Who's buying Skylake-X? (You may now change your vote)

[Wyrm]

I think this whole VRM thing is overblown. Results are showing that most boards are doing just fine, including my own TUF Mark 1.
As long as there is 8+4 power on the board for CPU power, there is no reason to avoid this platform. After I return from my vacation I will run some benches and stress tests myself.

Der8auer used a different chip 7900X with all cores running at 4.6GHz.

 

[Archer27]

Der8auer used a different chip 7900X with all cores running at 4.6GHz.

Right, that's correct. But he also only tested a couple of boards. The VRMs themselves are not the problem. The heatsink designs and armor can be depending on the board.
Yes a 10 core will certainly push them harder than my 6 core but the TUF Mark 1 has much better VRM design than the Aurous and the Prime boards.

 

[Archer27]

Lets keep in mind these are the boards he tested ^^
And two of these boards only have a single 8 pin CPU connector.

 

[DrMrLordX]

63C ain't bad. Looks like the 8c and 6c parts are doing okay at least.

 

[coercitiv]

But he also only tested a couple of boards. The VRMs themselves are not the problem. The heatsink designs and armor can be depending on the board.

He did mention this 2 times in the video, the VRMs are not the problem, the heatsinks are.

People just got stuck with the idea der8auer is dissing Intel and the X299 platform as a whole, so we had an immediate push back effect in the forums. None of the messages in that video came through, except those that could easily be invalidated. (like the faulty wire temp measurement)

Of course, people using the video to discredit the platform as a whole did not help either.

 

[jmelgaard]

He did mention this 2 times in the video, the VRMs are not the problem, the heatsinks are.

People just got stuck with the idea der8auer is dissing Intel and the X299 platform as a whole, so we had an immediate push back effect in the forums. None of the messages in that video came through, except those that could easily be invalidated. (like the faulty wire temp measurement)

Of course, people using the video to discredit the platform as a whole did not help either.

Well when you give you video the title "The X299 VRM Disaster" ... Then you have kind of already set out on what the focus is, so perhaps that is partly his own fault...

It is a shame that it hasn't evolved in a constructive way though, and instead ended in people throwing mud back and forth, because I don't doubt what der8auer is saying, nor whats Tom is saying...

So considering that even though Tom (He did not remove the heastsinks either), tested multiple boards, tried removing airflow etc... All in all tried to reproduce the same results, and didn't even get close, then the question stands... Why did der8auer experience what he did?...

Blame is irrelevant, cause is important.

 

[Timmah!]

Quick question, i guess i am about to order stuff, the thing is, if i am not going to delid, should i just stick with the regular TIM (i have unused "syringe" of arctic MX4 at home) between the CPU IHS and AIO pump heatsink, or should i get something like Conductonaut? Or is that pointless, if i am not delidding, as the heat transport issue is between the die and IHS anyway?

If i should go with liquid metal, how much is needed? There is 5g syringe for almost 40 EUROs :-O And then 1g alternative? I guess 1g is not enough, but 5g seems to be too much?

 

[Arachnotronic]

Quick question, i guess i am about to order stuff, the thing is, if i am not going to delid, should i just stick with the regular TIM (i have unused "syringe" of arctic MX4 at home) between the CPU IHS and AIO pump heatsink, or should i get something like Conductonaut? Or is that pointless, if i am not delidding, as the heat transport issue is between the die and IHS anyway?

If i should go with liquid metal, how much is needed? There is 5g syringe for almost 40 EUROs :-O And then 1g alternative? I guess 1g is not enough, but 5g seems to be too much?

Don't bother delidding, IMO. It's risky, voids your warranty, and doesn't buy you that much performance.

The 7820X is my sig is running with a pretty low-cost AIO (Liquid Freezer 120) in a case with reasonable (though not great) airflow, and performance is excellent.

A few hundred more MHz would be nice, but...meh. I like knowing that I'll be able to resell the chip for a good amount of $ once a new shiny comes along that I want to upgrade to...

 

[ManyThreads]

I think this whole VRM thing is overblown. Results are showing that most boards are doing just fine, including my own TUF Mark 1.
As long as there is 8+4 power on the board for CPU power, there is no reason to avoid this platform. After I return from my vacation I will run some benches and stress tests myself.

