My first build. Good temperatures?

[Kaeishiwaza]

I'm sorry. . . posts like this seem to take up a lot of threads here. But this is the first time that I've built a computer myself and I want to make sure things are OK. I loaded the computer with the prime95 stress test for 30 minutes. Here are my temps:

Idle 35C
Load 49C

Setup:
P4 3.4 (650)
Asus P5WD2
SI-120 with 120mm Panaflo (down)
Antec P180 (no front fan, top fan intake, rear fan exhaust)

I've prowled the web and searched the forums to find out what good temperatures are for the P4 3.4. I *think* these temperatures are OK, but I was expecting them to be lower. Most of the benchmarks I've found have been for AMD processors or super-overclocked P4s. Are my temperatures in line? If not, I will re-set the heatsink just to be sure. And if they are, what kind of room do I have for overclocking?

Let me know if I should include more information.

 

[BonzaiDuck]

I have a 3.2E OC'd to 3.5. It fits the older socket-478 spec, while yours is LGA775.

Based on scattered readings -- some at THG -- I had been under the assumption that the newer socket-design was supposed to reduce thermal power by reducing electrical resistance. There is too much conflicting data around to confirm that as a reliable assumption.

The first thing I would as, for anyone posting results on cooling, is that they should report the room-ambient or room-temperature in addition to their idle and load values. Motherboard temperatures are also important.

Even so, whatever the room temperature, a CPU-cooler's "idle-to-load" spread should be constant, either in Glacier Bay or a Baghdad summer. Yours is about 14C.

If your processor has a maximum thermal power of around 115W (watts), then your temperatures correspond to a thermal resistance value of 0.12 C/W.

The best reviews for the ThermalRight SI-120 show a thermal resistance of 0.14 C/W. If your thermal power is somehow "higher" -- the 14C spread will make for an even lower thermal resistance, and lower is better.

Fan speed and fan CFMs will change the thermal resistance of a cooler over a range from "barely running" to some limit beyond which even doubling the CFMs shouldn't have much effect on the cooler's effectiveness. This "threshold" of minimum thermal resistance should be somewhere shy of about 2,900 rpm for the SI-120.

So -- for assuring that the cooler is "effective" -- your temperatures are "just fine." "Perfect."

You should still strive to reduce motherboard (and therefore CPU-idle) temperature through better ventilation and ducting -- unless of course you are reporting these temperatures at a room ambient above 75 or 80F !!

Your idle temperature (about 95F) is a tad higher than mine -- 90F at room-ambient of 71F.

Another thing I did was to "sink" my Mosfet, PLL and Southbridge chips. I am still mystified by the effect that only this could account for on my idle-to-load spread. My CPU temperature only changes by about 5.5C degrees when loaded up with either PRIME95 or S&M. Someone should write a white-paper on the thermal effects of sinking Mosfet chips.

Motherboard ducting also gives you a few degrees improvement in Celsius, but it didn't -- and wouldn't -- change the spread between idle and load temperatures.

We're also pretty confirmed in our opinion that if you cool down other components, like your graphics card, it pulls all idle temperatures down somewhat. I wouldn't bother doing anything special to your Northbridge, because the SI-120 and whatever fan you use should effectively keep that cool, although motherboard ducting will really enhance cooling on the Northbridge and memory modules.

 

[BonzaiDuck]

PS

"room for over-clocking."

This is partly a matter of choice and risk acceptance. I've heard some people report that they've over-clocked a 3.2E to 4.0 with only air-cooling, but various people have different standards of rigor on the issue of stability. For me, I would validate an over-clock setting by running PRIME95 for at least six hours or S&M for at least four before I accept the OC setting as "stable."

For me, I don't really want to bump up my VCORE voltages beyond Intel's own limit. If I do, I am willing to believe that I can get to 4 Ghz, and I have an advantage already if my own temperatures are also as low as they are. But the idle-to-load spread will rise with additional over-clocking.

Maximum thermal power rises linearly with frequency and varies with the square of voltage. At constant thermal resistance, this means that the idle-to-load spread will also increase. I can't tell you precisely how this will impose an over-clocking limit, but it will -- to some degree.

