Heatsinks' TDPs can be trusted?

[Imago]

Hello everyone!

This isn't a question, it's more a "what do you think" thing...

I have in my PC a R7 5700G which is quite toasty, it easily almost reach 90 C° in no time and it takes some minutes to go back to 40C° idle.

I know that's a 65W TDP chip while my cooler is rated for 140W TDP with MX-4 applied on it... In theory it should be a bit overkill keeping my CPU fresher than that.

Now, if I buy a 260W TDP CPU cooler (one of those colossal tower 18 CMs tall, 2 Kgs, eight heatpipes, three fans monsters) I still have same temps?

It's almost the double TDP of my cooler but my cooler is already more than double of TDP of the CPU. This makes me doubt a lot of the stated TDPs values on products.

[Levent]

I don’t because it’s mostly due to how shit AMD and Intel’s definition of TDP is. Common sense is the only thing I use when I am buying coolers.

[TatamiMatt]

4 minutes ago, Imago said:

Hello everyone!

This isn't a question, it's more a "what do you think" thing...

I have in my PC a R7 5700G which is quite toasty, it easily almost reach 90 C° in no time and it takes some minutes to go back to 40C° idle.

I know that's a 65W TDP chip while my cooler is rated for 140W TDP with MX-4 applied on it... In theory it should be a bit overkill keeping my CPU fresher than that.

Now, if I buy a 260W TDP CPU cooler (one of those colossal tower 18 CMs tall, 2 Kgs, eight heatpipes, three fans monsters) I still have same temps?

It's almost the double TDP of my cooler but my cooler is already more than double of TDP of the CPU. This makes me doubt a lot of the stated TDPs values on products.

Have you applied the thermal paste right and have the correct tension on the cooler, i have a little slim 4 heat pipe tower with one fan, was overclocking my 10600K and hit right up near 100C, but when i stopped the test it dropped right back down to 40s/50s in a matter of seconds. If it takes minutes to cool and idle chip by 50C then how is it performing when actively cooling a hot chip

[will0hlep]

7 minutes ago, Imago said:

Hello everyone!

This isn't a question, it's more a "what do you think" thing...

I have in my PC a R7 5700G which is quite toasty, it easily almost reach 90 C° in no time and it takes some minutes to go back to 40C° idle.

I know that's a 65W TDP chip while my cooler is rated for 140W TDP with MX-4 applied on it... In theory it should be a bit overkill keeping my CPU fresher than that.

Now, if I buy a 260W TDP CPU cooler (one of those colossal tower 18 CMs tall, 2 Kgs, eight heatpipes, three fans monsters) I still have same temps?

It's almost the double TDP of my cooler but my cooler is already more than double of TDP of the CPU. This makes me doubt a lot of the stated TDPs values on products.

So...

 

1) Those cooler TDPs can't be trusted (neither can the chip TDPs)

2) If your getting 90C at idle (or close to idle) on a R7 5700G with a better than stock cooler then I'm not sure the cooler is the issue. Have you tried removing the cooler, checking for plastic on the cold plate, reapplying the thermal paste (following AMDs or your coolers manufacturers instructions) and then reinstalling the cooler?

[lafrente]

There a lot of variables. Quality and application of paste, how good is it mounted, ambient temp, case temp, case airflow etc.

 

If customer environment were exactly the same as testing environment of manufacturers, maybe.

[porina]

14 minutes ago, Imago said:

Now, if I buy a 260W TDP CPU cooler (one of those colossal tower 18 CMs tall, 2 Kgs, eight heatpipes, three fans monsters) I still have same temps?

It's almost the double TDP of my cooler but my cooler is already more than double of TDP of the CPU. This makes me doubt a lot of the stated TDPs values on products.

CPU TDP is the cooling required to guarantee you get base clock. The CPU at stock can use more power than TDP. Better cooler means possibly better boost clocks. AMD 65W TDP CPUs are typically set to 88W PPT limit.

 

AMD CPUs have a relatively high power density. Moving the heat away from the silicon is difficult. If it can't be move efficiently from the CPU to the cooler, having a better cooler might not help as much as expected. This is not a new problem.

