Jump to content
Search In
  • More options...
Find results that contain...
Find results in...
NumLock21

Intel recalls Xeon...

Recommended Posts

24 minutes ago, Fasauceome said:

4th gen was the last to see them. I actually managed to snag a cheap one for my brother's 7700, works a treat.

The copper core Haswell cooler had a slightly different fit configuration as well which made it a lot quieter under high fan speeds. Back in the Core2Duo/Quad days the C2D had a shorter all aluminum cooler and the C2Q had a taller copper core cooler, the copper core tall cooler would easily handle an OC'd C2D. I believe there were also taller and shorter coolers for first gen i series as well, but I didn't mess with those stock coolers as much.

 

Funny story, I passively cooled a G3460 (Haswell Pentium) with a stock copper core Haswell cooler with no fan. Only mod was a slightly lapped and polished copper base on the cooler for better contact. 20W does not need much cooling. It wasn't even snapped down, it was just resting on the CPU!

Link to post
Share on other sites

So did Intel forget how to make these?

 

image.thumb.png.53dafdca6d07f45ba4da0de92ac18189.png

 

^ This came with my i7-960, darn sight better this this joke, and both are terrible but still...

QAPylXoApw46nFuC.jpg

 

Link to post
Share on other sites
2 hours ago, NumLock21 said:

The 1 year is just speculation since it's an OEM CPU, which only carries warranty of just 1 year.

Until they say otherwise I will assume the part they are replacing with only has the warranty they specify for that part.


QuicK and DirtY. Read the CoC it's like a guide on how not to be moron.  Also I don't have an issue with the VS series.

Sometimes I miss contractions like n't on the end of words like wouldn't, couldn't and shouldn't.    Please don't be a dick,  make allowances when reading my posts.

Link to post
Share on other sites
Posted · Original PosterOP
5 minutes ago, leadeater said:

So did Intel forget how to make these?

 

 

Spoiler

image.thumb.png.53dafdca6d07f45ba4da0de92ac18189.png

 

 

^ This came with my i7-960, darn sight better this this joke, and both are terrible but still...

 

Spoiler

QAPylXoApw46nFuC.jpg

 

 

Intel does have the same one as the bottom, with the copper core.

SCR_05.thumb.JPG.bcec0427da32e6c68a60bbffa9e45773.JPG


Intel Xeon E5 1650 v3 @ 3.5GHz 6C:12T / CM212 Evo / Asus X99 Deluxe / 16GB (4x4GB) DDR4 3000 Trident-Z / Samsung 850 Pro 256GB / Intel 335 240GB / WD Red 2 & 3TB / Antec 850w / RTX 2070 / Win10 Pro x64

HP Envy X360 15: Intel Core i5 8250U @ 1.6GHz 4C:8T / 8GB DDR4 / Intel UHD620 + Nvidia GeForce MX150 4GB / Intel 120GB SSD / Win10 Pro x64

 

HP Envy x360 BP series Intel 8th gen

AMD ThreadRipper 2!

5820K & 6800K 3-way SLI mobo support list

 

Link to post
Share on other sites
19 minutes ago, leadeater said:

and both are terrible but still

Are they both terrible though? Having used both I'd only classify the all aluminum heat sink as terrible. Any cooler that maintains temps adequately and doesn't have to sound like a jet engine taking off is ok in my book


I WILL find your ITX build thread, and I WILL recommend the SIlverstone Sugo SG13B

 

Primary PC:

i7 8086k (won) - EVGA Z370 Classified K - G.Kill Trident Z RGB - Force MP500 - Jedi Order Titan Xp - Hyper 212 Black (with RGB Riing flair) - EVGA G2 650W - Black and green theme, Razer branwashed me.

Draws 400 watts under max load, for reference.

 

Linux Proliant ML150 G6:

Dual Xeon X5560 - 24GB ECC DDR3 - GTX 750 TI - old Seagate 1.5TB HDD - Dark moded Ubuntu (and Win7, cuz why not)

 

How many watts do I need? Seasonic Focus threadUserbenchmark (Et al.) is trash explained, PSU misconceptions, protections explainedgroup reg is bad

Link to post
Share on other sites
18 minutes ago, NumLock21 said:

Intel does have the same one as the bottom, with the copper core.

