Core i9-7900X Skylake-X Review Shows Up

[EarthDog]

A breaking point is 100C where it throttles. We say to keep away from there (by 10%). The REAL breaking point is 110C when it shuts down to protect itself. Popular, and correct, advice is to keep it under 90C while stress testing. Stress testing is, outside of F@H type applications an unrealistic operating situation. Rendering and such tends to be less, gaming WAY less... etc. Most nothing can match a P95 run (which uses the AVX extensions). AIDA64 using FPU only is a few C behind P95..

 

[Deleted member 172152]

So much this!!!

A 95W CPU running at 100* is going to put out the same heat as a 95W CPU running at 15* (assuming both are consuming 95W/at load).

(Also, 200mm fans are useless in a PC - dunno why so many TPU users are under any other impression)
While this is true, Backblaze's disks, on average, run at a maximum 31*, with a few disks running as high as 38*, so it doesn't really tell us what happens to a disk at >40*

If you look to Google's and Microsoft's results you'll still find little correlation (except at cold temperatures; see links at BB source), but you will see that drives at high temperatures fail sooner than drives at low temperatures in the case of Google, and in Microsoft's case, where drives were allowed to run closer to their operating limit, there's a more obvious correlation at high temperatures.

Actually cooler chips produce less heat and can consume less power since the resistance is lower in a cool chip than a hot one.

Also, the problem isn't necessarily the chip temps, but the TIM temps. 90c+ isn't good for cheap TIM like intel uses and you'll have to dial back an overclock or even stock clocks eventually or you'll have to void your warranty in an obvious way by delidding and replacing the TIM.

 

[trparky]

And what do you think will happen when the TIM breaks down like we have seen happen? The chip will either burn up or end up throttling itself every time it gets even a bit of load on it. The TIM will break down, we know this. Even if it's not catastrophic failure it still won't be good for the end user.

 

[mac007]

intel has its problems too..... remember DDR3 problem and cooler problem

 

[notb]

Actually cooler chips produce less heat and can consume less power since the resistance is lower in a cool chip than a hot one.

Are you aware of just how small this effect is? It's usually a lot less than 1 W/K (ΔP/ΔT).

Also, the problem isn't necessarily the chip temps, but the TIM temps. 90c+ isn't good for cheap TIM like intel uses and you'll have to dial back an overclock or even stock clocks eventually or you'll have to void your warranty in an obvious way by delidding and replacing the TIM.

Again, why do you insist that 90*C is bad for TIM or anything inside a PC? We're talking about electronics - they are really designed to handle this.

TIM is a standard solution in electronics - not just CPUs. Even for those designed to work at temperatures above 100*C.
And since you're so afraid about epoxies, ceramics and polymers working at 100*C, I really don't know how you 're even able to make tea, let alone cook or bake.

And what do you think will happen when the TIM breaks down like we have seen happen? The chip will either burn up or end up throttling itself every time it gets even a bit of load on it. The TIM will break down, we know this. Even if it's not catastrophic failure it still won't be good for the end user.

The issue here is that people really don't see this "breaking down". CPUs work perfectly well for years if not pushed over what they've been designed for.

 

[Captain_Tom]

yeah yeah whatever. Worse IPC is worse. No matter how Intel wanna spin it they failed hard.

Exactly my friend!


This is that "Secret tech Intel is hiding away in a vault for AMD" everyone was talking about. The miracle secret CPU is an overheating Broadwell part with less IPC. WOW.

 

[R0H1T]

Okay let's get this straight, the OC potential of SKLx is just a headline grabbing bullet point. In the prosumer space very rarely do people, or corporations, run (almost) server grade processors out of spec, delidding is totally out of the picture as well.

At this point in time, the base clocks & turbo speeds will determine how attractive Intel's offering is for the intended audience. TR has solder, HT, price & potentially (default) clock speeds in its favor, not to mention ECC which is a big selling point in this space.

If AMD don't mess up all of'em there's still a good reason to buy AMD, 64x PCIe 3.0 lanes & so TR should still be heck of a lot competitive regardless of what SKLx max OC is. You don't OC your production machines, that's what 99% of people won't ever do their source of livelihood.

 

[Deleted member 172152]

Are you aware of just how small this effect is? It's usually a lot less than 1 ΔP/ΔT.


Again, why do you insist that 90*C is bad for TIM or anything inside a PC? We're talking about electronics - they are really designed to handle this.

