The Threadripper Report (or the lack of it)

[JPQY]

When i see your last screen..then it should also handle my Chess-bench!

JP.

[chew*]

Just wanted to point out why I am not @ 4 gig.

https://m.hardocp.com/article/2017/0...k_comparison_1

https://m.hardocp.com/article/2017/1...coolers_review

This is the reality of things and 4 gig claims and currently the enermax is the best TR aio but its loud and still sux compared to a custom loop.

Might want to point out that although the bykski and xspc managed to get 4 gig....it did so above AMDs recommended max of 68c ( 1600 rpm fan speed ) as mentioned per The Stilt.

I have a bykski on the way as i did not feel like paying $50 more for 1c.

Aaron ordered something else that he mentioned looked easiest to modify.

[bluestang]

Waiting for your thoughts after you've tested that Phanteks Glacier C399a CPU block

[chew*]

Waiting for your thoughts after you've tested that Phanteks Glacier C399a CPU block

Heres my take on it.

Thin plastic or brass sheet or copper sheet or even maybe stainless cover directly over plate. highlighted in black completely changes the way this block can or will flow.

forcing it to hit the outer edge of plate first

also forces water to exit via channels thus forcing air out of the block which a user who has it has mentioned is an issue.

Still not optimal as all designs should be done in a "quadrant" so it covers the epyc and TR market not to mention since AMD was not forthcoming we are not 100% of which dies are active.....

all blocks should exit at top so as to not trap air.....or divert flow so as to force out air.



I'm fairly certain Aaron can/will walk circles around me and my idea.

Here's my horrible paint idea for a TR/EPYC block to expand the market potential.

Problem is the heat exchanger would be crazy expensive to make/machine but if money was no object this is my idea.

A copper wall would separate all Dies into a quadrant. I would extend the wall down at bottom to direct flow.

This design would also not be able to trap air in the block as it would be forced out.

central fired inlet smack dab in the middle of the Quadrant. top mounted outlet.

[chew*]

Mail came. Clean it and plumb it tonight.

Think I may lap it first to...

[chew*]

#3 contender.

[bluestang]

Heatkiller?

[chew*]

Phanteks

[chew*]

Room for improvement.

[chew*]

Just a 3x120 slim mcp 35X pump lapped bykski block.

[BeepBeep2]

Lapping help temps quite a bit? Wonder how flat the IHS is.
Aaron was looking for a WC top for the LN2 evaps, but I never had one

Looking good chew*.

Dual outlet waterblock top also might be worth considering, though it would be a PITA

[chew*]

Its an mcrx320 has like 3 inlets so would not be to hard. I grabbed it for flexibility/space reasons. Integrated mcp35x for high head and integrated res.

I may lap the cpu all this stuff is going to Aaron soon almost done then I will screw around with my setup.

Way behind schedule. 1 person and so much to do.

[chew*]

Nothing is flat apparently.

[Tom128]

I decided after many years to do an all out upgrade and went with a 1950X on a Zenith Extreme. I have some high hopes I may have a decent chip. At the very least if it's not a good max clock chip, it's a good chip for low power at modest clocks. I've been running it at 3.8Ghz 1.15v vCore and 0.85v SOC 100% stable and have not even tried if the vCore could be lower, but I bet it could. Currently seeing what sort of ceiling it has and I am stability testing for 4Ghz at 1.275vCore and 0.95v SOC for a while now and it looks promising. I don't have a stellar loop either, MPC355 with XSPC top from many years ago (XSPC D5 170 proton arrived today since I don't trust this long term), slim swiftech 360 rad, and a XSPC Raystorm. I am using it for a work station so part of my thinks I should just run with the 3.8Ghz and be happy but I want to see if I can get 4Ghz stable and then how much past that it can really go

[Buckeye]

I don't expect going past 4GHz will be worth it.
Word is even extreme cooling like Phase doesn't buy you much in OC.

I am running a ASUS X399 Zenith Extreme, it has a built in 4GHz OC. Using one of Thermaltakes Triple Rad AIO's which seems to work well. Maybe sometime I will switch to a real WC system but not sure if its worth the investment and time. The built in 4GHz OC works pretty darn well, maybe a tad high on vcore but oh well, it works and hasn't crashed on me with everyday stuff and gaming.

[Charles Wirth]

Brian is already at 400W 3.9Ghz with some kind of custom chilled loop showing -17c/73c

I planned to use two cascades, I had the mounting designed and just about to start cutting then amd unfriended me.

[NEOAethyr]

Brian is already at 400W 3.9Ghz with some kind of custom chilled loop showing -17c/73c

I planned to use two cascades, I had the mounting designed and just about to start cutting then amd unfriended me.

Oui, that's not good.
You'd think regardless of results they'd still be enthused about it.

[Tom128]

4Ghz seems to be a pretty hard ceiling on these chips. I got 4Ghz stable at 1.35v vCore and 0.95v SOC but the temps were leveling off at 67c. Sort of disappointing, but considering I can get 3.8Ghz while undervolted, I will just run with that.

[bluestang]

chew*, have you finalized the cooling setup yet?

[Opteron96]

@Chew*

Have you noticed any different in overclocking, by grounding all the holes in the motherboard?

[lowdog]

R.I.P...........thread

[DavidMMM]

Maybe I found the King of the hill? - 1950X air-cooled with Noctua UH14S-TR4 15cm fans push-pull, Vcore 1,185 in BIOS, 4GHz Cinebench, not stable for this frequency / voltage, otherwise works like a charm 3,9GHz with 1,175V.

[Charles Wirth]

[Buckeye]

WOW, very nice !!!