[Mito [PL]]

Yeah, I know that I could ask this on any PC tech forum... But I just feel more comfortable asking it here. Because reasons.

So, I'm a (not so) proud owner of a Phenom x4 9500 mounted on an ASUS M2N-X Plus motherboard with a stock cooler. All that powered by a Chieftec 500W power supply.
I was using this stuff with a really ancient BIOS version and a HD5770. The BIOS supported my CPU, but didn't recognise its temperature properly - looking at HWINFO32 sensor data resulted in constant reading of about 54C. Well, everything worked fine (but the performance wasn't so great...).
But once upon a time, I got a R7 265 GPU (not my call). Once it arrived, I decided to update my BIOS to the latest version (was supposed to have a fixed temperature calculation algorithm for some CPUs) and replace my GPU. Well, i did manage to update the bastard, but I will never do it without an external flash programmer. Never ever again. (Irrelevant). Also replaced my GPU, cleaned and installed the drivers etc. Also started using a Phenom TLB workaround tool that I've found somewhere on the net (that performance boost is sweet). Everything works fine...
...except CPU clocks under load. Once the temperature reaches 57C, the clock is cut to half (1105 MHz) for several seconds, then it goes back to 2210MHz, everything is fine for half a minute (or less), then the drop repeats. On a graph, it looks like a PWM modulation of CPU clock speed in order to never exceed the temperature of 57C. The CPU has Tjmax of 70C, VRMs are barely warm, chipset temps look OK,  I don't really think that this PSU will drop its voltages under that kind of load. The TLB workaround doesn't affect the problem (well, it does drop temps by about 2C in idle when the workaround is used).
So I've replaced the thermal compound under my cooler (to a cheap silicon one). Now the CPU throttles at about 55/56C.
Fiddling around with BIOS settings like Cool'n'Quiet and microcode updation (fixed someone's hecta-core t1060 with a similar problem), nothing.
Now, before I underclock my CPU a bit/slap a jet engine to my cooler/buy a new cooler with better thermal compound, I want to ask if any of you are familiar with such a problem, maybe someone might come up with something?

And well, if I had the cash to buy a good cooler, I would rather get some more cash and grab a 775 mobo with a modded 771 xeon and be over with it .

[LHN91]

First thing I would say is to try to get a quality spare PSU to try.

If the PSU wasn't a terribly good one, and it's now several years old, it may be not able to push out enough voltage to push the CPU and the R7 265 - as heat increases so does power draw, and that might be what's causing the behaviour.

[Mito [PL]]

Well, I don't really think that's the issue: the same throttling happened both during playing different types of gaming:
TF2 - large CPU and marginal GPU utilisation
Mass Effect 3 - both stressed,
Stalker Shadow of Chernobyl with Autumn Aurora graphic mod - large stress on a single CPU core, GPU crying for mercy...

It also happened when watching Youtube in large screen mode when it was warmer than usual in my room...

So i would seem that it isn't the PSU (especially when not even 50% utilised), but in any case I am still in possesion of a Corsair GS600 that I repaired myself and still need to test it... I've seen some funky funky stuff happening with electronics, so I don't exclude the possibility of my PSU going nuts.

[Mito [PL]]

Sorry for a double post.

I just found out that cooling down my room enough would cause the CPU to hit the max of 52C... because it would throttle there. Damn.
I'm going to take all possibly related HWINFO32 sensor readings (if that interests anybody).

[jr2]

http://www.tomshardware.com/answers/id-2807550/max-core-temp-phenom-9500-folding-home.html

So the max operating temp on that chip is 70deg C, some older AMD chips could hit like 90. That said at 115 it should have either throttled to a crawl or crashed, so I have a feeling that program is not giving you the correct information. Install AMD Overdrive on the computer and monitor thermal limit, "0" is 70, and whatever number negative is how far below you are.

If somehow this chip is running above 70 degrees it is not long for this world.

http://www.teamliquid.net/forum/tech-support/186697-good-computer-unplayable-fps-during-engagements

Your GPU is too hot (and liable to kill itself) but what's limiting your frame rates at your current settings is actually the CPU. The first-generation Phenoms had a TLB bug which Windows 7 works around at the cost of a massive performance drop. As a result they're hopeless for SC2.

