for you AMD fans, sheesh!

How much power?

http://www.gamepc.com/labs/view_cont...=pee840&page=4
One of their other review talk about heat.

Donnie

[Donnie27]

Sorry to resurrect this from the dead but Ii have a few more questions.

With a Pentium D, Am I able to lower the Multiplier if I want to (Or is it fully locked)

Also I was thinking about finally sucking it up and stepping into water cooling.

is 4333Mhz Possible with a Pentium D on water?

4333MHz (13x333)
1333FSB (333x4)
DDR2 @ 1000 (3:2)

Is this even Possible? Thanks

NO! Only the Extremely Expensive model is unlocked.

Donnie

[Puni Puni Poemy]

NO! Only the Extremely Expensive model is unlocked.

Donnie

Not even the Lower Multis huh. That sucks

Well what If I were to get an EE? All cost aside just for purposes here for my general knowledge... Are they binned to OC higher? and then I could play with the Multi.. so then what I said before be possible?

[Lithan]

Binned to OC higher? No, no cpu manufacterer does that. As long as it can run stock speeds in their testing and they need it, it's sold as a top end chip. If they don't need it and/or it cant run top end, it's dropped down. Not really sure how HT works internally, whether they make all the D's with HT and just disable it or whether they Make the EE's seperately and aren't (yet) selling the ones that won't run stock.

According to intels spec, the 840's use ~ 175watts maximum @ stock.
According to intels spec when compaired with the measurements donnie linked to, they use ~ 220watts maximum @ stock (assuming the testers were loading all processors equally.

[mr_knowitall15]

wow thank you people like fugger and donnie, and nasgul, for helping me prove a point. and lithan, thank you for FINALLY being civil. the last two posts you made were actually good.

[Lithan]

Everyone in this thread except nasgul and yourself have been civil.

[socrilles]

Sooo, I was thinking about a Dell OCing project #3. So far I have a 9300 with PM 1.6 @ 2.13, a Dell 8400 630 @ 4 ghz, and am now considering a 9100 2.8 P4D @ 3.73 any opinions for doing this on air at stock voltages? earlieer posts seemed to think 3.73 would be possible on air

[Lithan]

Not in a Dell. Dells are designed to just barely get by with their cooling (for silence). You'd definately need to upgrade the cooling for stability/not throttling. 3.73 is possible with good air cooling though. Not sure about stock vcore. So far the only overclocks I've seen with stock vcore are simple Suicide screens or else running on cascades. I'd say 3.6ghz with upgraded air cooling on stock vcore wouldn't be too hard, but that's mostly a guess. 3.73 I'd suspect would require a little extra vcore.

[socrilles]

Actually you would be surprised at how good the cooling on Dells are these days, if you glanced at my post more thoroughly, you would see I have a 3 ghz chip running at 4 ghz in a Dell. The HSF included has 4 heat pipes, and is probably 2.5X2X4 LWH, something that would cost $50ish at any retail store. Since the Dells support the higher end processors ie Dual Core EEs, and P4 3.8 ghz the HSFs they use have to be able to cool the high ends, and are still included with the low end processors.

look at the HSF at the bottom of this auction for reference

[Charles Wirth]

Lithan, might I suggest owning a 820 or 830 before running it into the ground.

It is a fact they run cooler than the 840 by a large margin and I do not believe any binning was done. The 840 has HT enable whitch usually led to a bit cooler temps but that trend does not follow with the dual core.

Do not compare what THG did, they used a heatsink from a 2.8Ghz CPU (check the amps) and it was one that required a shroud to duct in air. If you go off on using a shround then I will go off on having to use clips or having to use a power supply. Its part of the spec and should be followed for best results. We all know a hot case is not good for the dual core as you are running two prescott cores.

I said 266x14 was easy without raise in vcore but to reach 4Ghz a slight raise might be needed for weaker boards.

Taking into consideration you are slightly educated and use a good case like the antec lanboy with dual 120's and a good zalman heatsink there is not a problem disipating the heat load and no throtteling will take place.

Enough of the nay saying, unless you own one and report in on results bagging on this topic will be frowned upon.

[Lithan]

I don't own a Humvee, but I know they are gas guzzlers. You and everyone talk like 4ghz stable on air is a simple matter, but you've yet to provide a single example of this. Whereas at least a couple examples of much less on air and only slightly more on phase change have already been posted. If you say that noone can say bad about a product without first buying it, then that kind of reduces the discussion of Intels Duals to fanboy gushing, since after the results of every single review (including some from well-known intel biased sites) no objective buyer would purchase one, especially not overclockers, seeing as single core p4's destroy these things in just about everything when both are overclocked, without even considering amd's offerings.

