i9 7900X and/or X299 Motherboard water blocks?

[Noel 4,487]

The error LEDs didn't indicate a problem w/ CPU, memory, or VGA.  I did r/o my Titan as I had an old GTX 280 that still worked.  After a few other maneuvers I eventually got the thing to beep, and the beep pattern was one continuous followed by 4 short beeps, and that equates to 'Hardware Component Failure', whatever that means.  The 3930K has been flawless and runs cool, and memory is always stock volts and rated frequency, so I am guessing it's something in the mobo.  The HX850 Corsair I don't think I can blame at only 4.3y of flawless use, run off of a APC IR-1200, but perhaps it's still in the mix.  The mainboard and CPU were the most stressed components in the system and it has a lot of 4.42Ghz hours on it, so I have to think it's likely either the CPU or mainboard.   The CPU was subjected to 1.32v for half of its 4.3y of use, or 1.29 for the other half.

With that I think it's time to start putting together a new box so yes I look forward to what you learn w/ your new i9-7900X machine.  Initially I will use my GTX Titan in a fresh install of V3.4 so I can get a better sense of what the new CPU offers.  And then pick up 1080Ti or Titan Xp and call it good hopefully for another 4-5y.  I'll likely skip the overclock if I can get 20% better CPU performance just on turbo 3, 4.5Ghz at stock voltage over my current 3930K at 4.42Ghz.  Apparently that's doable w/ air cooling, the 4.5Ghz.  

 

[Sartanius 399]

4 hours ago, Rob Ainscough said:

Just a quick update, my 7900X shipped and will arrive Tuesday, the Asus PRIME X299 Deluxe MB will arrive Wednesday, and G.Skill DDR 32GB 4266 Wednesday also.  

I've heard good and bad regarding the 7900X in it's default mode "Auto" (everything) it out performs the 6950X in both single thread and multi-thread operation.  I've read about no problems reaching 5Ghz but again no actual details on exactly how that was done and what BIOS/EFI settings were used ... have you ever wondered why some people who claim higher GHz never actually publish their settings (all of them)?  I publish my EFI/BIOS settings for others to use, but it does raise BIG doubts why others claiming to be "experts" don't ... they just show frequency results and "look it booted" (for 30 seconds).  Anyway, once I get the 7900X up and running and spend some time reaching my OC goals, I'll post some detail results and settings.

But like I said before, I have fairly low expectations it will perform that much better than my existing 5960X @ 4.6Ghz ... but who knows, maybe I'll be pleasantly surprised, trying to stay optimistic.  Objective is to have all 10 real cores (HT disabled) running at 5Ghz.

Great to hear you snagged a 7700X.  Looking forward to your testing.  I have a lesser processor (5930K) than you currently run, so any of the improvement you realize would be even greater for my installation.

I'm still reading a good bit about Coffee Lake, how it will feature 6c12t at the base end, run on current m/b's (i.e., 1151), and likely be able to O/C to greater than 5Ghz.  No one knows for sure right now, but I'm patient enough at this stage to see what it has to offer before jumping on the 7700x.

Thanks again for all the effort you put in to testing and sharing your experiences.

[Noel 4,487]

On 6/19/2017 at 2:14 PM, Rob Ainscough said:

I expect very minor/marginal improvements in terms of FPS over my 5960X so my decision points are:

1.  Less heat
2.  Looks to be relatively easy to hit 5Ghz and is very OC friendly
3.  Improved chipset (X299) with slightly improved bandwidth and lower latency
4.  Less expensive than the 6950X and seems to perform single thread faster with OC

Cheers, Rob.

This article suggests cooling requirements are steep:

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-11.html

[Guest]

4 minutes ago, Noel said:

This article suggests cooling requirements are steep:

Most likely because they're using Prime95 for testing ... I believe ASUS recommend NOT using Prime95 and I agree with them.  I've stopped using Prime95 (I used to use it for testing) as it doesn't really emulate real world application usage.  I'd be more inclined to use Adobe CC 2017 and a complex render project that will use up all my Cores or just use a good old P3D V4 test loop I know will be very hard on my CPU/GPUs, Orbx NorCal or SoCal with FB/FSDT comes to mind.

But will just have to wait and see ... my dual loop setup is pretty efficient as I rarely see water temps above 27C and CPU temps even on "low to normal" fan speeds peaking around 54C (average around 44C) under P3D V4 extreme loads.  I do like the idea of a liquid chiller though as it has re-use potential.  But wonder how the 18/36 SkylakeX processors are gonna do if the 10/20 heats up that much under Prime95?

Cheers, Rob.

[Noel 4,487]

1 hour ago, Rob Ainscough said:

Most likely because they're using Prime95 for testing ... I believe ASUS recommend NOT using Prime95 and I agree with them.  I've stopped using Prime95 (I used to use it for testing) as it doesn't really emulate real world application usage.  I'd be more inclined to use Adobe CC 2017 and a complex render project that will use up all my Cores or just use a good old P3D V4 test loop I know will be very hard on my CPU/GPUs, Orbx NorCal or SoCal with FB/FSDT comes to mind.