SCR_05.thumb.JPG.bcec0427da32e6c68a60bbffa9e45773.JPG

That's exactly the two variants of the Haswell cooler, both the same size but a different fin layout and a copper slug in one, neither cools any better than the other actually (in my testing) but the copper one is a lot quieter due to the curved fins and how the air flows through them. The fan can hit higher speeds with less noise and SOUND like it's a better cooler.

Link to post
Share on other sites
1 hour ago, Fasauceome said:

Are they both terrible though? Having used both I'd only classify the all aluminum heat sink as terrible. Any cooler that maintains temps adequately and doesn't have to sound like a jet engine taking off is ok in my book

Well it was pretty bad for the i7-960 and up but it functioned as an adequate cooling device if your aren't pushing the CPU hard for ages or tried to use the unlocked multiplier. I kept it because of how bad the current ones were getting, if you need something it'll work for anything.

 

1 hour ago, NumLock21 said:

Intel does have the same one as the bottom, with the copper core.

No increase in surface area and no increase in thermal transfer to air so both are exactly the same cooling capacity. One is better at absorbing short spikes keeping tdie and tcase lower for better boost curve but this is purely a short term function not long term.

Link to post
Share on other sites
Posted · Original PosterOP
33 minutes ago, leadeater said:

No increase in surface area and no increase in thermal transfer to air so both are exactly the same cooling capacity. One is better and absorbing short spikes keeping tdie and tcase lower for better boost curve but this is purely a short term function not long term.

Taking Core i7 8700 and 8700K as an example

I must have confused myself thinking both have the same TDP and if the Intel stock cooler is able to manage to cool the 8700, then it should also be able to cool the 8700K. I was wrong about that, 8700 only has a TDP of 65w, 8700K has a TDP of 95w.

Just went to look at some older Intel CPUs with the highest TDP that comes with a stock cooler.
 

Intel Core i7 4790K has a TDP of 88W and comes with a stock cooler as shown in that retail box in the newegg link.

https://ark.intel.com/content/www/us/en/ark/products/80807/intel-core-i7-4790k-processor-8m-cache-up-to-4-40-ghz.html

https://www.newegg.com/intel-core-i7-4th-gen-core-i7-4790k/p/N82E16819117369

 

The Xeon Intel is asking back only has a TDP of 83W (not 88W, I've corrected that in the OP).

https://ark.intel.com/content/www/us/en/ark/products/191042/intel-xeon-e-2274g-processor-8m-cache-4-00-ghz.html

Stock cooler can manage to cool 4790K with tdp of 88w should also be able to cool that Xeon with a tdp of 83w.


Intel Xeon E5 1650 v3 @ 3.5GHz 6C:12T / CM212 Evo / Asus X99 Deluxe / 16GB (4x4GB) DDR4 3000 Trident-Z / Samsung 850 Pro 256GB / Intel 335 240GB / WD Red 2 & 3TB / Antec 850w / RTX 2070 / Win10 Pro x64

HP Envy X360 15: Intel Core i5 8250U @ 1.6GHz 4C:8T / 8GB DDR4 / Intel UHD620 + Nvidia GeForce MX150 4GB / Intel 120GB SSD / Win10 Pro x64

 

HP Envy x360 BP series Intel 8th gen

AMD ThreadRipper 2!

5820K & 6800K 3-way SLI mobo support list

 

Link to post
Share on other sites
3 minutes ago, NumLock21 said:

Stock cooler can manage to cool 4790K with tdp of 88w should also be able to cool that Xeon with a tdp of 83w.

It'll be failing the task due to desktop/workstation usage being burst load and the server workload profile is much more sustained, meaning the cooler never gets time to dissipate extra heat above the capacity of the cooler. 5 minutes of 140W load on a 80W cooler isn't actually much of a problem, you may or may not throttle slightly but once the load drops the built up heat will dissipate and the cooler is ready to handle another peak load. On the other hand the Xeon could be sustaining 100W load on a 80W cooler, it'll throttle always forever.