As to a shroud, I saw the heatsink, it's design was NOT one that would benefit more than slightly from a shroud. In fact it is in layout almost the same as their current retail heatsinks, designed in such a way as to achieve maximum effect from the way an unshrouded fan directs air. (a star pattern so to speak).


Socrilles, a couple points.

First, Dell does not have a universal heatsink. If you bought a 2.8ghz sc400 you got a different heatsink than if you bought a 3.2ghz (I know because I bought both). I owned the heatsink you are talking about, It had too few fins, spaced too far apart, and the joint between the copper and aluminum was poor. You could buy a similar heatsink these days for under $20 at coolergarage most likely. Also, it relied on indirect airflow from a case fan ducted to ~ 1cm above the heatsink. Almost identical to the cooling on my old P200 gateway. You might also mention that your 4ghz overclock was not stable as I recall reading. It was impressive for a dell, but I'm sure that in a warm room, you would have had temperature issues even if the overclock was stable. Also, I don't believe you were monitoring throttling on that machine... were you? And we are talking about a Single core Prescott 6, not a multicore prescott (which I believe is based off the 5 model though I'm not sure). There is ~ 100 watts difference at stock here.

Second, Dell uses temp-adjusted fans. But these fans aren't very powerful even at full speed. So there is a ceiling to the wattage they can deal with. It's a safe bet that a singnificantly overclocked Dual core will hit this ceiling and hard.
Third, Dell's case cooling (at least on their affordible servers and home pc's) is absolutely horrendous. Unless they have a direct duct out of the PC from the heatsink, this will be a huge thermal issue no matter how effective their cpu cooler is.

[Charles Wirth]

Lithan

The shroud is to pull fresh outside air and that is important.

Yes I have tested the 820 and in my threads you will find results. I had no problem pushing to 4Ghz on air with a Zalman. Eshbach as well has breached 4Ghz with stability on air.

The dual core is superior to HT in many ways, one is lower latency with arbitration and the other is true multi core processing. It will work circles around a HT or single core in threaded applications not just synthetic benches. Sure single cores are good for games and single stuff and can perform better in this scenario. But of you need dual CPU, the dual core is an option.

All of the lame reviews are on the EE part from what I have seen, this same also holds true about THG getting fresh air to the CPU from outside the case. No matter how much you do not want to accept this, its part of the spec.

The memory timing is so critical and to dumb the Intel down with crap timing hurts it very bad. And to throw a memory ratio in the mix compounds the problem more. And the AMD gets 2-2-2-5 1T...

Lets get away from reviews, this is XS and we do not live with SPD timing at stock FSB. That being said, the 820 or 830 at 266FSB 1:1 3-2-2-4 1T the picture changes drasticly and gives the X2 a run for a lot less. We do not have to run crappy ass settings the reviewers run nor do we have to run inferior heatsinks.

[Charles Wirth]

Quoted Lithan:

"Good luck with the chip. God, Xtremesystems has gone to the trolls. It's like the Intel version of anandtech. If I see one more imbicile claiming 4ghz is easy @ stock vcore and stock cooling with 820's, someone's gonna die."

Your input on this topic is understandable but your negativity is unacceptable. I am not sure whats your beef with Intel but you are the one trolling here from those who reported your posts.

My retail chip was from Fry's and was $299, not a hand picked sample in any way. I used a P5WD2 mobo. The machine now lives/runs at 4.4Ghz 1:1 3-2-2-4 turbo and was pushed to 4.9Ghz for benching.

Please reread, this imbicile clearly stated upgraded heat sink and small increase in vcore for 4Ghz

Im not mad at you Lithan as I see where your comming from but please dont disrespect us like that or you will lose all respect with the community here.

Calling us trolls and imbiciles requires an apology from you. That was uncalled for and did not go un noticed.

http://www.xtremesystems.org/forums/...ighlight=p5wd2

[Lithan]

Lithan

The shroud is to pull fresh outside air and that is important.

The side was off in toms test, and the 840ee w/ 820 HSF was therefore getting fresh outside air and still breaking 90*C before it would crash.

Yes I have tested the 820 and in my threads you will find results. I had no problem pushing to 4Ghz on air with a Zalman. Eshbach as well has breached 4Ghz with stability on air.

Links to SS of dual primes?