TIM is a standard solution in electronics - not just CPUs. Even for those designed to work at temperatures above 100*C.
And since you're so afraid about epoxies, ceramics and polymers working at 100*C, I really don't know how you 're even able to make tea, let alone cook or bake.


The issue here is that people really don't see this "breaking down". CPUs work perfectly well for years if not pushed over what they've been designed for.

Cheap and some expensive thermal paste isn't designed to run near 100c, it's designed to run at 80-90 max. Even nt-h1 only can handle 100-110c PEAK temps and that's a GOOD thermal paste. No way the stuff intel's using is designed to handle nigh 100c temps and it will probably even degrade relatively quickly at 80-90c. We saw that the 6700k had somewhat decent paste and the new 7700k stuff is worse, so I haven't got high hopes for the skylake-x TIM lasting too long.

 

[notb]

If AMD don't mess up all of'em there's still a good reason to buy AMD, 64x PCIe 3.0 lanes & so TR should still be heck of a lot competitive regardless of what SKLx max OC is. You don't OC your production machines, that's what 99% of people won't ever do their source of livelihood.

If you don't OC, what exactly does soldered IHS give you? Intel's TIM is not failing at factory clocks.

Cheap and some expensive thermal paste isn't designed to run near 100c, it's designed to run at 80-90 max.

Man! Where do you get all this stuff?

Even nt-h1 only can handle 100-110c PEAK temps and that's a GOOD thermal paste.

Do you really think that a consumer overpriced product for computer geeks is a "good thermal paste" overall?
What about materials used in industry and science?

Moreover, finding this took me literally 2 minutes:
https://www.lairdtech.com/products/tgrease-2500

EDIT:
I've just checked Arctic Silver: Ceramique and 5 are both good up to 130*C.https://www.lairdtech.com/products/tgrease-2500

 

[Deleted member 172152]

If you don't OC, what exactly does soldered IHS give you? Intel's TIM is not failing at factory clocks.

Man! Where do you get all this stuff?

Do you really think that a consumer overpriced product for computer geeks is a "good thermal paste" overall?
What about materials used in industry and science?

Moreover, finding this took me literally 2 minutes:
https://www.lairdtech.com/products/tgrease-2500

3.8 W/Mk!!!! Wow, that's rubbish! Even intel stuff might be better!

 

[R0H1T]

If you don't OC, what exactly does soldered IHS give you? Intel's TIM is not failing at factory clocks.

Lower temps with warranty, unless you're arguing Intel's paste is the best TIM they can afford?
Also something called VFM, though it's subjective for every other person, I'd like to think that the successor to 6950x shouldn't be any worse of a package (figuratively & literally) than it's predecessor.

 

[notb]

3.8 W/Mk!!!! Wow, that's rubbish! Even intel stuff might be better!

Hahaha :-D.
That's an industrial product. If they claim 3.8, it most likely is 3.8.
It's a bit different with consumer products. Check this report:
http://www.nrel.gov/docs/fy08osti/42972.pdf
Arctic Silver 5 - a paste praised by computer enthusiasts and reviewers - while rated at 8.7 W/mK - only managed 0.94 W/m in a controlled experiment. Another consumer product from Thermaxtech was just as awful.
Industrial-grade compound from Dow Corning (Intel's supplier) performs just as promised: 4.0 W/mK.

You can't compare specification of consumer and industry products. In consumer stuff it's usually a best-case scenario. In industry/science it's always realistic and hard conditions (including worst-case stress testing).
Just to give you an example: most MTTF figures for consumer PC fans (including CPU and GPU coolers) are at "room temperature" (usually 25*C). Think about it. ;-)

 

[EarthDog]

Actually cooler chips produce less heat and can consume less power since the resistance is lower in a cool chip than a hot one.

Also, the problem isn't necessarily the chip temps, but the TIM temps. 90c+ isn't good for cheap TIM like intel uses and you'll have to dial back an overclock or even stock clocks eventually or you'll have to void your warranty in an obvious way by delidding and replacing the TIM.

And what do you think will happen when the TIM breaks down like we have seen happen? The chip will either burn up or end up throttling itself every time it gets even a bit of load on it. The TIM will break down, we know this. Even if it's not catastrophic failure it still won't be good for the end user.

LOL, they are still going. Its amazing... its like... Sew/litwicki24 posts...

The answers are all right there, but, the same words keep getting spit out...