Ideally you'd upgrade the CPU, as they're not great for SC2 even when uncrippled, but it's not easy to work out what's compatible with OEM motherboards. You could try disabling the TLB fix by following these instructions:
http://www.rebelshavenforum.com/sis-bin/ultimatebb.cgi?ubb=get_topic;f=13;t=000111

http://www.tomshardware.com/forum/253341-31-athlon-6000-problems

To anyone reading this post, hear this: I've got the same problem as u all! Got a brand new computer in december 07, an athlon 6000 and got problems with it ever since. Bought it for gaming but games like bioshock or overlord requiring a lot from the cpu made it stutter like crazy even with a lot of exhaust fans (I've got a good gaming case). So... been searching and searching here and there. An ok solution I saw on an amd forum was to put a fan or the upper left corner of the motherboard to help the heatsink. It works but only to 95% while the demanding games are still ''lagging'' some here and there.

Now listen, what I found is the athlon 6000 is generating a lot of heat as you already know, so much that with stock fan/heatsink that comes with it your motherboard will have a hard time tolerating all this. What you need to fix it (hope it works for you) is to buy an arctic cooling freezer 64 pro, you can have this on ebay for 32 $ canadian, and apply the right amount of thermal paste (if u buy it new a pre-applied paste will be given). You install it with the fan facing the front of your case and ... tadam! problem solved for me! So before listening to others who will tell you to play in the bios or changing your motherboard (which can solve it I heard it can in some cases) try that easy solution to see if it works for you cause for me it works like a charm! Hope it helps!

Will a 212 Evo Plus or whatever work for you?

http://www.tomshardware.com/forum/332941-28-cooler-master-hyper-cooler-master-hyper

willardApr 26, 2012, 10:07 PM
Get the 212+ and spend the extra cash on a second 120mm fan. Will perform measurably better than the Evo.
darkclownjoeApr 26, 2012, 10:09 PM
willard said:
Get the 212+ and spend the extra cash on a second 120mm fan. Will perform measurably better than the Evo.

+1

212 Plus, $19.99 after $10 MIR:  http://pcpartpicker.com/product/gpw323/cooler-master-cpu-cooler-rrb10212pg1 

EDIT: You might want to put stuff like this in Gaming Discussion, as Gen Disc doesn't show up on unread replies, which is how some people check for new posts.

[Mito [PL]]

Oh, thanks for the answer.

1. I am absolutely aware of the 70deg limit, but the throttle happens here at anywhere between 52 and 59 (was able to push it that hard a few days ago). However, I have read somewhere that that component composition can give some fake temperature readings...
2.

Also started using a Phenom TLB workaround tool that I've found somewhere on the net (that performance boost is sweet).

Using that already, for some time now. And no, I haven't yet found a way to push my Sapphire R7 265 over 65deg (except removing the fans and/or the heatsink).
3. That's an idea - but before I spend money (I can find similar used coolers around my country for quarter the price) I need to try out some other ways. I may be mean as hell, but life isn't that easy (for me, at least at the moment).

After taking a look at voltages in HWINFO32, I found out that the Chieftec PSU actually dropped 12V rail voltages all the way to 11.4V when under load. So I've replaced my PSU with a Corsair GS600 that I bought for about 2$ and repaired myself. While input voltages are simply perfect now, I *think* the problem is a little bit less frequent, but I am unsure if it isn't just placebo effect.
Now, after taking a look at Vcore readings, there's something interesting. Nominal Vcore is 1.25V, but under load the CPU receives something varying between 1.23V and 1.26V. Drops to 1.23 usually when at 2.2GHz and gaming, and stabilising at 1.26 as soon ans the chip throttles to 1.1GHz.
That's why I suppose I should cool down my VRMs... somehow. There isn't a lot of space to fit a fan over them and not much to attach it to. I will try something out before buying stuff (though a nice cooler will be needed in summer).

Oh, and... May I ask somebody to move the thread to Gaming Discussion then, please?

[jr2]

Can you get your hands on some old discarded laptops?  They usually have copper heatpipes, maybe you could run a heat pipe (just bend it to fit, heat it up with a torch if you have to beforehand) from the VRMs?  You'll have the heatsink attachment and fan built in as well, just point it at an exhaust fan.

