Just a tiny little detail.. you should be comparing the 6870 to the 5850, not the 5870. The HD 68xx series are mid-range. The HD 69xx will be the top of the line (replacing 5870, etc).

IMO the most difficult part in cooling a hot gfx card (using aftermarket coolers) is the VRM cooling.
For the GPU there are currently several good performers that can handle a hot mid-range card like the 6870.
But for the VRMs AFAIK only the Thermalright solutions are getting under 100C temps.
So please, in your gfx card reviews, try to mention compatibility with aftermarket VRM coolers.
In this case, are the Thermalright solutions R1 - R5 compatible with the 6850 / 6870 ?
This information is not yet available (AFAIK) in the forums and I believe it's important. Thank you.

Weird... many other sites are citing a very high idle noise for the stock cooler, while here it looks like it's actually very good. Seems they tuned fan speed accordingly in the newer BIOSes.
Good to hear about the VRM temps, with both the 5870 and 5850 they were an issue even with the stock cooler at default frequencies.

It seems to me that TechPowerUp! and HT4U's findings should be consistent with SPCR ones.

Any chance to fit it in my rig? Is my psu ok?

Only if you mail me your 5750, otherwise I'm sure the 6870 won't fit.

I don't want to argue with you any more but TPU finds the cooler noisy, and in HT4U the two 6870 cards tested, while using the same reference layout, are clearly performing differently in the noise department, confirming there might be at least two different BIOS revisions (or cooler/fan/whatever else variances, but this seems more unlikely, the differences are way too clear).

Don't worry, I'm not such a kind guy as flapane!


You're right, I mistook 6850 results with 6870 ones.

Hi there.

I´ve read your articles with great interest for a couple years.

There is one thing that puzzles me though: When I read your articles, I think the temperatures in your system is overall very, very high.

I cannot at all recognize these temps in my own system. Which is Athlon II x2 250 with an 92 mm Arctic Freezer, XFX HD4770, Gigabyte GA-MA770 UD3 and a Antec Sonata 500 case with three fans, Antecs own Tri-Cool at the rear and two Coolermaster 120mm in the fornt. All running at 5 Volts, it is very quiet, just a breeze-sounding noise.

I have disengaged C&Q and clocked the processor to run at 3225 mhz.

And my temps are appr.: IDLE Cpu:17, HDD 34, Graphics 38, celsius
LOAD Cpu 34, HDD 38, Graphics 55, celsius.

And this is after gaming CoD for several hours, 1680x1050, all maxed out.

Of course I´m happy with this, it seems my system is running cool and quiet, I could just enjoy that. But when I seek information for a new GFX, which is my next upgrade, it´s a bit annoying, that the temps you have are unrealistic high

The other day I was scared away from the HD 6970 because it was both noisy and very hot. Until I checked out the review of my own card HD 4770, and realised, that your system is extremely undercooled and therfore also very loud, compared to my world.

Which is not the North Pole, by the way, ambient 20 degrees:-)

So in order to have a more real-world scenario, I think you should work heavy on your test-system.

Don´t misunderstand me, I realy enjoy reading here, the articles are great, and since I use my PC for music production, the subject of quiet computing is important to me.

Thanks.

Best regards

Uh? Which system? The SPCR test bed?


IMO you should note that according to your readings at idle your CPU would run 3°C below ambient... and it's not worth to mention that it shouldn't be phisically possible.


Usually the SPCR test beds are sort of "worst case scenario".
Anyway as already said, check your readings, as they look like suspicious.

Well, it certainly must be... Thats ok, I´l do the math when ever I read an review.

I just checked with Open HW monitor, HW Monitor and the BIOS, they all agreed, +/- 2 degree to the same as before. It will rise a little bit over the next period though, winter is coming where I live, and I have underfloor heating in my house.

The CPU and GPU temps look way too low. I doubt they are accurate readings. There's no way a CPU core will run at 17C, that is an obvious telltale. Only way you could get such low temps on both CPU and GPU after being "maxed" out gaming for hours is to have ridiculously loud fans on uber heatsinks -- and that's not what you describe, so the sensor readings are suspect. While a HD4770 isn't going to run that hot... nor a Athlon II x2 250, those temps just don't look like anything we get, though admittedly our testing setups demand more from both CPU & GPU.

You might just try the combo of prime95 + furmark at their most extreme settings until temperatures stabilize (usually a half hour or more) simultaneously for an apples-to-apples comparison. That's how we test.

Probably you miss the point here, kevith, maybe due to my not so good english.

Even if the ambient were 17°C, your CPU - as any current CPU all over the world - could not run on air at 17°C, as it is not possible.
I'm not saying neither "I think it's not possible", nor "I don't believe to it": it's just plainly phisically impossible (as well as flying donkeys).

AMD officially never recognized it in the Errata 319 of the 10h Revision Guide (so, if not a cpu one, maybe it is a rather common motherboards design issue), but the suspicious temperature readings is a well known and rather widespread issue for the Regor/Rana/Propus CPUs: so, your readings are actually inaccurate, and you do need to calibrate them (not to mention you don't explain how you are able to check at the same time a nearly equal reading onto HW Monitor and BIOS).

At anyway, these things are quite OT here (a discussion about the Radeon HD6870), maybe you would open a dedicated thread about (if in case, please use a more proper quotation, it was so difficult answering your post in a polished way).

Well, I can only say, that this is what I read out all the time. In win 7, win XP and Kubuntu. 32 bit and 64 bit.

Except when I experiment with fans and their placement, then the temps of couse changes. And during the summer they rise 3-4 degrees.

But ok, thanks for your time and the answers, I will leave it here, youŕe right, maybe it was a wromg place to write this, in a GFX review.

But those ARE my temps, whether its possible or not.

Best regards

I didn't want to be rude or harsh: I apologize if I looked like so.

About what's possible or not, those are your actual readings (I mean, measurements: AMD, since the Brisbane cores, has a long story of incorrect temperature measurements), and I trust you about them.
Just your real temps however, even if currently unknown, ARE a tad higher: you only need to find out which is the actual offset. If you mind, of course.

P.S.: I've taken two screenshot from a 45W X4. Look at them: idle and load (4 x Prime95 + FurMark, 15mins):


A > 29°C offset (ambient ~23° C) from AMD readings to have a more reasonable measurement.