The dual core is superior to HT in many ways, one is lower latency with arbitration and the other is true multi core processing. It will work circles around a HT or single core in threaded applications not just synthetic benches. Sure single cores are good for games and single stuff and can perform better in this scenario. But of you need dual CPU, the dual core is an option.

In every review I've seen, HT performs better for intel in more instances of multithreading than Dual Core. Throw in that you are clocking higher and unless you have an 840ee running quad high util apps or dual high util multithread apps, it's no contest. What's your source on lower latency in arbitration? That's counterintuitive, so I'd like to see some evidence.

All of the lame reviews are on the EE part from what I have seen, this same also holds true about THG getting fresh air to the CPU from outside the case. No matter how much you do not want to accept this, its part of the spec.

ee adds about 6watts of consumption, so its highly doubtful it adds more than that in heat.

The memory timing is so critical and to dumb the Intel down with crap timing hurts it very bad. And to throw a memory ratio in the mix compounds the problem more. And the AMD gets 2-2-2-5 1T...

Memory timings affect 3dmark 2001 and spi pretty severely, but in most everything else I've seen, it wont be more than a couple % points. (Edit: also certain compression/decompression tasks get hit pretty hard with loose timings.)

Lets get away from reviews, this is XS and we do not live with SPD timing at stock FSB. That being said, the 820 or 830 at 266FSB 1:1 3-2-2-4 1T the picture changes drasticly and gives the X2 a run for a lot less. We do not have to run crappy ass settings the reviewers run nor do we have to run inferior heatsinks.

P4's are competitive with loose timing ddr2, they get a little better with tighter timings. How do you maintain that a chip of essentially the same design goes from not even close to competitive to competitive with the same changes?

Quoted Lithan:

"Good luck with the chip. God, Xtremesystems has gone to the trolls. It's like the Intel version of anandtech. If I see one more imbicile claiming 4ghz is easy @ stock vcore and stock cooling with 820's, someone's gonna die."

Your input on this topic is understandable but your negativity is unacceptable. I am not sure whats your beef with Intel but you are the one trolling here from those who reported your posts.

I have no beef with Intel. All I do is point out that just because a chip doesn't live up to expectations, you can't just pretend that it does and challenge people who advise against purchasing it to prove you wrong by buying one. That's asinine.

My retail chip was from Fry's and was $299, not a hand picked sample in any way. I used a P5WD2 mobo. The machine now lives/runs at 4.4Ghz 1:1 3-2-2-4 turbo and was pushed to 4.9Ghz for benching.

Now was that on Cascade or just single stage? Hell people are benching Dothans near 4ghz. I'd hardly call 4.9ghz a success for netburst architecture. Aren't you benching close to 6ghz with single cores these days?

Please reread, this imbicile clearly stated upgraded heat sink and small increase in vcore for 4Ghz


Im not mad at you Lithan as I see where your comming from but please dont disrespect us like that or you will lose all respect with the community here.

Calling us trolls and imbiciles requires an appology from you. That was uncalled for and did not go un noticed.

If you didn't say that 4ghz was easy on stock vcore and cooling, then I wasn't talking about you. But to make such an outrageous claim is trolling. To maintain it against all reason and logic is imbecility.

http://www.xtremesystems.org/forums/...ighlight=p5wd2

Once again we are confronted with the difference between Screenshot stable (and almost synonymous, sandra stable) and 100% stable. You've worked with P4's surely more than I have. And even I know that on a p4 Sandra stable is Hundreds of mhz away from actual stability.

[Charles Wirth]

Side off is not pulling fresh air to CPU, its stagnant and no air flow at all is happening and the EE runs a lot warmer, I know as I have the 840EE and it can only do 3.73Ghz on stock vcore.

best you will get from me is a 32m pi. I dont run prime for hours.

My source on arbitration is Intel, I am an Intel developer. This was knowledge we had back before you even heard of dual core.

Think back when HT was introduced and people complained in the performance hit when compared to cores without HT??

where did you come up with ee adds 6 watts?

Like I said, forget about reviews and crap timing. Quit with the nonsense on loose timings is good stuff.

If you look at the other screen shot it shows a cinebench but that is a short intensive run as well.

When you are talking trash about XS you are talking about me and I take it personal.

Next post better be an apology from you, the lack of one already shows your intent.

[Lithan]

Side off is not pulling fresh air to CPU, its stagnant and no air flow at all is happening and the EE runs a lot warmer, I know as I have the 840EE and it can only do 3.73Ghz on stock vcore.