How are you liking your TUF 1? There is so little info available for it, not many people seem to have bought it (yet). I haven't been able to find any VRM details on it either other than it's a 8+2 phase.

 

[Timmah!]

Don't bother delidding, IMO. It's risky, voids your warranty, and doesn't buy you that much performance.

The 7820X is my sig is running with a pretty low-cost AIO (Liquid Freezer 120) in a case with reasonable (though not great) airflow, and performance is excellent.

A few hundred more MHz would be nice, but...meh. I like knowing that I'll be able to resell the chip for a good amount of $ once a new shiny comes along that I want to upgrade to...

Thanks, i wont be delidding for sure. The question still stands though, Arctic MX4 (or some other, better non-conductive TIM alternatively, OR liquid metal - between IHS and cooler

BTW i will probably go for your board, as you suggested, pro carbon, so i will eventually be all over you in a dire need of advice which sliders to move in the BIOS - i guess i will be able to OC the core on my own, but not mesh and RAM (cause i have 2 32GB packs, causing instability on my current system just by activating XMP, requring manual OC). So i hope you wont not mind if later i will harass you with questions a little bit.

 

[UsandThem]

Thanks, i wont be delidding for sure. The question still stands though, Arctic MX4 (or some other, better non-conductive TIM alternatively, OR liquid metal - between IHS and cooler

Really not much a difference any more like there used to be 15 years ago with Artic Silver was above their competition. I use the included Noctua paste that came with my coolers, and it has been great.

One thing of note though, depending on how old the MX4 is, you might want to toss it and buy a new tube of something. It loses some performance as it ages, and if it's older than say 4-5 years, it's worth the <$10 to get a new tube.

http://overclocking.guide/thermal-p...ted-with-air-cooling-and-liquid-nitrogen-ln2/

 

[Timmah!]

Really not much a difference any more like there used to be 15 years ago with Artic Silver was above their competition. I use the included Noctua paste that came with my coolers, and it has been great.

One thing of note though, depending on how old the MX4 is, you might want to toss it and buy a new tube of something. It loses some performance as it ages, and if it's older than say 4-5 years, it's worth the <$10 to get a new tube.

http://overclocking.guide/thermal-p...ted-with-air-cooling-and-liquid-nitrogen-ln2/

I got it 10 months ago for free when i got parts for my X99 machine. There is copyright 2015, all rights reserved on the back side of the package.

Anyway, its not really that much worse than Conductonaut?

 

[UsandThem]

I got it 10 months ago for free when i got parts for my X99 machine. There is copyright 2015, all rights reserved on the back side of the package.

Anyway, its not really that much worse than Conductonaut?

I think they had the MX4 listed in the result chart in the link provided, even testing several liquid metals. You actually looked at that, right? If not, let me link directly to the results and recommendation page:

http://overclocking.guide/thermal-p...d-with-air-cooling-and-liquid-nitrogen-ln2/6/

 

[Timmah!]

I think they had the MX4 listed in the result chart in the link provided, even testing several liquid metals. You actually looked at that, right? If not, let me link directly to the results and recommendation page:

http://overclocking.guide/thermal-p...d-with-air-cooling-and-liquid-nitrogen-ln2/6/

Nope, i did not see that, thank you! So just about 2C between liquid metal and regular TIM? Its settled then.

 

[DrMrLordX]

Quick question, i guess i am about to order stuff, the thing is, if i am not going to delid, should i just stick with the regular TIM (i have unused "syringe" of arctic MX4 at home) between the CPU IHS and AIO pump heatsink, or should i get something like Conductonaut? Or is that pointless, if i am not delidding, as the heat transport issue is between the die and IHS anyway?

If i should go with liquid metal, how much is needed? There is 5g syringe for almost 40 EUROs :-O And then 1g alternative? I guess 1g is not enough, but 5g seems to be too much?

First off, don't use any liquid metal TIM if aluminum is involved.

Personally I swear by stuff like Conductonaut and CLU. I've used both, and they both work nicely. But I have only ever used that stuff on soldered chips and/or chips that underwent a delid/relid operation. Using great TIM between the IHS and cooler will only affect IHS surface temp. That temp delta between IHS surface and die will still be there regardless.