Link to post
Share on other sites
6 hours ago, NumLock21 said:

Taking Core i7 8700 and 8700K as an example

I must have confused myself thinking both have the same TDP and if the Intel stock cooler is able to manage to cool the 8700, then it should also be able to cool the 8700K. I was wrong about that, 8700 only has a TDP of 65w, 8700K has a TDP of 95w.

Just went to look at some older Intel CPUs with the highest TDP that comes with a stock cooler.
 

Intel Core i7 4790K has a TDP of 88W and comes with a stock cooler as shown in that retail box in the newegg link.

https://ark.intel.com/content/www/us/en/ark/products/80807/intel-core-i7-4790k-processor-8m-cache-up-to-4-40-ghz.html

https://www.newegg.com/intel-core-i7-4th-gen-core-i7-4790k/p/N82E16819117369

 

The Xeon Intel is asking back only has a TDP of 83W (not 88W, I've corrected that in the OP).

https://ark.intel.com/content/www/us/en/ark/products/191042/intel-xeon-e-2274g-processor-8m-cache-4-00-ghz.html

Stock cooler can manage to cool 4790K with tdp of 88w should also be able to cool that Xeon with a tdp of 83w.

“Should” is strong I think.   The problem with tdp ratings is that they’re fuzzy.  There is no strong and ubiquitous measurement system.  The result is the numbers are vague.  “Should” there be a strong and ubiquitous measurement system?  Yes.  The problem is there isn’t one.  The problem is more accuracy is applied to it than actually exists.

Something with a 150w tdp is going to have more cooling than something with an 80w tdp because while tdp is fuzzy it’s not THAT fuzzy, but 88w vs 83w?  The most accurate thing you could say is that they’re “somewhat similar”. The 88w device could actually have less cooling than the 83w device because of measurement system differences.  Also there is application.  The “actual” tdp of a cooler varies quite a bit as to how it is applied.  There are several heat transfer connection points, and if one of them is compromised effective functional TDP can go down.  Sometimes drastically.


Life is like a bowl of chocolates: there are all these little crinkly paper cups everywhere.

Link to post
Share on other sites
6 hours ago, Bitter said:

TDP, where the watts don't matter and the numbers are made up.

spacer.png

Binary thinking.  It’s not useless, it’s just not nearly as accurate as it claims I think.

 

the old solution was to look at a very rough measure of the cooler construction.  How many heat pipes does it have?  How much does the cooler weigh? What is The connection system between the various parts of the cooler?

This was abandoned though because it doesn’t work either.

That system can be cheated too.  The hyper 212 by this measure should have poorer cooling than it does and there are some aliexpress twin tower 6 pipe coolers that by that measure have much better cooling than they actually do.

 

TDP was an attempt to improve on inaccuracy of the old system.  It may be better, it may not be better.  It isn’t as accurate as it claims to be, which seems to be by the watt.  Neither one is even 80% accurate it seems to me.


Life is like a bowl of chocolates: there are all these little crinkly paper cups everywhere.

Link to post
Share on other sites

Problem is there are ZERO standards for CPU TDP measurement among manufacturers and ZERO standards for cooler manufacturers. That 88W Intel CPU is only 88W under some very weird specific circumstances which are different than the circumstances that make the other 83W Intel CPU 83W. AMD does it different than Intel. CoolerMaster does it different than both Intel and AMD, and everyone else does it different too. There's ZERO standardization with any of it so TDP numbers do not matter, not even within Intel's own product lines (I have no experience with AMD lately to comment). I can have two CPU's from Intel both rated at within 1W of one another. The i3 will run at over 54W and the G3460 will run at under 30W both under the same test conditions in the same board with same ancillary hardware.

https://ark.intel.com/content/www/us/en/ark/products/77480/intel-core-i3-4130-processor-3m-cache-3-40-ghz.html

https://ark.intel.com/content/www/us/en/ark/products/83428/intel-pentium-processor-g3460-3m-cache-3-50-ghz.html

The i3 routinely ran at around 65-70W and the G3460 routinely ran at around 25-30W in the same board. Intel TDP numbers mean very little in real world conditions and should be treated as strictly marketing wank until a real measurement is taken. I used both Intel XTU software as well as HwInfo to check real reported draw from the motherboard sensors. The board did not have any user adjustable CPU power options and was an Asus OEM board. Some Intel CPU's actually do run at or very near the rated TDP, I've found the low wattage options are often the closest with an i5 4590S currently running in a HP system is rated at 65W and measures a peak of 65W.