That's not true unless the fan only flows ~ 10cfm or less or is in a very small case. With the flow a typical 80mm fan gets, the air is easily reintegrated with room air and within a degree or two of ambient when it get's fed back to the fan. The much more important purpose of a shroud is to direct airflow (hard to see on heatsink fans, but with a box fan, the effect is very evident). By preventing intake at an angle you decrease the angle of the exhaust. But due to Intel's heatsink design (which takes advantage of angled exhaust from the fan), this wouldn't be desirable. The only thing Intel could hope to gain from an intake shroud would be to keep fresh air at the cpu in spite of case temps. In actuallity, it's impact on airflow would probably do more damage than good when compaired with an open case.

best you will get from me is a 32m pi. I dont run prime for hours.

My source on arbitration is Intel, I am an Intel developer. This was knowledge we had back before you even heard of dual core.

32pi takes what, 20min? How about dual runs of Prime for 20min?

I'm not saying you are lying. I believe you were told that. But I suspect it was PR BS used by all Marketing people, because they don't understand it and they assume you won't either. I'd like to see some white pages that mention this if you can find/provide them.

Think back when HT was introduced and people complained in the performance hit when compared to cores without HT??

So we're compairing undeveloped HT with unoptimized programs versus the current incarnation of HT and programs that have been building on HT for years? You know that isn't fair. With a few Unoptimized HT programs and in certain Multitasking situations (due to problems with HT, HT does have problems with task assignment, but this is a flaw of HT (A bug if you will) and fairly uncommon. It's not a matter of Dual cores performing better. It's similar to putting a single core w/o HT into some of the tests we've seen (high priority encoder + low priority game and use games FPS to decide winner in a 3200+ versus 3.2c situation for example) you are setting it up to lose intentionally, so of course it isn't a fair compairison.


Edit: It's very likely Dual cores may perform better in most tasks once we are using apps optimized for dual cores instead of HT-enabled single cores. But by then Intel will have much better dual core cpu's out, so I can't recommend buying a D based on that prediction.

where did you come up with ee adds 6 watts?

System consumption of identical systems with 840 ee (ht) on/off @ full load.

Like I said, forget about reviews and crap timing. Quit with the nonsense on loose timings is good stuff.

I never said it was good. I said it wasn't as bad as you make it out to be. If it was as bad as you try to make it sound, there would have been a massive public outcry when Intel switched to DDR2. As it was, there was a minor performance hit that mostly slipped under the public's radar.

If you look at the other screen shot it shows a cinebench but that is a short intensive run as well.

I did miss that.

When you are talking trash about XS you are talking about me and I take it personal.

I've seen some pretty abject behavior by certain members here and I comment on that. You don't have to ban people for lying, trolling, or verbally assaulting others in your forums but then you also can't blame the people who complain about this behavior if you don't.

Next post better be an appology from you.

I'm sorry that I was in a bad mood and made the mistake of referring to the forums when I only meant to refer to the small group of members who were making the loudest noise here at the time. There are plenty of good members who didn't deserve to be lumped in with those.

[Charles Wirth]

Thanks,

http://www.xtremesystems.org/forums/...ght=smithfield

Pretty much all the results I cannot match or even get close with my 3.73EE at stock speed (14x266) same as my air stuff in that thread.

dual core arbitration is better than HT, by how much depends on the application. I mean if you are running web server in background and playing doom3 vs ripping DVD and playing doom3 the dual core would fare much better at the DVD + game though both would do good at web serving and game.

In professional applications, like Postershop (a program to resize images to billboard size) an HT processor would appear as a single CPU and the application would not run as a true dual CPU even though you could start two instances. The dual core would show up as two processors and you could run in dual cpu mode correctly there are a few other applications that react this way that I have run across, another is wasatch.

You have two of the top dual core clockers and a link to another member running it now. That should be a pretty good indication that the 820 and 830 are good clockers. I can say for sure that the 840EE is a much heavier load on the system for some reason in the design, maybe its the HT logic heating it up a bit more where that is dead space in the 820/830.

[Charles Wirth]

I just ran these at 3.73Ghz, the same speedas the dual core. Its a single and shows that you lose 2 seconds with semi crap timing, I could add a few 5's in there to see how much worse it gets.

I wonder how much OC on the CPU I would need to make up those 2 seconds.





For bandwidth intensive apps this is a huge performance hit. And I am sure most of the reviews would have favored Intel in a lot more tests that were run.