 

Having said that, I would not expect any half decent cooler to struggle with this particular CPU. Try using monitoring software and observing what the actual power usage is. Temps should drop rapidly when CPU returns to idle after load, but not necessarily all the way back to minimums.

[mariushm]

The TDP of cooler is estimated at a particular amount of airflow through the fins of the heatsink. 

 

Motherboards tend to have the fan curves set to keep the system on the quiet side, so the fan is kept at lower rpm and isn't spinning up at maximum rpm. The fan rpm also tends to be gradually increased instead of just switching to full speed in an instant. 

 

Also, processor may have 65w TDP but that doesn't mean it consumes that much or under that, it actually is allowed to consume more power and produce more heat - the idea is that the heatsink has a particular amount of mass which takes time to heat up, so the cpu can boost for short periods consuming more power because the cool heatsink will absorb that extra amount of heat 

[Imago]

28 minutes ago, Levent said:

I don’t because it’s mostly due to how shit AMD and Intel’s definition of TDP is. Common sense is the only thing I use when I am buying coolers.

Ok... But how common sense could tell me if the 300$ high end industrial grade cooler I'm going to install will keep my CPU fresher than the 10$ cooler from Temu?

Noctua rates the dissipation using NSPR... This is even more confusing!

Anyway, to answer the guy that asked about the installation, I'm sure I applied the paste correctly with a good mount pressure and no protective plastice laft on the direct contact plate.

[TatamiMatt]

21 minutes ago, Imago said:

Ok... But how common sense could tell me if the 300$ high end industrial grade cooler I'm going to install will keep my CPU fresher than the 10$ cooler from Temu?

Noctua rates the dissipation using NSPR... This is even more confusing!

Anyway, to answer the guy that asked about the installation, I'm sure I applied the paste correctly with a good mount pressure and no protective plastice laft on the direct contact plate.

Most high end heatsink tower coolers should max out around a premium of like £100 or so, anything more is paying extra and if you find cheaper options they can sometimes just be as good, Noctuas NH-D15, the £100 option, has quite a few competitors now in the £40-60 price range that actually in some cases match performance and in some scenarios can actually outperform, and the NH-D15 is/was cream of the crop for the longest time, seen as gold standard.

 

Also on your heatsinks, i dont doubt youve checked all the mentioned, but something isnt right if it takes that long to drop at idle. It should be seconds, even 10s of seconds at a push, minutes seem to be way off the chart. And its not the solution to the issue but, do you have any case fans moving any of the hot air away from the heatsink? Also, out of curiosity, what cooler do you currently have?

[Mark Kaine]

1st its an estimate,  2nd if your case cooling sucks (no fans because i already have a cooler???) then it's not going to perform as advertised. 3rd its an estimate. 

 

get what im saying?  its an estimate for "perfect condition" and will obviously vary depending on on a gazillion factors like fan speed, etc.

 

 

[Mark Kaine]

4 minutes ago, TatamiMatt said:

Most high end heatsink tower coolers should max out around a premium of like £100 or so, anything more is paying extra and if you find cheaper options they can sometimes just be as good, Noctuas NH-D15, the £100 option, has quite a few competitors now in the £40-60 price range that actually in some cases match performance and in some scenarios can actually outperform, and the NH-D15 is/was cream of the crop for the longest time, seen as gold standard.

nuh, noctua never was seen as "the best", more as the "not quite as loud as others" which is indeed still the case...  others may perform better or similar,  sound like a jet engine though. 

Noctua forever!  (tm)

[TatamiMatt]

8 minutes ago, Mark Kaine said:

nuh, noctua never was seen as "the best", more as the "not quite as loud as others" which is indeed still the case...  others may perform better or similar,  sound like a jet engine though. 

Noctua forever!  (tm)

Fair enough, have used anything from noctua myself, but £100 is still quite a premium for a cooler that other £40 coolers can also achieve, albeit not as quiet, but £60 for some noise reduction is quite a bit for me personally. And in fact Thermalrights new phantom spirit has both marginally better cooling and better noise levels accross all temperature ranges right at that £40 price

[GarlicDeliverySystem]

The thing with coolers and TDP is that it tries to put a whole process into a single number, which is doomed to fail (unless you specify the exact conditions).

