High Temperatures

[WobbleWobble]

I have a Q9550 (C1) on a EP45-UD3P and have been getting what I feel are higher than normal temperatures.

I can't remember what temps are like with stock volts, but with between 1.35V and 1.4V with LLC enabled overclocked to 3.6-3.8GHz, I will get idle temperatures with around 40C (give or take a couple degrees). When loading my CPU with Linpack in OCCT, I will average temps in the high 70s and sometimes even break 80C by a couple degrees.

I've tried the following heatsinks:
- Lapped Thermalright Ultra Extreme 120 with both a 800RPM 120mm Scythe Slipstream and a 1200 RPM Noctua NF-P12 (I know these aren't ideal for heatsinks)
- Kingwin RVT-12025 (same as Xigmatek HDT-S1283) with the stock 1350RPM fan

Both heatsinks give me almost identical temperatures. I'm using OCCT 3.0, RealTemp and Everest to monitor temps (they're all pretty much reporting the same).

I've been using Arctic Cooling MX-2 thermal compound and tried the various application methods. For the Thermalright, I've been following the Arctic Silver method with a line on the CPU. I've also tried the small bead method as well. For the HDT cooler, I put a thin line down each heatpipe.

My cable management seems OK. Motherboard temperature is about 50C.

The rest of my system is:
- P182 (closed case)
- 800RPM Scythe Slipstream (top)
- 1200RPM Noctua NF-P12 (rear)
- 1000RPM Nexus (front)
- 1200RPM Noctua NF-P12 (bottom)
- 3xHDDs (2xSamsung 250GB, 1xWD 640GB)
- GTX260
- Corsair 620HX
- 4x2GB Patriot Extremes (2.2V)

Any thoughts where I can improve my system? I don't really want to lap the CPU.

Thanks!

 

[zagood]

Any major differences with your case open vs. closed?

 

[Kraeoss]

ambient temps ? are you living in a hot area ? that can affect that sometimes and also u could have used too much tc, you can also try changing the orientation of the fans for different airflow

 

[F1shF4t]

Your temps are fine especially considering the 1.35 vcore (I get 86C load at 3.8ghz and 1.3 vcore). Can your chip run stable with lower volts?

I've tested mine yesterday (config in sig) and got 78C under linpack, opening the case usually makes 3-5C difference. I'm using a TRUE with two 1200rpm fans, which is comparable to your setup.
80C is roughly the max temp to aim for under linpack, means nothing you ever do on the comp (besides running linpack) will get to those temps.

 

[chizow]

Lapping the CPU will certainly also help lower temps, the other suggestion is changing the mounting system for your TRUE. Does the TRUE slide around with very little force with the stock mounting system? I found the stock mounting brackets on the TRUE didn't provide nearly enough pressure because the screws bottom out into the base bracket.

What I did was I took a spare TT Big Typhoon back plate (you can either make the holes bigger with a drill bit on the TRUE backplate or buy another bracket kit for ~$5) and used four 2" 8/40 bolts inserted from the back. Then I inverted the TRUE's top X-bracket and used 8/40 wing nuts to secure the X-bracket to the screws. I also placed a neoprene washer between the X-bracket and heatsink base for additional pressure and to prevent scratching and shifting.

After doing this and re-lapping my Q6600 (I had already lapped the TRUE), my temps dropped 10-14C in Prime95/OCCT/IBT. Before doing so I wasn't happy with my temps as they were literally identical to the Tuniq Tower I replaced. If you're OK with the mounting pressure I'd highly recommend lapping the CPU, its very safe if you have that little plastic cover and some masking tape.

 

[BonzaiDuck]

Originally posted by: WobbleWobble
I have a Q9550 (C1) on a EP45-UD3P and have been getting what I feel are higher than normal temperatures.

I can't remember what temps are like with stock volts, but with between 1.35V and 1.4V with LLC enabled overclocked to 3.6-3.8GHz, I will get idle temperatures with around 40C (give or take a couple degrees). When loading my CPU with Linpack in OCCT, I will average temps in the high 70s and sometimes even break 80C by a couple degrees.

I've tried the following heatsinks:
- Lapped Thermalright Ultra Extreme 120 with both a 800RPM 120mm Scythe Slipstream and a 1200 RPM Noctua NF-P12 (I know these aren't ideal for heatsinks)
- Kingwin RVT-12025 (same as Xigmatek HDT-S1283) with the stock 1350RPM fan

Both heatsinks give me almost identical temperatures. I'm using OCCT 3.0, RealTemp and Everest to monitor temps (they're all pretty much reporting the same).