 

Also remember that as the CPU heats up it does begin to consume more power which makes it heat up more and so on in a self feeding cycle until throttle, so a poor cooling solution actually makes CPU overheating much worse in a steady load situation than just the numbers might indicate. If you've ever dabbled with passively cooling a CPU I'm sure you've ran into this situation where it seems fine but then the temp starts to creep up and power use goes up and then temp climbs faster and power use climbs faster and it runs away to throttle pretty fast. Small bursts of heavy load, as mentioned, are easily handled by a cooler which may be inadequate for sustained loads which is why a normal desktop cooler can be pretty dinky on a "88W" CPU, but in a server environment a CPU is more likely to see heavy sustained use likely surpassing that "83W" rating and swamping the cooler completely till throttle and then staying throttled and possibly also shortening the CPU and possibly the board life span.

 

TDP discussion is also interesting because once you start really looking into it, it's all pretty much BS. I miss the old synthetic heat load tests some places used to do on heat sinks because it was an even across the board test. 200W load into 20 different heat sinks at a set ambient temp all with the same fan speed and measure the thermal dissipation and fan noise. I'd love to see that testing methodology come back because it made comparing coolers very easy, I suspect that cooler manufacturers didn't care for it because it was a bit of a bullshit exposer.

 

I'm mostly curious how Intel will be handling the warranty and if there's any compensation for the loss of cooler since it was included in the original purchase price.

Link to post
Share on other sites
6 minutes ago, Bitter said:

Problem is there are ZERO standards for CPU TDP measurement among manufacturers and ZERO standards for cooler manufacturers. That 88W Intel CPU is only 88W under some very weird specific circumstances which are different than the circumstances that make the other 83W Intel CPU 83W. AMD does it different than Intel. CoolerMaster does it different than both Intel and AMD, and everyone else does it different too. There's ZERO standardization with any of it so TDP numbers do not matter, not even within Intel's own product lines (I have no experience with AMD lately to comment). I can have two CPU's from Intel both rated at within 1W of one another. The i3 will run at over 54W and the G3460 will run at under 30W both under the same test conditions in the same board with same ancillary hardware.

https://ark.intel.com/content/www/us/en/ark/products/77480/intel-core-i3-4130-processor-3m-cache-3-40-ghz.html

https://ark.intel.com/content/www/us/en/ark/products/83428/intel-pentium-processor-g3460-3m-cache-3-50-ghz.html

The i3 routinely ran at around 65-70W and the G3460 routinely ran at around 25-30W in the same board. Intel TDP numbers mean very little in real world conditions and should be treated as strictly marketing wank until a real measurement is taken. I used both Intel XTU software as well as HwInfo to check real reported draw from the motherboard sensors. The board did not have any user adjustable CPU power options and was an Asus OEM board. Some Intel CPU's actually do run at or very near the rated TDP, I've found the low wattage options are often the closest with an i5 4590S currently running in a HP system is rated at 65W and measures a peak of 65W.

 

Also remember that as the CPU heats up it does begin to consume more power which makes it heat up more and so on in a self feeding cycle until throttle, so a poor cooling solution actually makes CPU overheating much worse in a steady load situation than just the numbers might indicate. If you've ever dabbled with passively cooling a CPU I'm sure you've ran into this situation where it seems fine but then the temp starts to creep up and power use goes up and then temp climbs faster and power use climbs faster and it runs away to throttle pretty fast. Small bursts of heavy load, as mentioned, are easily handled by a cooler which may be inadequate for sustained loads which is why a normal desktop cooler can be pretty dinky on a "88W" CPU, but in a server environment a CPU is more likely to see heavy sustained use likely surpassing that "83W" rating and swamping the cooler completely till throttle and then staying throttled and possibly also shortening the CPU and possibly the board life span.

