The Ryzen "ThreadRipper"... 16 cores of awesome

[csbin]

Ryzen Threadripper 1920，12C24T，TDP 140W

 

[mattiasnyc]

Ryzen Threadripper 1920，12C24T，TDP 140W

It'll be interesting to see where it lands in terms of price. There's a $30 difference between the 1700 and the 1700x, and I would expect the gap to be bigger here. $700? $650?

 

[AdamK47]

Going to pair two hot off the press links :

and my own analysis

Conclusion..
None of the AIO water cooling solutions being pushed with the product launch cover the dies sufficiently.
The center of mass of the cold plate is over dead air and up to 50-70% of the active die is left out of contact w/ the copper surface. This is absolutely bogus and unacceptable.

This is the very reason why Noctau designed custom contact interfaces to thread ripper (Because nothing on the markt covers the dies sufficiently) ... Right is ryzen .. left is threadripper. Per the spec, the the Noctau covers the full Heatspreader :

If an AIO water cooling solution with a circular base actually covers the active dies in a diagonal configuration, they have to be 51.9mm in the inner diameter. Nothing on the market is that diameter... I hear EK is making a custom interface for threadripper. This is what's needed for all the AIO watercoolers... It seems this is the only solution that was ready at launch which is why I was hoping AMD actually included a cooler themselves. They're trusting their flagship product in the hands of a bunch of cost cutting third party cooling companies. I shunned Noctau's poo colored fans but at least they engineer a spec'd product.

That answers my earlier question and confirms my suspicions about it.

I'm actually surprised that no one else thought about or was concerned about this.

 

[scannall]

That answers my earlier question and confirms my suspicions about it.

I'm actually surprised that no one else thought about or was concerned about this.

I don't think it will be a huge deal. The dies are soldered to a metal cover, so there is good thermal transfer all around. Optimal? No, but not bad either. For those that want to push it, a little lapping should do the trick.

 

[AdamK47]

I can imagine it creating hot spots on the outer edges of the dies that aren't covered. The ideal path for the heat is up through the IHS and into the block. In this case it would have to traverse the IHS and then move into the block. Heat will build up.

 

[MarkPost]

I suppose that if AMD has validated these coolers, its due to they do the job, but for sure it will be better with specific TR coolers....

 

[Ajay]

Any word yet on whether these are the B2 stepping or not?

 

[formulav8]

They are B1's only for now.

 

[Ajay]

They are B1's only for now.

Thanks.

 

[Pandora's Box]

I spoke to James Prior (works for AMD) this afternoon. I specifically asked him if the EVGA CLC 280mm definitely works with Threadripper. He said AMD has internally tested a handful of AIO coolers, they all work with threadripper. I think these issues gamers nexus is bringing up are a non-issue.

 

[Malogeek]

I spoke to James Prior (works for AMD) this afternoon. I specifically asked him if the EVGA CLC 280mm definitely works with Threadripper. He said AMD has internally tested a handful of AIO coolers, they all work with threadripper. I think these issues gamers nexus is bringing up are a non-issue.

Of course they work. No one would tell you otherwise. Also GN hasn't done any testing, all they were doing is looking at how well the existing AIO cooler plates cover the hotspots. Nothing more. There is no issue yet because no testing is done.

 

[ub4ty]

I can imagine it creating hot spots on the outer edges of the dies that aren't covered. The ideal path for the heat is up through the IHS and into the block. In this case it would have to traverse the IHS and then move into the block. Heat will build up.

And guess what it will traverse over? An already thermally exhausted area over the dies... Thus, you'll have heightened heat in the areas not in contact with the heatsink but over the die and you'll get hot areas on the portions of the die in contact with the heatsink. Bad thermal efficiency plus you're drawing extra heat across and up areas that are already thermally exhausted.. We haven't even gotten into the internal fins occupying an even smaller area on the copper plate that are supposed to be over the die and even worse the orientation.

This is an absolute clown show IMO. Those that don't call it what it is are either being apologist or lack a basic education in physics and heat transfer... Though you don't even need this, a little common sense goes far here.

 

[ub4ty]

I suppose that if AMD has validated these coolers, its due to they do the job, but for sure it will be better with specific TR coolers....

There's a difference between doing the job and doing the job right or efficiently.
I channel the intel mayo paste instead of TIM solution fiasco as a referenceable example of (does the job) vs (does it right). This is near the degree of the intel mayo paste fiasco. Though at least this one can be solved by buying a properly engineered cooling solution and not somebody slapping on a contact plate meant for one single die to two in a diagonal gapped configuration equivalent to the spacing of 4 dies.

Common sense would tell you that something meant for one die sure as hell isn't going to work efficiently on 4 spaced out dies....

Put one of these AIO Water coolers on out the box and leave prime95 running for 24 hours and tell me if your house burns down.

 

[Kenmitch]

....Doubt it would burn down the house.

Guess we'll see if it has the mayo effect soon enough.

