• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Minor Upgrades for my X99 Platform

  • Thread starter Thread starter Deleted member 50521
  • Start date Start date
D

Deleted member 50521

Guest
Got a 6950X to replace my "old" 5820K. Now this X99 platform should be able to serve me until 16 core 32 threads become mainstream.


Out with the old 5820K, which has helped me powered through loads of my work. It should show up in the for sale section eventually. It was a good processor with an awesome IMC, handled 128GB RAM beautifully :D

File_000.jpeg



And in with the new. Somehow the 6950X feels lighter than the 5820K. The PCB is also thinner. Weird.

File_002.jpeg


Thermal paste up and ready to go!
File_003.jpeg



Clearing UEFI settings and powering up. No problem at all. The Sabertooth X99 took it in just fine.
File_004.jpeg



After a little bit of tweaking I settled for 1.19V @ 4.2GHz for all 10 cores, 3.5GHz Uncore and DDR4-3000 as usual for the RAM.

I tested briefly on DDR4-3200. Welp this thing is flying fast. I am seriously impressed by the improvement of IMC from HWE to BWE. Also I am able to run DDR4-3000 on 100 BCLK now!

2017.07.24-20.42.jpg
2017.07.24-22.53.jpg


Did a quick stability test. Checks out as well. All in all I am very satisfied!
2017.07.24-22.26.jpg



EDIT1:

Totally forgot, the Noctua D15 seems to be able to handle the heat from 4.2GHz 6950X, just barely though. As you can see the temps are creeping uncomfortably high. Most likely I will be getting a very good AIO down the road.
 
Last edited by a moderator:
The AIO won't do much for temps that noctua equals and bests most of them.
 
cdawall said:
The AIO won't do much for temps that noctua equals and bests most of them.
Click to expand...

Even something like the yet to be released EK new MLC AIO? Well then i guess i just need to tweek the voltage a bit more. So far im not liking the BWE vid based vcore setup.

Or i may try triple fan setup on the D15
 
I would go for either custom loop or a Swiftech AIO to help on the temps, price is almost the same.
 
Ebo said:
I would go for either custom loop or a Swiftech AIO to help on the temps, price is almost the same.
Click to expand...
Just worried about Swiftech's QC. Many horror stories about leaking unit or DOA pumps.

Full custom loop is no go. Have little knowledge in setup full loop. A good AIO is all I'm shooting for
 
is this the max temp ya got 58° ?
i think will be hard to beat(maybe 2-4° better)-i would be super happy running 10 cores 20 threads at 4,2ghz with this good temp.
hail noctua
 
basco said:
is this the max temp ya got 58° ?
i think will be hard to beat(maybe 2-4° better)-i would be super happy running 10 cores 20 threads at 4,2ghz with this good temp.
hail noctua
Click to expand...
The Max was 90. And yes, BWE is quite a hot one compared to HWE, especially the more cores it has.
 
xkm1948 said:
Even something like the yet to be released EK new MLC AIO? Well then i guess i just need to tweek the voltage a bit more. So far im not liking the BWE vid based vcore setup.

Or i may try triple fan setup on the D15
Click to expand...

The EK will probably do better I mean the generic asetek based stuff. I would try three fans first and go from there.
 
OK BWE overclock was not "cut and dry" as I thought. I used overclocking knowledge from HWE and boy was I wrong.

It seems Intel used some sort of VID mechanism for built in turbo. So even if I set additional final turbo voltage to 1.19, during load the CPU will still refer to its built in VID for vcore. In my case this voltage is a god damn 1.255V! Way too high for my liking. Good thing is 4 hrs of reading and tweaking gets me to understand how this works. By utilizing negative voltage offset, I get the adaptative vcore to come down to 1.21V under full load. This is with 1.175V set as adaptative vcore in UEFI. Not as good as I thought, but a massive improvement in terms of temperature during RealBench Stress test.

With the same room temperature I am seeing accross the board ~10C decrease for CPU package during stress test. Also I used Noctua's NT-H1 this time instead of AS5. Apparently NT-H1 works better than AS5.

So here is my final results. 4.2GHz full load around 1.21V vcore, max CPU temp is around 80C. Pretty good I would say for a freaking 10 core and an air cooler. If I have time I would totally fine tune it to the max. But for now this machine needs to go back to work. Already have loads of work piled up for the shiny new 6950X.

@cdawall, If I am going to with triple fan, does it matter if I mount a slim 14mm fan from a different vendor there? With my FuryX radiator right behind the D15 the space is quite limited. Thanks in advance!

2017.07.25-13.26.jpg

IMG_1094.JPG
 
4.2GHz seems to be the sweet spot for 6950X. 100MHz more would require 1.29V vcore and the temperature will become too ugly for D15 to handle.

And finally putting the beast to good work! Having 20 logical processors surely helps quite a lot. Assay time reduction from ~6hrs per data set to ~3hrs per data set. Cut time needed in half! Now I wish I have 256GB of RAM though.

2017.08.03-22.47.png
2017.08.03-22.47_01.png
 
Damn, seeing 120 of 128 GB or RAM in use is pretty impressive I've got to say
 
xkm1948 said:
@cdawall, If I am going to with triple fan, does it matter if I mount a slim 14mm fan from a different vendor there? With my FuryX radiator right behind the D15 the space is quite limited. Thanks in advance!
Click to expand...

Don't know how I missed this. It would be more efficient to run a blank spacer and try to seal it to the rad as best as possible.
 
Nuckles56 said:
Damn, seeing 120 of 128 GB or RAM in use is pretty impressive I've got to say
Click to expand...

He's the only legit use case for that kind of ram I've seen on this forum outside of servers or VMs. Pretty epic I must say.
 
Nice setup.

the D15 is a really good air cooler.

how about mounting a fan on the outside of the case, I have used that reversed teknik in/on my case... :)
 
cdawall said:
Don't know how I missed this. It would be more efficient to run a blank spacer and try to seal it to the rad as best as possible.
Click to expand...

OK I must be dumb. What exactly are you proposing to do here? Do you have any illustration of the design? Also why would a blank spacer work better? Any source of proof for that? Thx!

JalleR said:
Nice setup.

the D15 is a really good air cooler.

how about mounting a fan on the outside of the case, I have used that reversed teknik in/on my case... :)
Click to expand...


Hmm, that is surely an interesting idea
 
Last edited by a moderator:
xkm1948 said:
OK I must be dumb. What exactly are you proposing to do here? Do you have any illustration of the design? Also why would a blank spacer work better? Any source of proof for that? Thx!
Click to expand...

Stacking fans causes interference you would be better just channeling the heat through especially since you already have what I assume is a good static pressure fan on the rear.

A 3d printed fab shroud for your needs would be a great option imo.

http://www.performance-pcs.com/phobya-radiator-shroud-20mm-for-120mm-fans-black.html
 

Attachments

  • 20170805_104416.jpg
    20170805_104416.jpg
    1.9 MB · Views: 203
  • 20170805_104424.jpg
    20170805_104424.jpg
    1.8 MB · Views: 290
  • 20170804_102635.jpg
    20170804_102635.jpg
    1.5 MB · Views: 191
You must log in or register to reply here.
Back
Top