[Lithan]

Those are all synthetic's though. How about some running some Ht-enabled encoders and some Single-threaded encoders + Heavy net use and Single threaded encoders + Playing (with whatever player you use) a dvd ripped onto a hard disk on a seperate channel from the drive you are encoding from/to?

The simple matter for this thread is that 4ghz is far from easy. You say you hit it, but your stock vcore max (if I understand you correctly) was 3.73ghz. Obviously you weren't too interested in your air max, since you're going to be PC'ing it 24/7, and you've said you aren't a Prime-er, so even if you hit 4ghz, you can't really say that it was 100% stable. I'd wager the best you can say is that it was benchmark stable. Which with the complexity of dual cores combined with the heat concerns of this processor isn't enough to point to it as stable in my opinion. Also, so far you are the only person I've seen to even claim to personally have hit 4ghz stable on air. Whereas I've seen 2 people who failed as much with water, 4 who failed as much on air, and about half a dozen who only couldn't get 4.5ghz stable on single stage. Even if you can get 4ghz completely stable, you must conceed that isn't enough to advise someone to purchase the same chip on the expectation of the same.

A much bigger thread would address the memory bandwidth issues that it's rumored this architecture will face once dual core is fully supported by software. The lackluster 64bit performance (which will very likely come almost hand in hand with dual core support for home users). The heat and power issues, that even though you say aren't sufficient to be a concern, are certainly at issue for some when compaired with AMD's offerings. And the near certainty of a much much much better processor (and at an acceptable pricerange) from intel within a year. It's like telling someone to buy a $250 Athlon Xp in 2002.

[Lithan]

I just ran these at 3.73Ghz, the same speedas the dual core. Its a single and shows that you lose 2 seconds with semi crap timing, I could add a few 5's in there to see how much worse it gets.

I wonder how much OC on the CPU I would need to make up those 2 seconds.





For bandwidth intensive apps this is a huge performance hit. And I am sure most of the reviews would have favored Intel in a lot more tests that were run.

Be fair though, short of compression utilities, Spi is about the most affected by timings of any benchmark.


Edit: Compair these two reviews. For the AMD x2 vs Intel D, Intel has 4-4-4... but for the Amd 64 vs Intel D (2.8ghz) vs Intel 630, Intel has 3-2-2. You can get a fairly good grasp of the impact timings have by cross-referencing.

http://www.anandtech.com/cpuchipsets...spx?i=2410&p=9

http://www.anandtech.com/cpuchipsets...spx?i=2389&p=7


1st multitask test. Intel 3.2ghz D (4-4-4) no HT beats Intel 2.8ghz D no Ht (3-2-2) by 8.25%

2nd: can't be compaired (they changed the method)


EDIT:


Scratch that. Found a compairison with a 3.2ghz D @ 3-2-2... so compare

http://www.anandtech.com/cpuchipsets...spx?i=2388&p=7

with

http://www.anandtech.com/cpuchipsets...px?i=2410&p=11

Only difference I can tell is 4-4-4 5:3 vs 3-2-2 4:3

[Charles Wirth]

99% of the machies that we build leave at stock speeds. I get the most out of them with memory timing and OS tweaks.

The other overclocks you mention, were they on P5ND2 or P5WD2 or ? The AW8 has 300+ FSB issues

Gimme a few days to finish my builds, I will be using a aftermarket heat sink.

[Lithan]

99% of the machies that we build leave at stock speeds. I get the most out of them with memory timing and OS tweaks.

The other overclocks you mention, were they on P5ND2 or P5WD2 or ? The AW8 has 300+ FSB issues

Gimme a few days to finish my builds, I will be using a aftermarket heat sink.

No problem. I look forward to it. I've actually got 3 or 4 others across a number of forums who just got 820/830's assuring me they'll let me know their results.

[Lithan]

Results

5:3 loose - 4:3 tight
21.3 - 21.3
143 - 143
197 - 197
171-171
34.3-34.7 (Multimedia content creation winstone 2004, Tight timings won)
257-257
336-336
259-259


After that it starts jumping around between tests and they change their methods a lot across the reviews. I did one more. Xvid with autogk.

36.2-36.2

So as far as I can tell, Anand didn't do amd any favors by running intel at a positive divider 4-4-4 instead of 1:1 3-2-2 (though in gaming and compression, it probably did hurt intel a little, you'll have to check those because all this alt-tabbing hurts my head). In fact by running AMD 1:1and intel 5:3, it looks like if anything they gave Intel an advantage in many cases.