 

TDP discussion is also interesting because once you start really looking into it, it's all pretty much BS. I miss the old synthetic heat load tests some places used to do on heat sinks because it was an even across the board test. 200W load into 20 different heat sinks at a set ambient temp all with the same fan speed and measure the thermal dissipation and fan noise. I'd love to see that testing methodology come back because it made comparing coolers very easy, I suspect that cooler manufacturers didn't care for it because it was a bit of a bullshit exposer.

 

I'm mostly curious how Intel will be handling the warranty and if there's any compensation for the loss of cooler since it was included in the original purchase price.

I like the whole synthetic benchmark thing.  It was not problem free though.  If a synthetic benchmark is used companies aim to maximize that synthetic benchmark over real world performance.  Standards are hard.  Multiple benchmarks maybe?  Synthetic benchmark and “tdp” and rough cooler construction?  What irritates me is there are likely engineers within the companies who know exactly what coolers are more useful for what tasks and why.

 

Makes me want to go off on my standard rant about what is wrong with privatized science and the crippling of public science since the 1980’s


Life is like a bowl of chocolates: there are all these little crinkly paper cups everywhere.

Link to post
Share on other sites
8 minutes ago, Bombastinator said:

I like the whole synthetic benchmark thing.  It was not problem free though.  If a synthetic benchmark is used companies aim to maximize that synthetic benchmark over real world performance.  Standards are hard.  Multiple benchmarks maybe?  Synthetic benchmark and “tdp” and rough cooler construction?  What irritates me is there are likely engineers within the companies who know exactly what coolers are more useful for what tasks and why.

 

Makes me want to go off on my standard rant about what is wrong with privatized science and the crippling of public science since the 1980’s

I mean they COULD build for the synthetic rig I guess but the rig's I remember being popular (I seriously forget whom) were a controlled heating element under a heat spreader the same area as the CPU's of the time either those with IHS or bare die (if that tells you how old I'm talking about). Was it frostytech maybe? Anadtech? They had a HUGE list of like...every cooler ever ranked by capacity, temp drop, noise, etc you could scroll through and click to the review of, it was a fantastic resource.

Link to post
Share on other sites
30 minutes ago, Bitter said:

and ZERO standards for cooler manufacturers

Cooler manufacturers do actually have very good standards for measuring thermal performance of coolers and it's common across that industry, the problem is the devices they are designed to cool (CPUs/GPUs) do not use the same measurement specification nor use the TDP rating for the same intended purpose. It's a bit like measuring the horsepower at the flywheel versus at the wheel, different things telling you different information.

 

Gamers Nexus has a video on their channel where they go to a cooler manufacture and show the equipment used to measure TDP of coolers and how the process is done. While this method is scientifically accurate a CPU/GPU is not a fixed power output heating plate so if you were trying to 1 to 1 the TDPs you'll end up with a greatly over spec cooler.

 

Intel does also have detailed thermal specification documents for mobile CPUs, desktop CPUs and server CPUs. Yes how Intel rates these products with TDPs is different across these product ranges, a desktop CPU with 88W TDP is not the same as a Xeon with 88W TDP.

 

What it boils down to is that there is zero point in general for consumers to be discussing TDPs because they are, for us, useless specifications that are not understood well enough to convey meaningful information.

 

The relevant video:

 

Link to post
Share on other sites

Standards internally for their own cooler product lines or standards across different manufacturers? If I buy a car with SAE rated 200HP it has the same SAE rated 200HP whether it's from Chevy or from Nissan or from Ford, etc. If I buy a 200W rated cooler from Thermaltake is it the same 200W rating from CoolerMaster, etc? I'd really love to see both cooler and CPU (maybe even GPU?!) manufacturers sit down and all hammer out some kind of standard rating system for their products so that if you buy something that says it has a thermal output of 200W and you buy a cooler with a thermal capacity of 200W both products will be evenly matched. I'm sure this kind of thing exists in the industrial/server/professional market, I'd like to see if trickle down to the consumer level to make life easier. I doubt it ever will because marketing departments love their marketing wank.