 

On a high level, the heat energy needs to be transferred from the chip to the IHS, then via the thermal paste to the cooler and then distributed among the cooler. Only then do you transfer it to the ambient air.

 

Now, each of these transfer steps requires a difference, or gradient, between the two parts or regions; the stronger the gradient, the faster the heat transfer. This is where you could see TDP come in, since it is a defined as "power" (aka energy per unit time). However, you would also need to know the temperature gradient, or in other words the temperature difference between hot and cold (very much simplified).

 

A stock cooler will be able to dissipate 150W of thermal power to ambient air, but only by having a large thermal difference and thus high temperature on the hot side. To put it bluntly: a red-hot glowing Intel stock cooler will probably dissipate a lot more thermal power than a 360 AIO with the hot side at 50C. 

 

Other factors remain like how well the cooler itself conducts heat from the cold plate (hot side, the place in contact with the CPU) to the fins or radiator. AIO/liquid coolers have the advantage here because they can physically transport the medium, rather than relying on conduction alone (plus rad size limitations are different etc.). So a cheap cooler with the same surface area of aluminium fins might perform worse than a pricy competitor simply because a higher temperature difference is required to drive the transport of heat energy from CPU to fins. I think this is usually called thermal resistivity or thermal resistance.

 

So, when manufacturers put TDP values on anything: that is just the amount of thermal power to be dissipated. Under which conditions, that is the real question here.

[OddOod]

2 hours ago, Imago said:

R7 5700G

I have had AMD chips with absolutely bonkers temp readings. three pastings and nothing helped. It would bounce by 10C every second. Got an RMA for it

 

[RevGAM]

Please provide complete specs of your system. 

[Imago]

3 hours ago, Mark Kaine said:

2nd if your case cooling sucks (no fans because i already have a cooler???) then it's not going to perform as advertised

 

Well, I have a Chieftec CI-01B-OP v2 with one fan in front sucking air in and one in the back pulling air out (Both are Arctic F12) plus another one on the top pushing air directly on top of the CPU Cooler (Another ARCTIC F12). The cooler is a Mars Gaming MCPU220 with the original fan replace by an Arctic Bionix P120. Dust filters removed to increase airflow. Fan curves are set in the BIOS to maximize airflow when under heavy load, the fans gradually reach 100% when the CPU goes over 75C°.

I did the "smoke test" once and the air seems to flow correctly inside the case, I didn't find and "hot pocket".

Unless I did some mistake in the fan setup, the airflow should be good in my case.

[RevGAM]

9 minutes ago, Imago said:

Well, I have a Chieftec CI-01B-OP v2 with one fan in front sucking air in and one in the back pulling air out (Both are Arctic F12) plus another one on the top pushing air directly on top of the CPU Cooler (Another ARCTIC F12). The cooler is a Mars Gaming MCPU220 with the original fan replace by an Arctic Bionix P120. Dust filters removed to increase airflow. Fan curves are set in the BIOS to maximize airflow when under heavy load, the fans gradually reach 100% when the CPU goes over 75C°.

I did the "smoke test" once and the air seems to flow correctly inside the case, I didn't find and "hot pocket".

Unless I did some mistake in the fan setup, the airflow should be good in my case.

What are the temps like if you run it with the panels off?

[Imago]

2 minutes ago, RevGAM said:

What are the temps like if you run it with the panels off?

I have to try.

I'll do that now and see what  get.

[Imago]

Two degrees less than usual and about 50 seconds to be back to initial 40C° instead of a full minute, so the case isn't the cause of those temps.

I'm 100% sure about the cooler's installation and pasting, so there are two cases:

1) the CPU isn't a 65W (or 88W as some say) but more like 200W.

2) the cooler isn't a real 140W

Even if the CPU were 105W the cooler should be able to cool down the chip.

[Imago]

5 hours ago, VennbergsU said:

Instead of changing the cooler, after I checked the HSF mounting system and I found I haven't seated corner screws at the very end, I repositioned them as should be and also rotated the cooler radiator completely by 90 degrees -> result after 20 minutes of CPU-Z stress test: 65C° max!

I tried repasting and sadly it didn't worked so magical.