I've been using Arctic Cooling MX-2 thermal compound and tried the various application methods. For the Thermalright, I've been following the Arctic Silver method with a line on the CPU. I've also tried the small bead method as well. For the HDT cooler, I put a thin line down each heatpipe.

My cable management seems OK. Motherboard temperature is about 50C.

The rest of my system is:
- P182 (closed case)
- 800RPM Scythe Slipstream (top)
- 1200RPM Noctua NF-P12 (rear)
- 1000RPM Nexus (front)
- 1200RPM Noctua NF-P12 (bottom)
- 3xHDDs (2xSamsung 250GB, 1xWD 640GB)
- GTX260
- Corsair 620HX
- 4x2GB Patriot Extremes (2.2V)

Any thoughts where I can improve my system? I don't really want to lap the CPU.

Thanks!

Wobble-x-2, there are several factors to consider. Other posters here, such as Chizow, are giving good advice.

There had always been issues about Penryn Wolfdale and Yorkfield temperature sensors. You aren't specific about which sensors are giving you those temperatures. There is a legacy sensor known as "TCase," which measures temperature at the center of the IHS or processor cap, and there is a "TJunction" or core sensor for each Wolfdale or Yorkfield core. The pattern since the Conroe C2D's and Kentsfield C2Q processors has been such that TCase would be 5C to 10C lower than the core average.

The Intel thermal spec is defined to reference TCase. I can't speak for others, but while some core sensors on the Wolfdales and Yorkfields seem to be "stuck" at some value above what would be a sane and normal idle temperature, the TCase sensor, at least on my Wolfdale, works fine.

However, there are other things to consider. With the Penryn cores [Wolfies and Yorkies], Intel implemented a revised algorithm for interpreting the sensors. Non-Intel motherboard and BIOS makers did not easily keep up with the revision (as described in a March, 2008 Anandtech article here on the E8500), so there was a lag in BIOS revisions for Penryn-compatible boards from the time the CPUs (especially the C1-steppings) hit the street. You are advised to see if there are more recent BIOS versions than what shipped with your board, and -- if not familiar with the process -- do a little reading and research on flashing the BIOS PLCC chip on your motherboard.

Now, given those factors and caveats, your temperatures don't seem too far off what they should be with the levels of VCORE voltage you are applying. I think the INtel thermal spec for those CPUs is about 72C, and there is still wiggle-room above that. And keep in mind -- that spec applies to the TCAse sensor. Fact is, if you are just clipping 80C under Linpack, you'll never come within 10C of that temperature under normal use.

But you CAN improve things. I assume you lapped the TRUE cooler so that it is flat, and if you did so, you ground off the nickel-plating. As Chizow suggests, unless you're worried about your Intel 3-year-warranty (which you've violated anyway by manually setting the voltage) -- carefully lap off the nickel plating on the IHS processor cap.

Also, for a 100W-TDP processor such as yours, you'll get about 2C to 4C improvement in load temperatures by using:

Micronized or "nano-diamond" Thermal Paste

At least, that improvement will show over the use of Arctic Silver 5. It will either be as good or better as the compound you say you're using.

I'm guessing that lapping the IHS flat to bare copper will give you maybe 4 or 5C improvement, and the diamond paste may add a little to that.

You might then try pushing CFMs through the TRUE corresponding to nearly double the rpm speeds you cite, and if you can thermally control the speed, a good fan for the system won't even be that noisy at 2,500 rpm.

Take a look at the DELTA tri-blade 120x38mm fan -- providing you can control it thermally. [It tops out at 3,700 rpm and 51 dBA]. I'm using a lower-powered 0.50 Amp San-Ace 120x38mm fan, which doesn't exceed about 2,500 rpm. If you can, find a way to use these beefier fans as case exhaust fans and duct them to the cooler. That lessens the weight of the TRUE+fan.

On that Gigabyte board, I haven't thoroughly researched, but the "Ultra-Durable" aspect of the board design should eliminate the risk of greater weight on the cooler -- so you could also just hang the fan on the cooler as you would normally.

 

[WobbleWobble]

I have the latest F8 BIOS. I'm pretty sure I'm readnig Tjunction temperatures.

Room temperature is about 19C. I was running Linpack last night and it's staying around the mid 70s with 1.36V BIOS (LLC on, 1.34V actual), which I guess is acceptable. It's just that when I read reviews of various heatsinks, people are idling in the mid 20s and loading in the 50s.