Link to post
Share on other sites
1 minute ago, Bitter said:

Standards internally for their own cooler product lines or standards across different manufacturers? If I buy a car with SAE rated 200HP it has the same SAE rated 200HP whether it's from Chevy or from Nissan or from Ford, etc. If I buy a 200W rated cooler from Thermaltake is it the same 200W rating from CoolerMaster, etc?

The equipment and process shown in the video is standard across the industry, like I said it's however not relevant to the devices they are going to cool. CPUs aren't fixed heat output and a cooler with 200W TDP can perform worse than a 180W TDP cooler for an Intel CPU operating in Intel spec (not common on gaming MB) where by the boost duration is actually honored and the 180W TDP cooler has better heat conduction from the IHS to the cooler base plate and to the fins.

 

TDP ratings are unfortunately not fully representative of a coolers performance when placed on a CPU, that is why cooler TDPs were taken off box packaging and marketing. There just isn't a single unit of measurement that can be used to show which cooler is better than another for all applications.

Link to post
Share on other sites
34 minutes ago, leadeater said:

The equipment and process shown in the video is standard across the industry, like I said it's however not relevant to the devices they are going to cool. CPUs aren't fixed heat output and a cooler with 200W TDP can perform worse than a 180W TDP cooler for an Intel CPU operating in Intel spec (not common on gaming MB) where by the boost duration is actually honored and the 180W TDP cooler has better heat conduction from the IHS to the cooler base plate and to the fins.

 

TDP ratings are unfortunately not fully representative of a coolers performance when placed on a CPU, that is why cooler TDPs were taken off box packaging and marketing. There just isn't a single unit of measurement that can be used to show which cooler is better than another for all applications.

Seems like it would be pretty easy to fudge those test chamber numbers with a few very slight alterations in conditions but good that they're all generally using the same testing and standards, I had missed that video. I understand that a CPU isn't a fixed heat load and that not every CPU is ran the same in every motherboard either, with enthusiast boards discarding guidelines laid down for clock duration and voltage and probably even some OEM's doing so as well. I suppose I always want the absolute worst case numbers so I can build to be thermally infinite and never throttle under the most demanding loads for the longest duration.

 

I must test my stuff more harshly than most, Prime95 small torture test with system assembled for about 24hrs to check cooling and power delivery (with Furmark if GPU is included in system) and then another 24 hours of Prime95 large and small ran concurrently to check memory and continue stressing the CPU and cooling system. After that 48 hours I'll deem a system stable and ready to go if it hasn't throttled or reached max fan speeds to not throttle (well if there's a discrete graphics card I'll leave it looping some kind of benchmark on very high settings). During that 'burn in' is when I look at TDP as reported by software when it's cool and when it's warm. I don't want to see anything I assemble throttle even under the most absurd loads and even when neglected and filled with dust. People generally view PC's as a no maintenance item until something goes wrong enough for them to have problems.

 

But we're getting pretty far off topic now.

Link to post
Share on other sites
20 hours ago, leadeater said:

It'll be failing the task due to desktop/workstation usage being burst load and the server workload profile is much more sustained, meaning the cooler never gets time to dissipate extra heat above the capacity of the cooler. 5 minutes of 140W load on a 80W cooler isn't actually much of a problem, you may or may not throttle slightly but once the load drops the built up heat will dissipate and the cooler is ready to handle another peak load. On the other hand the Xeon could be sustaining 100W load on a 80W cooler, it'll throttle always forever.

Ha I'd love for him to slap that 4790k cooler on a xeon equivalent and let it  try to cool the xeon while it does it's usual tasks. There's a reason server sit in air conditioned rooms. Even with those high rpm fans they'll throttle faster than a consumer chip when they get bitch slapped with a sustained load.