Finally, your choice of memory modules will either expand or constrict your ability to OC your processor.

 

[andrewbabcock]

P4s don't oc that well especially if its a prescott.

 

[RKS]

those temps actually seem low for a Prescott. My 3.0 idles in the upper 40s in a P180 on an Intel 915pbl.

 

[BonzaiDuck]

Yeah, but it depends on what objective you're trying to achieve.

I've thought that the P4 processors -- Northwood as well as Prescott -- had an upper limit of 25% over the stock settings. Compare that with the AMD "Venice" 3000+ processor, which has a great reputation for over-clocking to 50% above stock.

With a Prescott and the right mobo, you can drop the loose-jointed frequency multiplier down two integer notches (and only two) -- then bump up the external frequency to run high-end memory at full spec. This can't be done so easily with a Northwood unless it is a Northwood in the 2.4C to 2.8C range. You're stuck with the locked multiplier and options limited to changing only the FSB in all cases with the Northwood.

The Prescott is definitely a "mixed bag," but the "pros" among the "cons" in that bag do provide some performance leeway, even if the INtel processor line is trumped by AMD. With high-end memory, you need less L2 cache to buffer system memory to L1 and the processor, but with the 1 MB L2 of the Prescott, there is a bit more performance potential and a bit more "scalability."

 

[BonzaiDuck]

RKS --

Are you using a cooler like the ThermalRight SI-120 or Scythe Ninja?

Here are my temperatures with a 3.2E @ 3.5 Ghz, FSB (800) @ 1,000 Mhz:

room CPU-idle CPU-load
70F 90F 100F
75F 95F 105F
80F 100F 110F

The last two load values are rounded up 0.5F degrees. All load values taken from PRIME95 Large FFT and S&M @ 100% load setting.

Yes, the Mosfets are "sinked," and there is a motherboard duct. But with the Prescott toaster, I can't get it to go higher than 43C.

 

[sarcasticone]

BonzaiDuck:
Sorry if this is a stupid question but could you explain what you mean by "sinking"? Yes, I'm a newb, sorry.

I have a Prescott P4 3.0 socket 478. Currently, with a room temp of about 72*F it's running about 37*C at idle and I don't think I've seen it go over 45*C at load (I'll double check those numbers). The case temps stay around 28*C. I have a side air duct positioned over the HS to pull cool air from outside the case directly to the processor, it's set to the recommended distance from the HS like Intel lists on their site. I also have 2 80mm front fans as intakes and a 120mm rear fan as exhaust as well as another 80mm on top as an exhaust. I'm using a Masscool copper HS and not the stock Intel that came with the processor. I feel these temps are decent but wouldn't mind dropping them if I could.

So if you could explain that for me I would appreciate it.

Thanks

 

[RKS]

Originally posted by: BonzaiDuck
RKS --

Are you using a cooler like the ThermalRight SI-120 or Scythe Ninja?

Here are my temperatures with a 3.2E @ 3.5 Ghz, FSB (800) @ 1,000 Mhz:

room CPU-idle CPU-load
70F 90F 100F
75F 95F 105F
80F 100F 110F

The last two load values are rounded up 0.5F degrees. All load values taken from PRIME95 Large FFT and S&M @ 100% load setting.

Yes, the Mosfets are "sinked," and there is a motherboard duct. But with the Prescott toaster, I can't get it to go higher than 43C.

I have the copper Zalman 7700. I use the Intel utility to guage temps with current CPU temp of 50 under light/moderate load.

 

[JEDIYoda]

I am currently running my 3.2 o/c`d to 4.0 on air...XP120.

My idle temps stay the same until I go past 3.6 then the temps go up to around 30c over 3.6mhz....

Under laod the spread is still the same as if I was using stock 3.2 speeds!!

Under load even at 4.0mhz I have yet to see anything over a 40c.....

Mind you I have tremendous airflow and real sweet heatsink and fan that displaces 90+ cfm........

Bonzai is on the right track