i9 7900X and/or X299 Motherboard water blocks?

[Guest]

I probably should clarify, for the graphics settings I use to keep 30 FPS the adjustments I make are to keep the GPU below 100% load.

DD's NYC is actually not a location that will induce maximum CPU loads, Orbx SoCal with FSDT KLAX is my current hot spot that will produce maximum CPU loads if I max out my AG sliders.

But, I find it really easy to manage my CPU load based on location and add-ons, AG density and shadow option reduction while keeping a good draw distance for AG and leaving high res terrain on.  V4 can still look really good even with "normal" AG density.  So it's fairly easy for me to manage CPU side, but 4XSSAA and Dynamic Lights will be a problem for my GPU regardless of CPU load.

Cheers, Rob.

[supersym 52]

Oh I see 

I don't use dynamic light because of bugs with a few france scenery  "home made" by some people. And I fly a few with orbx (North california, san Fransisco...) So I spoke about NY because it's very known, but clearly  it's not the most heavy...

For me the most heavy location are some France aera with France VFR 3DA. There,  my CPU is hard working but my GPU use only 50 % from all his power...(4k 4XSSAA)

Thanks for clarify

[Noel 4,501]

54 minutes ago, Rob Ainscough said:

I probably should clarify, for the graphics settings I use to keep 30 FPS the adjustments I make are to keep the GPU below 100% load....

By trial and error this is what I do to a tee:  keep the GPU somewhere not higher than 90% or so to give a little headroom, and modulate scenery sliders to keep the CPU from bogging down too much depending on where I am.   Sticking w/ vsync to my screen's 30mHz refresh rate delivers perfection in video smoothness--except I have to stop flying here and there to keep it that way!

I asked this question a while back:  but it seems to me the GPU side is begging for a realtime modulator where the user sets up priority on what gets dialed back or bumped up according to that moment's GPU load.  I know redraws don't need to happen as they do when changing at least some scenery sliders, for example you can change shadow draw distances w/o redrawing the entire scene.  So to have something that KEEP the GPU at around 95% by modulating sliders affecting GPU load sounds really useful.  I don't know how many of the controls affecting CPU load might also be modified during runtime but these two combined truly would be a utility everyone would greatly benefit from.  How many times do you have to stop a flight to make a setting change so to obviate that could be a huge plus.  

You might know how possible this is to develop?

[supersym 52]

hello Rob,

I guess you're meeting a few issues...

anything serious ?

 

[Guest]

5 hours ago, supersym said:

I guess you're meeting a few issues...

No issues so far, except time ... it's up and running, OS installed (Win 10 Pro 1703) ... now it's time to OC then re-install all my FS products (I have a lot of them, DCS, XPlane, FSW, AF2, P3DV4) and their respective add-ons, etc. etc. 

I'll probably not be in a position to start performance testing until the following weekend (if work or something else doesn't come up).

I've also started the slow process of transitioning information to my P3DWiki web site which has been a time consumer.

Cheers, Rob.

[supersym 52]

3 minutes ago, Rob Ainscough said:

 except time ...

I know what you mean 

 

4 minutes ago, Rob Ainscough said:

No issues so far,

Nice to know that.

So take the time you need for a good job 

[Noel 4,501]

On 6/29/2017 at 7:39 AM, Rob Ainscough said:

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.

Cheers, Rob.

Have you for sure ruled out my cooling solution approach?  While it's not exactly quiet, it offers serious value in other areas:

1.  Cools the whole room as desired, whereas as a chiller must be sending its heat into the room, unless you can isolate the chiller in another location.  

2. Cools the entire PC:  GPU, entire mainboard, memory, everything!

3. Works with all builds you will do going forward w/ no need to install parts into your next build.

4.  The noise it makes is ultimately white noise and if you want to be completely free of it you can use noise-cancellation headphones.  But here's another reality I find:  when was the last time you were in an airplane and it was all nice and quiet?  Not!  By the time I turn up the volume of audio for flying I can't hear my cooling solution any longer--plus, it's really a low pitched white noise, so fits just fine w/ flying.  I consider it the recirculation pump for my pressurized cabins :o)