But will just have to wait and see ... my dual loop setup is pretty efficient as I rarely see water temps above 27C and CPU temps even on "low to normal" fan speeds peaking around 54C (average around 44C) under P3D V4 extreme loads.  I do like the idea of a liquid chiller though as it has re-use potential.  But wonder how the 18/36 SkylakeX processors are gonna do if the 10/20 heats up that much under Prime95?

Cheers, Rob.

Rob, can you recommend a water cooler for me for SkylakeX?

That fits with what I've seen which is 130W while flying, i.e. that realworld test is a much more valid evaluation of what's needed for P3D or XP.  I'm fully out of order now so have to decide which route to go for CPU.  Looks like regardless of heat the 7900X is a serious processor.  I might also consider 7820X as well.  Beyond more cores, anything significantly different between the two that will materially affect our simming experience?

[Sethos 4,303]

13 hours ago, Rob Ainscough said:

have you ever wondered why some people who claim higher GHz never actually publish their settings (all of them)?  I publish my EFI/BIOS settings for others to use, but it does raise BIG doubts why others claiming to be "experts" don't ... they just show frequency results and "look it booted" (for 30 seconds).

Overclocking is extremely individual based on the silicone, RAM and the motherboard. Each overclock needs to be specifically tailored to your build, so there's no point in showing how you got there. Some might get there with a simple ratio change, plus some added voltage. Some might need BCLK overclocking and others will be doing it via XMP with custom offset voltage values. Overclocking is fairly straight forward in the first place, not much sense in copying others settings. It's just a time consuming process of finding your own stable values.

[Piotr007 164]

8 hours ago, dmiannay said:

Great to hear you snagged a 7700X.  Looking forward to your testing.  I have a lesser processor (5930K) than you currently run, so any of the improvement you realize would be even greater for my installation.

I'm still reading a good bit about Coffee Lake, how it will feature 6c12t at the base end, run on current m/b's (i.e., 1151), and likely be able to O/C to greater than 5Ghz.  No one knows for sure right now, but I'm patient enough at this stage to see what it has to offer before jumping on the 7700x.

Thanks again for all the effort you put in to testing and sharing your experiences.

True,

I am eagerly waiting for the CL chip too. I am already running a "nice" 6700K on my LGA1151 motherboard. But this tends to be insufficient for all my P3D V4 flights. I am hitting the 100% marker on all cores with this CPU over KJFK with numerous addons. The fun fact is that is is not constantly 100% but it has a load moment of 100% CPU usage (2 seconds or so) but this is enough to cause stutters. A 6-core CL chip at 5,0ghz could give more head space and not fill up all those graphs with 100% load. Otherwise the sim grinds to a halt each 5 seconds or so at multiple busy airport locations.

[Noel 4,487]

On 6/21/2017 at 7:27 PM, Driverab330 said:

Numerous reviews had mentioned the diffitulty they had experienced on cooling the 7900x which was partly due to the fact that thermal paste is used instead of soldered tin between the processor die and the heat spreader. 

I give, why on earth would Intel create and market a HEDT CPU, then use ordinary thermal paste IF IT REALLY MAKES A DIFFERENCE?   The 2 explanations I can imagine are:

1.  It really makes marginal difference--TIM v soldered tin

2.  The cost is significantly more to use soldered tin:  NOT!  the bloody 7900X costs $999, and a little solder makes any kind a meaningful dent in production costs?  NOT!

I have to think it's mostly #1.  What is the real increase in headroom post delid given the same cooling solution?

[Guest]

Hey Noel,

If I were to hazard a guess and from what I've seen of my own experience, you actually get more surface contact (for heat transfer) when you leave the chosen material alone ... some surface "roughness" is better for adhesion when combined with thermal paste.  Where you find the most variance in cooling will be how much thermal material is applied and a well designed the compression system that provides even distribution of contact.   Too much torque can be worse then too little torque.  So I'll suggest  #1 in your list is the answer along with other variables outlined above.

I think these elements provide far more variance cooling:

1.  Too much or Too little Thermal paste (Too much is often the case for many thinking it's safer when in fact it's really not)
2.  The Clamping system used
3.  Getting correct Torque specs (not too much and not too little)

A better CPU layout would be circular with a hole in the center (more like a donut) so that clamping forces can be applied in the inner as well as the outer perimeter.  And even the 3 above case, I suspect one would have to really mess it up to see any significant difference.

Cheers, Rob.

EDIT: Motherboard and water block (just need the block out of the A240 not the rest of it as don't see how one could cool anything with that tiny radiator) arrived today so I guess I'll be busy this weekend.

[Noel 4,487]

Thank you Rob.   Yes, should be fascinating to hear how it goes.

If you will, if I end up going 7820X or 7900X, what exact cooler parts do you recommend?  It sounds like the block from this A240, but a different radiator part?  