CPU: Intel i7 7700K | GPU: ROG Strix GTX 1080Ti | PSU: Seasonic X-1250 (faulty) | Memory: Corsair Vengeance RGB 3200Mhz 16GB | OS Drive: Western Digital Black NVMe 250GB | Game Drive(s): Samsung 970 Evo 500GB, Hitachi 7K3000 3TB 3.5" | Motherboard: Gigabyte Z270x Gaming 7 | Case: Fractal Design Define S (No Window and modded front Panel) | Monitor(s): Dell S2716DG G-Sync 144Hz, Acer R240HY 60Hz (Dead) | Keyboard: G.SKILL RIPJAWS KM780R MX | Mouse: Steelseries Sensei 310 (Striked out parts are sold or dead, awaiting zen2 parts)

Link to post
Share on other sites
12 hours ago, Bitter said:

I mean they COULD build for the synthetic rig I guess but the rig's I remember being popular (I seriously forget whom) were a controlled heating element under a heat spreader the same area as the CPU's of the time either those with IHS or bare die (if that tells you how old I'm talking about). Was it frostytech maybe? Anadtech? They had a HUGE list of like...every cooler ever ranked by capacity, temp drop, noise, etc you could scroll through and click to the review of, it was a fantastic resource.

My memory is it was anadtech and iirc there were complaints that some manufacturers DID do just that. And it wasn’t even a standardized test.  If the synthetic score was standardized it would likely get much worse much faster.


Life is like a bowl of chocolates: there are all these little crinkly paper cups everywhere.

Link to post
Share on other sites
12 hours ago, leadeater said:

Cooler manufacturers do actually have very good standards for measuring thermal performance of coolers and it's common across that industry, the problem is the devices they are designed to cool (CPUs/GPUs) do not use the same measurement specification nor use the TDP rating for the same intended purpose. It's a bit like measuring the horsepower at the flywheel versus at the wheel, different things telling you different information.

 

Gamers Nexus has a video on their channel where they go to a cooler manufacture and show the equipment used to measure TDP of coolers and how the process is done. While this method is scientifically accurate a CPU/GPU is not a fixed power output heating plate so if you were trying to 1 to 1 the TDPs you'll end up with a greatly over spec cooler.

 

Intel does also have detailed thermal specification documents for mobile CPUs, desktop CPUs and server CPUs. Yes how Intel rates these products with TDPs is different across these product ranges, a desktop CPU with 88W TDP is not the same as a Xeon with 88W TDP.

 

What it boils down to is that there is zero point in general for consumers to be discussing TDPs because they are, for us, useless specifications that are not understood well enough to convey meaningful information.

 

The relevant video:

 

Zero is an infinite number.  I would say there is a lot less usefulness than is implied by the precision of the number.  It’s not quite the same thing.  The amount of slop is huge, but it’s not the same as zero.  A cooler with a 300w tdp is going to cool more than a cooler with a 120w tdp.  The range is large though. Real world that 120 could wind up being between (somewhat randomly) 70 and 180, whereas the 300w cooler would be more like 200-350 perhaps.  This is assuming a reasonably competent install.  Screw something up badly and those numbers get much smaller very quickly.  So not totally and completely useless, but also not nearly as accurate in practice as is implied.

 

Dirty data is not always infinitely worthless.  All models are wrong but sometimes they are still useful.


Life is like a bowl of chocolates: there are all these little crinkly paper cups everywhere.

Link to post
Share on other sites
1 hour ago, Bombastinator said:

Zero is an infinite number.  I would say there is a lot less usefulness than is implied by the precision of the number.  It’s not quite the same thing.  The amount of slop is huge, but it’s not the same as zero.  A cooler with a 300w tdp is going to cool more than a cooler with a 120w tdp.  The range is large though. Real world that 120 could wind up being between (somewhat randomly) 70 and 180, whereas the 300w cooler would be more like 200-350 perhaps.  This is assuming a reasonably competent install.  Screw something up badly and those numbers get much smaller very quickly.  So not totally and completely useless, but also not nearly as accurate in practice as is implied.