5.  Is far less expensive than a chiller!  

It's simply a low BTU through-the-wall a/c unit that has 3 speeds w/ A/C w/ temp adjustments, and 3 fan-only speeds.  I live up near Chico in Paradise, where it's very warm right now, so A/C all summer works well.  With my Noctua/3930K combo during cool winters up here I would only use fan only which pulls in outside air so no need for A/C.  I've used this approach now for about 14 years or so--same A/C unit made by Sharp, and it was about $225 then.  I installed it myself right behind and about 8" from the back of the PC case which has the big 6" fan in the back panel--it's a Coolermaster HFA X case.  When I monitor CPU core temps w/ the a/c running at peak sim load it's easy to see them get into the low 50's, however I never run the A/C to maximum low it just never needs it, and that is w/ a 3930K at 4.42Ghz on air.  Also, in my playroom with a PC simming along there is considerable heat into the room--the a/c unit moves the heat thru the wall.   I'm shocked more folks haven't used this approach.  For my next build I may go water cooler unless the Noctua DH-14 will work on a 7820X in which case I will attach the radiator directly in the flow of the a/c unit.

 

[Guest]

On 6/28/2017 at 6:26 PM, Noel said:

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?

Do you mean this?  I honestly didn't think you were being serious, sorry.  If so, that's going to require hourly maintenance and a very good giant sized ice maker and you shouldn't run plain water in loop or else you'll acquire algae very quickly.  You'll want to run some type of extremely toxic coolant.

I've thought about attaching several Peltier to the radiator but that needs a separate power supply and more cooling on the "hot" side of the Peltier and will likely produce condensation which needs to be addressed but since you can place a radiator external or away from sensitive electronics it's doable.

Cheers, Rob.

 

[Noel 4,501]

3 hours ago, Rob Ainscough said:

Do you mean this?  I honestly didn't think you were being serious, sorry.  If so, that's going to require hourly maintenance and a very good giant sized ice maker and you shouldn't run plain water in loop or else you'll acquire algae very quickly.  You'll want to run some type of extremely toxic coolant.

I've thought about attaching several Peltier to the radiator but that needs a separate power supply and more cooling on the "hot" side of the Peltier and will likely produce condensation which needs to be addressed but since you can place a radiator external or away from sensitive electronics it's doable.

Cheers, Rob.

 

Not really I just wondered if the tubing to the radiator and the radiator itself could be submerged in an ice water bath, and how long that would last with one full charge, for example fill up a large cooler full of ice and a bit of water on the bottom.  

Quite frankly I'm disillusioned w/ the cooling requirements of these chips now seems like we've gone backwards what with all the focus on making them more efficient, low-power.  My actual cooling support concept I described above is a complete no-brainer for me as it solves more issues than cooling one or two components, plus the room stays just right.  Amazed I don't hear of others doing this simple but effective solution.  I might go cpu water block just to get the weight off the mainboard as now the DH-14 puts a log of weight on the mainboard and socket since it hangs off sideways.  

[Guest]

QUICK UPDATE:

I am having some issues with the ASUS PRIME X299 Deluxe motherboard which is making OC difficult. I also can't "Shutdown" my PC ... I hit the Power off button or via Win10 "shutdown" and the PC will shutdown, then about 3-5 seconds later will power up again.  The only work around is to shutdown, then just as it goes off, hit the PSU power switch to OFF before it boots back up again ... this is a new "First" for me, after building PC's for 30+ years I've never encountered this specific problem of not being able to turn my PC OFF.

I've got a support ticket out with ASUS.

I have several other issues with the motherboard also:
1. Set the EZ-XMP switch and system will not boot
2. Enable over voltage on CPU and my system will not boot (just enable via jumper but no changes in BIOS)
3. Set the XMP profile in EFI/BIOS and my system will not boot
4. Using EZFlash to 0402 BIOS/EFI - did not boot after successful flash and required CLR CMOS

So far I can't get a good OC and no XMP support so "Base" Auto everything is all I can really do. I'll run a baseline with NO OC later this week.

I expected bleeding edge and I found it ... to be honest, this ASUS PRIME X299 DELUXE motherboard seems VERY rushed out the door and full of some very basic bugs/issues (even in the BIOS/EFI user interface it's slow and unresponsive with dropdown choice popping up/down in the wrong direction and going out of window so I can't see the options). I may end up returning the motherboard and waiting for the ASUS ROG X299 EXTREME motherboard or something else?

Cheers, Rob.