Can one run the closed loop immersed in for example a party cooler full of ice and ice water?  It's not very elegant but seems like it could be really effective, no?  Take a nice good sized picnic cooler, drill an in and out hole in the lid and submerge the loop  all the way on the bottom and fill it with ice and water.  Could last a while!  I guess the worry is condensation dripping off the tubing, but that does't sound technically terribly challenging.  Maybe that's what custom loops are?

 

[Tom_L 522]

Hey Rob,

should you run into problems while overclocking, you might want to consider cooling the VRM as there seems to be an issue with nearly all current Mobos (according to this video). Looking forward to your expierence as I am (was?) also considering an upgrade to the X299 platform.

Rgds

Tom

[Guest]

3 hours ago, Tom_L said:

you might want to consider cooling the VRM

Yes, that is in my plan, either with a monoblock or dedicated MOFSET, but currently no one has anything available for the ASUS X299.  So for now it's CPU water cooling only.

I did see the "der8auer" video, and it looks like he just tested with a Gigabyte motherboard (not a brand I've had any great OC success with) and then claimed ALL motherboards are the same with just a single 8pin 12V MB pwr connector ... as I sit looking at my ASUS PRIME X299 Deluxe I have in my hands that has 8 pin and another 4 pin (both 12V).  So his information is not accurate and/or he didn't really test/look at the other Motherboard options or maybe he was working with Pre-release boards that weren't finalized products?  Or maybe AMD are subsidizing him ;)

13 hours ago, Noel said:

If you will, if I end up going 7820X or 7900X, what exact cooler parts do you recommend?

I haven't "Finalized" my cooling yet, but my current 5960X is dual loop, CPU loop is an EK-DCP 4.0 PWM Pump and an EK 480 XE Radiator (60mm thick) and that's what I'll be using for the 7900X.  A monoblock/MOFSET and water chiller unit is certainly on my near future radar.

On the AMD Threadripper front, competition is good, if AMD deliver the goods and it's faster than the 7900X, I would not hesitate to build an AMD based PC (I'm NOT really a brand loyalty kinda person, I go with what works the best).

Cheers, Rob.

[Noel 4,487]

Nice, thanks Rob.  I think I will look at water block coolers this time if nothing else to get the weight off of the mainboard compared to the DH-14 approach that got me here.  I'm hoping Asus can repair my old mainboard to give me a simple path towards waiting for the dust to clear on highest end processors.

This article says Intel didn't solder SkylakeX but uses TIM, which they say is a departure from prior (presumably performance-oriented CPUs) CPUs.  They also say this is a very good thing for the overclocking uber-enthusiast, but generally a bad thing for 'layman and enthusiasts'.   I could be tempted to pay for a professionally de-lidded SkylakeX just out of principle unless the entire de-lidding thing is ultimately hardly meaningful.  Presumably the material they use is some kind of liquid metal rather than solder per se I would think, and those two might not be fully equal in terms of heat conductivity between IHS and die.

http://wccftech.com/intels-core-x-series-skylake-x-and-kabylake-x-will-not-be-using-soldered-integrated-heat-spreader-ihs/http://wccftech.com/intels-core-x-series-skylake-x-and-kabylake-x-will-not-be-using-soldered-integrated-heat-spreader-ihs/

[Guest]

Hi Noel,

EK got back to me with their future plans for water cooling, looks like they'll provide a MonoBlock for the ASUS ROG Rampage VI Extreme and the ASUS ROG Strix X299-E Gaming, but NOT for my current ASUS PRIME X299 Deluxe.  Not great news, as it means I'll probably swap out my PRIME X299 for the Rampage VI Extreme when available ... probably do the Koolance 800W water chiller at the same time (provided it's not too noisy).

Anyway, I'm currently GPU limited (Titan XP) so I don't expect my 7900X to help much until nVidia release their next gen GPU end of year or early 2018.

I wish SLI would be more beneficial for P3D V4, but until LM move to DX12 EMA with SFR (not confirming this will happen, just my hope), we're not going to see much benefit from SLI (a little, but not much).

You have my take on de-lidding, it's useless IMHO and more likely to hamper heat transfer rather than improve it ... think in terms of  contact surface area, de-lidding actually reduces contact surface area. The tiny irregularities of the surface area (which de-lidding reduces) combined with the correct amount and type of thermal paste will provide more contact surface area and hence better heat transfer.  But ultimately you're not looking at anything "significant" and the measurement variances are more likely to come from other factors.

 If you really want to cool the CPU with minimal hassle then water chiller route or do a lot of MB prep and go with a Peltier.  Water chiller route is the least hassle and avoids condensation issues.  

I've done Peltier builds before and the heat and power requirements and MB prep just weren't worth the extra 400Mhz over a good water cooling only setup.

Cheers, Rob.

[supersym 52]

22 minutes ago, Rob Ainscough said:

 

 I'm currently GPU limited (Titan XP) so I don't expect my 7900X to help much

Hi,

Is that the case with heavy scenery too ? like New york X 2 (from D Design) for instance 

Because I have the same configuration like yours and It's really not my GPU with is my limit...