[Mito [PL]]

This is a very interesting solution, thanks! I may even find a spare heatpipe laying around... However, it would seem that the motherboard's layout is going to make such modification a VERY difficult one (take a look at VRMs https://www.asus.com/media/global/products/Baevh0NrwM6fG33b/P_setting_fff_1_90_end_500.png). Plus, I don't really think I'd have a way to attach the heatpipe to the board reliably...

My first idea was to grab some tiny heatsinks (like these https://liliputing.com/wp-content/uploads/2016/03/loverpi-heatsink.jpg) and slap them to the MOSFETs, as they come with a thermal conductive tape and are dirt cheap. Perhaps with a small fan blowing at them.

[jr2]

Hmm.  I've got a bunch of small heatsinks in a bag somewhere that should work rather beautifully, I *think*, let me see if I can find them (unless you can get them for free or dirt cheap, but let me see if I can take pictures, I think they'll pretty much exactly fit what you need.  Assuming I can find them.)

[Mito [PL]]

Since I was home early, I grabbed my PC, tossed it onto the table and hanged a single fan above my VRMs. Then I tested it with some games (i. e. Metro 2033 Redux on Medium or absolutely maxed out TF2). I was able to push my CPU to 65C, not even once its frequency dipped. Looking at HWINFO -> Vcore, the graph was much more consistent than normally. So it seems that the VRM cooling is the issue.

I should also have something like these small radiators laying around - and even if I don't, I can always purchase a pack of 6/8 with thermal conductive tape for something around 3 - 5$, so no trouble here. And even if I decided to get them from you... Do you happen to live in Poland? Shipping from other countries might well exceed the value of these heatsinks (I think).

[jr2]

Since I was home early, I grabbed my PC, tossed it onto the table and hanged a single fan above my VRMs. Then I tested it with some games (i. e. Metro 2033 Redux on Medium or absolutely maxed out TF2). I was able to push my CPU to 65C, not even once its frequency dipped. Looking at HWINFO -> Vcore, the graph was much more consistent than normally. So it seems that the VRM cooling is the issue.

I should also have something like these small radiators laying around - and even if I don't, I can always purchase a pack of 6/8 with thermal conductive tape for something around 3 - 5$, so no trouble here. And even if I decided to get them from you... Do you happen to live in Poland? Shipping from other countries might well exceed the value of these heatsinks (I think).

I actually completely forgot about this until just now.  I only mentioned because I have a bag of them cheap and I'll probably never use more than a couple of them in the rest of my life, lol.  They don't weigh much, if I could ship them for a few bucks I would.  However, I just went and looked, and they aren't in the box I thought they were in -- if you can't get some cheap enough, I'll do a thorough search and find them (they're pretty much useless to me at this point) but other than that, they're somewhere in my collection of old parts.

[Mito [PL]]

So, one more question from me guys.
I have recently purchased an Asus P5Q motherboard and a Xeon E5450 CPU (from China, 5 weeks of waiting). Yeah, the famous 771 -> 775 CPU mod. It works for me perfectly. The only issue is that I've got a crappy Intel stock cooler, so the temperatures are quite high. They're about 80C with case closed and 70ish with it opened (both at extended period of stress), so I thought of undervolting the chip a bit to lower them, so I took a look at my Vcore. Both HWINFO32 and CPU-Z told me that it's 1.032V when idle and 1.04 when stressed...

Now, the only thing I need to know is how high the typical Vcore of these CPUs are. From some topics on other forums I saw voltages at stock frequency at about 1.15V. Now I'm wondering, maybe my motherboard is somehow malfunctioning a bit, or that's just a good hit on silicon lottery?

[The E]

Investing into a better cooler sounds like a better idea, tbh.

[Mito [PL]]

It's summer, and it's really hot here. It will get a little bit better when it gets cooler outside.

And yeah, I've got a better cooling solution coming. I'm basically going to modify an AM2 socket water block (no problem with that), and I've already got the water cooler itself together with the pump. Since this is going to be a dirt cheap water cooling, I suppose I'll use some pneumatic tubing to connect everything together. Seriously, who would buy such a motherboard just to not overclock the proccessor?

It's going to be fun!