Yes, TRUE does provide some wiggle room but the HDT cooler seems to be ok. I'll check out some of that thermal paste BanzaiDuck linked. I have tried various amounts of thermal compound. Initially yes, I think I was using a bit too much but I've cut back.

I'm not so concerned about warranty, but resale of the chip once lapped. But I'll consider it if it helped you guys that much.

I think I'll try a push-pull configuration. Generally I go for quieter computing, so that's why I have 800RPM Slipstreams. I'll try a 1200RPM high static pressure fan as I know the Noctua I have isn't the best CPU fan (but a great case fan).

But I guess the main thing that has me concerned is the higher than what I percieve to be normal idle temps which result in high load temps.

 

[BoboKatt]

1.34- 1.36V? Have you tried fiddling with other voltage settings in your BIOS? I have a q9550 C1 as well running 3.7 GHz with 8 GB of DDR2 on an older P35 Asus p5K-E. My vcore is 1.2v in bios and shows slightly less in CPU-Z, same case as you, same TRUE and lapped (not the CPU tho).

Funny thing is though, my idle temps are same as yours if not actually higher but under load for hours, it wont go much over 60c.

I worked hard to find the sweet spot of my RAM and to make sure when I OC'ed my CPU that the RAM would not be an issue. Then I started playing with the other voltage options and kept trying to reduce the actual vcore as much as possible.

 

[TidusZ]

My Q9550 at 1.392v llc enabled hit mid 80's in linpack with the best of air cooling, and I am not concerned. Your temps are fine, go for 4 ghz imo or enjoy what you got now.

 

[BonzaiDuck]

Originally posted by: WobbleWobble
I have the latest F8 BIOS. I'm pretty sure I'm readnig Tjunction temperatures.

Room temperature is about 19C. I was running Linpack last night and it's staying around the mid 70s with 1.36V BIOS (LLC on, 1.34V actual), which I guess is acceptable. It's just that when I read reviews of various heatsinks, people are idling in the mid 20s and loading in the 50s.

Yes, TRUE does provide some wiggle room but the HDT cooler seems to be ok. I'll check out some of that thermal paste BanzaiDuck linked. I have tried various amounts of thermal compound. Initially yes, I think I was using a bit too much but I've cut back.

I'm not so concerned about warranty, but resale of the chip once lapped. But I'll consider it if it helped you guys that much.

I think I'll try a push-pull configuration. Generally I go for quieter computing, so that's why I have 800RPM Slipstreams. I'll try a 1200RPM high static pressure fan as I know the Noctua I have isn't the best CPU fan (but a great case fan).

But I guess the main thing that has me concerned is the higher than what I percieve to be normal idle temps which result in high load temps.

OK -- you have the latest BIOS. I still don't trust the idle temperature readings on these Penryn-based cores. True, I've only had two Wolfies to play with, but you'll find threads from others on this topic pertaining to the Yorkies.

For what it's worth, I have a B3-stepping of the Kentsfield quad, and just checked the core temperatures in Everest Ultimate at a room-ambient 72F. The core idle temps are between 35 and 38C. This is for the chicken-s*** over-clock of 20% above the stock 2.4 Ghz, and a voltage setting around 1.31V ~= idle sensor reading. By comparison, your Q9550 is OC'd 34% above stock. Even my Wolfie E8600 -- @ 4.25 Ghz -- only represents a 28% OC.

On the lapping issue, I can't see really how that would affect resale, if the buyer is "one of us" and appreciates useful work done toward welcome ends. There's not a lot of risk, if you take modest precautions. I use the black-plastic cover that clips onto the CPU's PCB, to keep the LGA-dots from contamination; hold it on there firmly as you lap the IHS; probably start with 240 or 320-grit; use a minimum amount of water on the sandpaper -- a few drops here and there. And frankly, by the time you've ground down the nickel-plate, the coarser-than-recommended sandpaper has also degraded so that it's more like 400-grit.

As for BoboK__, he may be lucky with his C1 stepping. But figure -- given the OC on my Wolfie -- that Wolfies should show less thermal dissipation than the (double-wolfie) quads.

By the way -- there is a CPU_VTT or CPU_FSB voltage setting. Is it still set to the default value? I'm guessing that with FSB-speed of around 447 Mhz, it should be higher than default as long as it's set so the sensor reading is below 1.4V. On some mobos, you actually have to tweak the termination voltage to get beyond 400 Mhz, and doing so might even reduce the VCORE voltage required for stability at the speed you're shooting for.