My cooler uses the stock AMD cooler bracket, it's like "click" and the cooler is in place, no need to fidget with adapters or screws. After a bad experience with another cooler with proprietary bracket, I prefer the stock system.

[RevGAM]

I'm afraid you lost the silicone lottery. As @OddOod said, you should get an RMA and have them replace the CPU. 

[Mark Kaine]

16 minutes ago, RevGAM said:

I'm afraid you lost the silicone lottery. As @OddOod said, you should get an RMA and have them replace the CPU. 

well, did they say what cooler now, a chiefteck? uh-oh...

 

 

i mean maybe the cooler is just bad, get a peerless assassin for 40 bucks instead before changing out a probably perfectly fine cpu?

 

 

1 hour ago, Imago said:

Well, I have a Chieftec CI-01B-OP v2 with one fan in front sucking air in and one in the back pulling air out (Both are Arctic F12) plus another one on the top pushing

also that sounds you don't have exhaust and intake fans, except one intake blowing on top of the cooler, which is less than ideal.... hot air goes up so you want it out in the top, not "in"

 

 

picture of your setup would help, but as for now i think I'll go with bad airflow in your case (no fans because you already have a cooler!!!) 

 

either that or its just not a good cooler.... clips on the amd bracket.... never heard of something weird like that ¯\_(ツ)_/¯ 

[Mark Kaine]

6 hours ago, TatamiMatt said:

Most high end heatsink tower coolers should max out around a premium of like £100 or so, anything more is paying extra and if you find cheaper options they can sometimes just be as good, Noctuas NH-D15, the £100 option, has quite a few competitors now in the £40-60 price range that actually in some cases match performance and in some scenarios can actually outperform, and the NH-D15 is/was cream of the crop for the longest time, seen as gold standard.

 

Also on your heatsinks, i dont doubt youve checked all the mentioned, but something isnt right if it takes that long to drop at idle. It should be seconds, even 10s of seconds at a push, minutes seem to be way off the chart. And its not the solution to the issue but, do you have any case fans moving any of the hot air away from the heatsink? Also, out of curiosity, what cooler do you currently have?

yeah but thing is you don't need a d15 for most cpus (certainly not amd) i have a Noctua u12s chromax for 50 bucks and its totally fine for my 5800x3D,  60-70c ,no throttling etc, idles at around 40c...

 

i also have a bunch of chromax fans in my case around 10 bucks each, so while not the cheapest, they're far from the most expensive and the sound, performance and looks are certainly worth it for me (i hate the beige noctuas, would never buy one of those lol)

 

 

ps: i even have those rubber "screws" (not from noctua) but i think it certainly helps with vibrations and also no nasty screwing and metal flakes everywhere, yuck!

[Mark Kaine]

1 hour ago, Imago said:

I tried repasting and sadly it didn't worked so magical.

My cooler uses the stock AMD cooler bracket, it's like "click" and the cooler is in place, no need to fidget with adapters or screws. After a bad experience with another cooler with proprietary bracket, I prefer the stock system.

yeah, that just sounds like it'll never make proper contact.  you need to screw it in! click isnt a thing that works, i don't think so.

 

 

ps: get a Noctua,  the most easiest retention mechanism ever, takes 5 minutes max to install the complete cooler! (well good luck with the fans, but the cooler itself 5 minutes tops!) 

[RevGAM]

1 hour ago, Mark Kaine said:

well, did they say what cooler now, a chiefteck? uh-oh...

 

 

i mean maybe the cooler is just bad, get a peerless assassin for 40 bucks instead before changing out a probably perfectly fine cpu?

 

 

also that sounds you don't have exhaust and intake fans, except one intake blowing on top of the cooler, which is less than ideal.... hot air goes up so you want it out in the top, not "in"

 

 

picture of your setup would help, but as for now i think I'll go with bad airflow in your case (no fans because you already have a cooler!!!) 

 

either that or its just not a good cooler.... clips on the amd bracket.... never heard of something weird like that ¯\_(ツ)_/¯ 

First rule of tech support: always consult the log before responding. 

 

The OP has 3 fans per prior statement. Already repeated multiple times, no plastic on cold plate, etc.  Also, running it with panels of barely helped.  Ergo, the CPU is crap...

 

I'm surprised you're not familiar with the mounting system described... it's real.