 

[Duality92]

The only thing I have to say is they state overclocking headroom, but this processor already turbo's at 4.5 GHz, so overclocking it to 4.7 should almost be a given. That's only a 4.4% overclock over turbo.

Ryzen, 1700 for example, turbo's to 3.7 GHz, but can be clocked to 4.1 (or 4.0 worst case). That's a 8.1% overclock (for 4.0) and 10.8%. I feel that Ryzen has better overclocking room here?

Last processors with "real" overclocking headroom was, Vishera. FX-8320, Turbo to 4.0, but clocked to 5.0, that's 25% and Sandy Bridge, 2500K, turbo to 3.7, overclocked to 5.0 GHz, 35% overclock. I don't see anything out lately going NEAR this.

*These numbers are assuming processors can hit their turbo clocks on stock voltage, which I've never seen not happen*

 

[EarthDog]

Ryzen can't get past their own XFR either...its 4/4.1... If you are going to do that, at least compare where all cores leave off (like to like) or go with XFR clocks and max turbo boost.

 

[Deleted member 172152]

Ryzen can't get past their own XFR either...its 4/4.1... If you are going to do that, at least compare where all cores leave off (like to like) or go with XFR clocks and max turbo boost.

Just use the average max speed to compare to all-core oc, although you should keep in mind i9 baseclocks are really low.

Really though we should ignore core clocks and overclocks and only look at stock/oc-ed performance and stock 7900x performance is meh, and oc-ed it's great, but who knows how long that intel TIM will last (sadly if it really does suck so much it will essentially become dry toothpaste, you'll only know after a few months-years).

 

[trparky]

The issue here is that people really don't see this "breaking down". CPUs work perfectly well for years if not pushed over what they've been designed for. ... Intel's TIM is not failing at factory clocks.

Then why are there so many people de-lidding these things? All I heard about is people de-lidding them. Lots of YouTube videos telling people how to do it. Lots of online tutorials. Specialty shops are even selling tools that do it for you, just put the CPU in, turn the crank, and you're done. Obviously lots of people are doing this and there's a high demand for knowing how to do it so there must be a problem.

Again, the TIM has been known to be absolute shit for some time now. We know this! People have taken their CPUs apart to find that the TIM is cheap shit. Again we would not be having this problem if Intel used some decent TIM like Arctic Silver, Thermal Grizzly, or any one of the hundreds of other enthusiast TIMs that one can get in the enthusiast space. But no, as we have said before, they instead chose to use the cheapest fucking garbage that they could find. It would probably cost them only one more cent to use some quality stuff under the IHS but noooooo... they have to save one cent and use garbage.

 

[efikkan]

Overclocking, even for normal enthusiasts is becoming pretty pointless these days, where the OC potential is marginal without extreme cooling, for both vendors. It's nice that we still have the option, even though it's becoming less relevant.

 

[trparky]

Then they need to get rid of the K-series of chips because the K-series has the unlocked multiplier than practically begs you to overclock the chip. People pay premium prices for these chips for the right to overclock them and then all of a sudden Intel tells us "You shouldn't overclock them" all because they were stupid idiots and cheaped out in the manufacturing of them? Well I have words for that kind of treatment... Fuck you Intel!

 

[Steevo]

If you don't OC, what exactly does soldered IHS give you? Intel's TIM is not failing at factory clocks.

Man! Where do you get all this stuff?

Do you really think that a consumer overpriced product for computer geeks is a "good thermal paste" overall?
What about materials used in industry and science?

Moreover, finding this took me literally 2 minutes:
https://www.lairdtech.com/products/tgrease-2500

EDIT:
I've just checked Arctic Silver: Ceramique and 5 are both good up to 130*C.

They are good to that temperature, but that is rated when the CPU is laying flat or horizontal, not vertical, and phase change compounds WILL tend to run out of the spaces given that it becomes a liquid at those temperatures and gravity works. Its the same reason they tell you to store Arctic Silver products in the fridge with the tip down as the mixture will settle out. Considering the temperature decrease many see with delidding there is no doubt that Intel is purposefully using poor quality TIM to reduce the life expectancy or performance over time.... gotta get the users to upgrade somehow.

 

[notb]

Then why are there so many people de-lidding these things?

Usually because they want to call themselves "enthusiasts". Being "an enthusiast" of something simply means doing pointless, insensible things.
Just to give you an example: someone doing extreme OC under liquid nitrogen is a hero for many people on this forum. However, for most population on Earth he's just a moron.