 

[Ajay]

Lets keep in mind these are the boards he tested ^^
And two of these boards only have a single 8 pin CPU connector.

Ah, der8auer's test results (graphed by extremetech) which have already been debunked due to a variety of errors.
Please link the article next time. Terrible click-bait article here -> https://www.extremetech.com/computi...mperatures-disaster-limit-skylake-x-potential
OFN!

 

[Wyrm]

Ah, der8auer's test results (graphed by extremetech) which have already been debunked due to a variety of errors.
Please link the article next time. Terrible click-bait article here -> https://www.extremetech.com/computi...mperatures-disaster-limit-skylake-x-potential
OFN!

 

[.vodka]

So he basically shut everyone up who "debunked" his initial testing. Very well worth the watch.

Those CPUs are power hogs and will throttle if not tweaked for them to specifically not to. When not throttling power consumption is up at the 300-400w range as already known, depending on the load. OC3D can now replicate his results. ASUS boards throttle the VRM at 105°C, Gigabyte does it at 115°C.

P95 27.1 is pre Haswell... I wonder what those numbers would be using 28.3 with FMA+AVX2 or 29.2 with AVX512 support.

VRM cooling on current boards just isn't enough for the worst case scenario and the 10 core part, if not the 8 core model pushed hard enough. Again, those upcoming 12-18c models when overclocked, will result in some pretty interesting posts across enthusiast boards

 

[ZGR]

He addresses all the questions I originally had. Glad to see he addresses these concerns. I love all the info about throttling. I could talk throttling all day!
The VRM sensor readings is quite interesting and I never thought about how different vendors will place the sensors in different areas.

edit: Ninja'd

 

[moonbogg]

And that's why he's a credible source. Everyone thought he was wrong, but he was simply aware of many things that no one else was. So it appears the issues with the platform may still stand. Simply waiting a couple months is probably the smart move.
That's a shame though because Intel spent over $100 million rushing these chips out and all that was for nothing since many people will likely wait on a purchase now anyway.
I still doubt that existing boards will be robust enough for the higher core count chips, especially 16-18 core variants, even with the VRM cooling issues being improved.

 

[maddie]

Well that was interesting. The following pages of posts should be interesting to read.

 

[beginner99]

Those CPUs are power hogs and will throttle if not tweaked for them to specifically not to.

That was the most interesting fact. As far as I understood if you leave a certain voltage (vcc-in?, called digi+ or something in Asus bios) on auto the cpu can't pull enough power and will perform slower than it should but all tools will still show the correct multiplier and frequency and you won't notice it unless you measure power use or the actually performance.

This is the most troubling aspect. You think the chip is stable and overclocked correctly yet it performs subpar.

The VRM issue is more problematic due to longevity. High VRM temps even at stock or low OCs means lower lifespan and he shows the VRM cooling on an mainstream board (don't z170 or z270 don't remember) that support max. 4 core CPUs that in extreme cases pull 150W, less than a 7900x at stock. Yet the VRM heatsinks are massive, probably 3-times the size than on these x299 boards. No wonder they overheat.

 

[krumme]

The nv fermi 480 gtx had tantalum capacitors. Extremely expensive stuff but it can stand high temperatures in the 110c area and still have a very high lifespan. The gpu was build for high temp.

Its horrendeous those boards and cpu doesnt seem to be. Lifespan is a genuine concern. The trotling is one problem but with gigabyte setting the limit at 115c there will be s stress in the environment where even solid high quality 105c capacitors is subjected to a heat that can potentially shorten the lifespan on those mb. You dont want your 105c capacitors subjected to more than 90c.

Man its scary and surpricing he was right. Looked in hindsight the explanations about ps and cables was also weak sauce. Talk about applying meaning backwards.

 

[tamz_msc]

Would like to see some tests with AVX workloads in BOINC.

It's also surprising that 'enthusiasts' used to assume a large size P95 FFT as the standard way of stress-testing the CPU. It's about time they actually learn about the algorithms that their programs use - increased time complexity with larger sizes may not stress the CPU adequately, and then there's the fact that non-power-of-2 FFTs are less efficient. Both of these things are reflected in those graphs.

 

[DrMrLordX]

502W? Holy crap!