 

Dirty data is not always infinitely worthless.  All models are wrong but sometimes they are still useful.

Competing coolers aren't really going to have much different TDP capabilities and with how computers are used thermal conduction is more important and shows up in cooler reviews more than a TDP would. Heat pipes are great at moving heat but do not dissipate that heat well, a copper base-plate/slug is worse than heat pipes, an aluminium block is worse than copper however coolers made using these varying different metals and components can have the same TDP rating but tdie and tcase temperatures will be quite different in actual operation.

 

TDP doesn't account for these factors, only how much heat can be radiated from the cooler to air/water/alternate medium.

 

This is known as the too clean issue, where you remove too many variables and create a test case which can no longer be applied to the real use case or situation. This is the TDP testing of coolers. While the TDP rating of a cooler does inform you well of how good it is at moving heat from a device to air it doesn't tell you accurately how good it is at cooling a CPU/GPU. 

 

But I think you've still missed the key issue, CPU TDP is not the same measurement as a cooler TDP and never can be unless we have a system design where P States and C States do not matter or are disabled i.e. HPC 100% utilization always or 0% idle waiting for jobs.

 

It is not possible to match up TDP of a CPU/GPU to a TDP of a cooler, they aren't the same thing. It is not the TDP of coolers that is the problem and have zero value to us it's the TDP of CPUs, these and only these are useless unless you understand exactly the technical specifications used to create these (how it's measured and test case conditions). I can almost guarantee 90% of the people on this forum have no idea what the CPU TDP represents and how it is measured.

 

Edit:

If you are after long term sustained load cooling information pull out a current clamp on the EPS or use XTU and get a readout of the power draw under load, then you can start matching that up to TDP of coolers.

Link to post
Share on other sites
2 minutes ago, leadeater said:

Competing coolers aren't really going to have much different TDP capabilities and with how computers are used thermal conduction is more important and shows up in cooler reviews more than a TDP would. Heat pipes are great at moving heat but do not dissipate that heat well, a copper base-plate/slug is worse than heat pipes, an aluminium block is worse than copper however coolers made using these varying different metals and components can have the same TDP rating but tdie and tcase temperatures will be quite different in actual operation.

 

TDP doesn't account for these factors, only how much heat can be radiated from the cooler to air/water/alternate medium.

[/QUOTE] agreed



But I think you've still missed the key issue,

very possible.  Wouldn’t be the first time.  Even recently and on this forum



CPU TDP is not the same measurement as a cooler TDP and never can be unless we have a system design where P States and C States do not matter or are disabled i.e. HPC 100% utilization always or 0% idle waiting for jobs.

well of course not. It’s a question of what one is using TDP for.  If you say “I’ve got a 65w tdp chip so all I need is a 65w TDP cooler it’s almost certainly going to go quite badly.  If on the other hand you say “I’ve got a 65w tdp chip so to be reasonably safe I need a cooler with at least 100w of TDP” it’s probably going to go much better.



It is not possible to match up TDP of a CPU/GPU to a TDP of a cooler, they aren't the same thing. It is not the TDP of coolers that is the problem and have zero value to us it's the TDP of CPUs, these and only these are useless unless you understand exactly the technical specifications used to create these (how it's measured and test case conditions). I can almost guarantee 90% of the people on this forum have no idea what the CPU TDP represents and how it is measured.

ok, but weren’t we talking about coolers not CPUs?  

I suppose they are intertwined.  I was under the impression that there was a lot more slop in cooler TDP than in CPU TDP.  You claim this is not true.  I have nothing to argue with on that one.  I don’t have any information.  Do you by any chance? 



 

This is known as the too clean issue, where you remove too many variables and create a test case which can no longer be applied to the real use case or situation.

yep.


This is the TDP testing of coolers. While the TDP rating of a cooler does inform you well of how good it is at moving heat from a device to air it doesn't tell you accurate how good it is at cooling a CPU/GPU. 

because the CPU/GPU TDP suffers from “too clean”?

 


Life is like a bowl of chocolates: there are all these little crinkly paper cups everywhere.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now


×