[Noel 4,501]

6 hours ago, Rob Ainscough said:

I expected bleeding edge and I found it ... to be honest, this ASUS PRIME X299 DELUXE motherboard seems VERY rushed out the door and full of some very basic bugs/issues (even in the BIOS/EFI user interface it's slow and unresponsive with dropdown choice popping up/down in the wrong direction and going out of window so I can't see the options). I may end up returning the motherboard and waiting for the ASUS ROG X299 EXTREME motherboard or something else?

Cheers, Rob.

Shoot!  I'm still waiting to see if ASUS can fix my old P979X WS so I can try to get more miles out of my old  3930K to buy some time until I have to commit to SkylakeX or Threadripper.   I am very much looking forward to how thermal management goes w/ your 7900X when using flight sims, not so much burn-in routines.

[Guest]

Another quick update:

I was able to resolve the shutdown and restart issue by removing my Sabrent USB 3.0 PCIe card (on a side note I had disabled SATA6G_7 port which is required in order to use PCIEX1_2 slot).  I've notified ASUS of the apparent conflict (note the same Sabrent USB 3.0 PCIe card worked in my ASUS RAMPAGE V extreme).  This might leave me short on USB power for my 36 USB devices attached to this PC so I may need to find alternate USB port solutions.

I was also able to resolve the EZ-XMP issue, required a MemOK! clear (button on the motherboard used to improve memory stability) and a few more reboots.  So the G.Skill F4-4266C19D-16GTZR 4 x 8GB RAM is now working well at it's correct XMP profile.

So the two big issues I was able to resolve.  Unfortunately ASUS support has NOT been very helpful, they are at least responding once a day, but it's clearly not an engineer and is someone just going thru scripted responses which aren't really addressing the issues.

I haven't really dived into much OC yet, a very quick tune with the basics and I had no problems running at 4.8Ghz at 1.2v all 10/20 cores/threads, VRM/CPU is cool, but PCH is hot (well within tolerances but is much hotter than VRM and CPU).  Hopefully EK release a monoblock.

I'll report back on P3D V4 performance with baseline clocks (which is 4.5Ghz boost) and OC clocks later this week.

Cheers, Rob.

[Noel 4,501]

31 minutes ago, Rob Ainscough said:

I haven't really dived into much OC yet, a very quick tune with the basics and I had no problems running at 4.8Ghz at 1.2v all 10/20 cores/threads, VRM/CPU is cool, but PCH is hot (well within tolerances but is much hotter than VRM and CPU).  Hopefully EK release a monoblock.

I'll report back on P3D V4 performance with baseline clocks (which is 4.5Ghz boost) and OC clocks later this week.

Cheers, Rob.

Well that's hopeful for sure.  What kind of core temps at what load at 4.8Ghz on all cores?  

ASUS says they will repair or replace my X79 SB-E motherboard so looks like I have a little hope of just plugging in parts and may get some more miles out of it which I'm fine with--means I can postpone going to V4 which I wanted to do when I did whole new build.  Shouldn't be too long before we start seeing Threadripper performance benchmarks.  Despite the hypotheses we spoke of re soldered tin v TIM I have think metallic solder literally binding IHS to die would have to offer the best possible heat transfer TO THE ISF, whereas other factors as you mention having to do w/ how well heat from the IHS gets to the HSF solution will matter hugely and so easily override the IHS interface material contribution to heat management.  As someone who would not ever want to have to screw around w/ delidding the soldered method would I would think have to be the best starting point.

[Guest]

Given the thermal limits of current CPUs (around 90-105C - 221F ... that's engine oil temps and/or coolant temps) I think you'll run into other subsystem issues before you hit temp throttling issues on the CPU.  Under Arthematic stress test the CPU on 20 threads did jump to a peak of 77C and stayed around 75C during the test ... this is a fairly unrealistic artificial test but will stress out all the cores.

I'm working on some graphic data for the loads, testing, etc. etc. ... but like I said, I think this CPU can easily hit 5Ghz (stable) with minimal effort on all 10/20, with the PCH being the hot spot. 

Here were some preliminary videos I recorded while testing (probably should have recorded via external camera rather than stress out my CPU even more with 4K video recording while testing).

Some basic stress testing videos at 4.8Ghz, runs at 5.0Ghz (again this is all basic OC settings no real effort put into working higher OCs):

Cheers, Rob.

[supersym 52]

very interesting ! Thanks for all your work