 

[ub4ty]

I spoke to James Prior (works for AMD) this afternoon. I specifically asked him if the EVGA CLC 280mm definitely works with Threadripper. He said AMD has internally tested a handful of AIO coolers, they all work with threadripper. I think these issues gamers nexus is bringing up are a non-issue.

LOL, you also could use your own brain and look at the contact area and figure out it isn't working in any scale of efficiency like a properly sized surface would.

See the hot spot and see the thermal trail off? You're getting highly inefficient thermal transfer if you don't have anything contacting the surface immediately above the heat source (dies). This is why the center of mass and the internal fins are directly in the center of all of the AIO water cooling contact plates. They are designed for one die dead center not 2 dies spaced out and at a diagonal. This is inefficiency on inefficiency. Also, think about covering half of the above heat source with a heatsink... Now tell me, where does the heat flow across to get up into the heatsink? Back across a max'd our thermal region .. multiplying the heat that's in it.

Sure it works and that is likely not under persistent heat stress. God knows what this manifests in the long run.. At the temperature processors run at when max'd out, I'm sure you're going to get warping and all sorts of issues down the road. A freshman in college out of physics could tell you this. I can't believe actual people who do this for a living are lying to people and saying otherwise.... I'm not sure what type of fools they think people are.

 

[Pandora's Box]

End of the day I doubt the Noctua TR4 cooler is going to be available for launch of Threadripper. I'll use my EVGA CLC 280mm cooler for now, will keep an eye on the Noctua cooler and may switch out down the line.

 

[Markfw]

End of the day I doubt the Noctua TR4 cooler is going to be available for launch of Threadripper. I'll use my EVGA CLC 280mm cooler for now, will keep an eye on the Noctua cooler and may switch out down the line.

Exactly ! And I will report loaded temps on both coolers.

 

[ub4ty]

Deeper down the rabbit hole we go...
Dell is using a much larger custom AIO surface mount and bracket than what is being offered by the compatibility list and AMD. They switched to it towards launch after using skimpy brackets in testing. Clearly the Compatibility list solutions and AMD's bracket wont cut it for a production system ... The fact that an OEM has already moved on from it some time ago shows they knew of this issue and know that these compatible list coolers aren't going to cut it.. This gets uglier the more I look into it.
Detailed here :
http://imgur.com/a/EonHL

 

[mattiasnyc]

This gets uglier the more I look into it.

Maybe you should take a break then. I don't think I've ever seen someone this "intensely" looking at a presumed issue before the testing has been done getting this "agitated". We're what, only 7 days away from the gag lifts on reviews, right? So, 7 days of waiting and then we'll know.....

 

[maddie]

LOL, you also could use your own brain and look at the contact area and figure out it isn't working in any scale of efficiency like a properly sized surface would.

See the hot spot and see the thermal trail off? You're getting highly inefficient thermal transfer if you don't have anything contacting the surface immediately above the heat source (dies). This is why the center of mass and the internal fins are directly in the center of all of the AIO water cooling contact plates. They are designed for one die dead center not 2 dies spaced out and at a diagonal. This is inefficiency on inefficiency. Also, think about covering half of the above heat source with a heatsink... Now tell me, where does the heat flow across to get up into the heatsink? Back across a max'd our thermal region .. multiplying the heat that's in it.

Sure it works and that is likely not under persistent heat stress. God knows what this manifests in the long run.. At the temperature processors run at when max'd out, I'm sure you're going to get warping and all sorts of issues down the road. A freshman in college out of physics could tell you this. I can't believe actual people who do this for a living are lying to people and saying otherwise.... I'm not sure what type of fools they think people are.

So a less than 5 Celsius difference? Based on top graphic.

 

[ub4ty]

Maybe you should take a break then. I don't think I've ever seen someone this "intensely" looking at a presumed issue before the testing has been done getting this "agitated". We're what, only 7 days away from the gag lifts on reviews, right? So, 7 days of waiting and then we'll know.....

I've provided the detail that are necessary ... I no longer have any comments.
It's the individual's choice from here on out

 

[EXCellR8]

interesting... i almost feel like a dual-chamber block would work better but then again the intersection in the middle of all four dies will likely still be the hottest spot. if anyone wants to fund my own personal research please step on up and I'll figure it out.

 

[moinmoin]

Do we know if the two dies and two spacers in TR are always in the same location, or can that change between chips? Depending on that there is plenty room for future optimized cooling solutions specific to TR.

 

[Markfw]

I am not going to worry about the water block, until I run some benchmarks ! And I will get the Noctura NH-U14S TR4-SP3 to test with also (when available) and decide THEN.

 

[DrMrLordX]

Put one of these AIO Water coolers on out the box and leave prime95 running for 24 hours and tell me if your house burns down.

Yay for hyperbole.

How about letting people run some tests with these coolers instead of trying to manufacture controversy? You sound like someone trying to distract from problems with some competitor's recent platform launch . . .