Lots of YouTube videos telling people how to do it. Lots of online tutorials. Specialty shops are even selling tools that do it for you, just put the CPU in, turn the crank, and you're done. Obviously lots of people are doing this and there's a high demand for knowing how to do it so there must be a problem.

No offence, but the way you're drawing conclusions shows just how bad is the state of humanity today. Even without mentioning YouTube this would look stupid enough.

Again, the TIM has been known to be absolute shit for some time now. We know this!

So share it with the rest.

People have taken their CPUs apart to find that the TIM is cheap shit. Again we would not be having this problem if Intel used some decent TIM like Arctic Silver, Thermal Grizzly, or any one of the hundreds of other enthusiast TIMs that one can get in the enthusiast space.

In one of my posts above you'll find a report comparing few TIMs - Arctic Silver 5 is on a list. It shows just how bad this material is - yet "computer enthusiasts" praise it (it clearly wins many comparison tests).
But if the TIM Intel uses is similar to the one from the report, it's clearly better than Arctic Silver and most likely better than anything that "enthusiasts" can buy in "enthusiast space". Surely, soldering or liquid metal will improve thermal conductivity (at a cost of other properties), but replacing what Intel has with a consumer TIM could be an awful idea.
I still hope you'll point me to a sensible research of Intel's TIM properties...

 

[trparky]

Considering the temperature decrease many see with delidding there is no doubt that Intel is purposefully using poor quality TIM to reduce the life expectancy or performance over time.... gotta get the users to upgrade somehow.

Exactly! People have delidded their processors and put various forms of enthusiast TIMs under the IHS and have seen dramatic reductions in overall processor temperatures. One person used some kind of liquid metal and was able to overclock his Core i7 far higher than he was ever able to before he delidded it and replaced the TIM.

I see this as a conspiracy to drive more sales. As many of us know, new PC sales are in the dumpster. It's the lowest we've seen in years. Nobody seems to have a need to buy new computers due to any number of variables that come into play. What better way than to purposely cheap out on the manufacturing of these processors to drive more sales. It's Corporate Bullshit 1010, it's been done for decades. Cut quality to the bare minimum to make people buy new products. You can thank the pencil pushing, know-nothing, shit for brains MBAs for this kind of corporate bullshit.

 

[Duality92]

Worst thing is, it's not really the TIM the problem, the TIM itself isn't half bad, the problem is the gap between the IHS and die, this even proved by VSG.

https://www.techpowerup.com/reviews/Intel/Delidding_the_Intel_Core_i7_7700K/

 

[Kyuuba]

Then why are there so many people de-lidding these things? All I heard about is people de-lidding them. Lots of YouTube videos telling people how to do it. Lots of online tutorials. Specialty shops are even selling tools that do it for you, just put the CPU in, turn the crank, and you're done. Obviously lots of people are doing this and there's a high demand for knowing how to do it so there must be a problem.

Again, the TIM has been known to be absolute shit for some time now. We know this! People have taken their CPUs apart to find that the TIM is cheap shit. Again we would not be having this problem if Intel used some decent TIM like Arctic Silver, Thermal Grizzly, or any one of the hundreds of other enthusiast TIMs that one can get in the enthusiast space. But no, as we have said before, they instead chose to use the cheapest fucking garbage that they could find. It would probably cost them only one more cent to use some quality stuff under the IHS but noooooo... they have to save one cent and use garbage.

If soldered you would want to OC the chip beyond its limits breaking the designed TDP it was meant for, or you are begging for a burned not only cpu but motherboard.

 

[EarthDog]

Just use the average max speed to compare to all-core oc, although you should keep in mind i9 baseclocks are really low.

Really though we should ignore core clocks and overclocks and only look at stock/oc-ed performance and stock 7900x performance is meh, and oc-ed it's great, but who knows how long that intel TIM will last (sadly if it really does suck so much it will essentially become dry toothpaste, you'll only know after a few months-years).

lol again...

If we ignore overclocking, there is ZERO point to talk about the paste...none. it even allows overclocking.
As stated ad nauseam, the paste will last throigg the useful life of the processor. Ypu guys keep thinoong its going to fall apart and outsode of VERY RARE circumstances, production flaws which has little to do with the paste, ir doesnt breakdown

The way you two describe it, people should be shitting their pants their tim is going to go bad, but its quite simply not true. You are using correlation to determine causation which, if people actually understood how things work, isnt the case.

You two are going on and on and on with absolutely zero merit... pure speculation and a lack of understanding...