ThreadRipper Silicon Dies revealed

[chaosmassive]

In a wake of Ryzen 3, now we've get another glimpse of ThreadRipper CPU
instead of two silicons pieces of Ryzen dies, we're greeted by four silicon dies.

This reveal the possibility of AMD could be release even more SKU for HEDT
as we know EPYC CPU have maximum of 32C/64T with 8 channels of RAM.

Below video posted by der8auer who apparently killed the CPU after delid

 

[Toothless]

Rip cpu

 

[FR@NK]

I guess with 4 dies they could better bin for higher clocks and lower power draw since half the cores get disabled.

Or could just be how all the ES chips are setup and retail will use only 2 dies.

And now its confirmed that there is "glue" between the dies holding everything together

 

[chaosmassive]

nd now its confirmed that there is "glue" between the dies holding everything together

still better than Intel's toothpaste

 

[DeathtoGnomes]

stealing other peoples lighting!

glue and toothpaste makes for smelly sticky salad dressing....

 

[MrGenius]

Good God! He totally slaughtered that thing. Claiming like he doesn't know what he did wrong...when it's PLAINLY OBVIOUS. He cut off 4 caps completely and severely damaged a 5th.

It's also pretty obvious which 2 dies are inactive. Gee...the 2 with full rows of caps missing next to them...ya think?

 

[DeathtoGnomes]

Good God! He totally slaughtered that thing. Claiming like he doesn't know what he did wrong...when it's PLAINLY OBVIOUS. He cut off 4 caps completely and severely damaged a 5th.
View attachment 90554

It's also pretty obvious which 2 dies are inactive. Gee...the 2 with full rows of caps missing next to them...ya think?

Hackers of the World UNITE!

 

[Komshija]

Intel is known to use the cheapest possible thermal paste. AMD made better solution by soldering IHS. Actually, this is a lot better option because IHS is "directly" touching the cores and thus providing better heat transfer.

 

[R0H1T]

So TR is basically EPYC rejects, wonder what happens when demand for EPYC & TR soars?

 

[cadaveca]

So TR is basically EPYC rejects, wonder what happens when demand for EPYC & TR soars?

No, not rejects. They are exactly the same; Epyc reduces clockspeed to enable the extra cores within the same power envelope.... by 1/2. Double the cores, halve the frequency = no rejects.

I'll also remind people that these CPUs aren't out there yet, so... don't let the guy that works at whatever tech company color your perspective. He's doing it for dollars. Your dollars.

 

[OneMoar]

here is something interesting likely fud but thats basicly what this thread is

http://www.pcworld.com/article/3211...en-threadripper-has-two-mysterious-chips.html


Two of the 8-core "dies" on the 16-core Ryzen Threadripper 1950x are just dummies to keep the heat spreader from bending out of shape.

So did AMD really waste two perfectly good "Zeppelin" dies? Nope.

Those other two "chips" are nothing more than spacers to help maintain the structural integrity of the gigantic heat spreader, PCWorld has learned from a source who declined to be identified.

The source didn't detail much more, but it makes a lot of sense when you consider the size of the CPU and its construction. Rather than a single monolithic die, like Intel uses on its Core i9, AMD builds its Ryzen and Ryzen Threadripper CPUs using multiple chips connected with a high-speed fabric.

If AMD had populated a Threadripper CPU with just two of the dies, the metal heat spreader would likely bend where the empty spots are.

 

[cdawall]

I wpuld be curious if they are harvested bad dies used as spacers.

 

[OneMoar]

it makes zero sense for amd todo that there are a dozen cheaper more effective ways they could do that

 

[cadaveca]

Good God! He totally slaughtered that thing. Claiming like he doesn't know what he did wrong...when it's PLAINLY OBVIOUS. He cut off 4 caps completely and severely damaged a 5th.
It's also pretty obvious which 2 dies are inactive. Gee...the 2 with full rows of caps missing next to them...ya think?

Then it should be safe to say that this guy either doesn't know what he is doing, really, or he's lying about it. Either way, credibility here is LOST. Hopefully more people see that for what it is.

 

[R0H1T]

it makes zero sense for amd todo that there are a dozen cheaper more effective ways they could do that

That's the other possibility I considered, clearly two dies won't be enough to fit inside the TR package. Now whether the extra dies are totally dud or dummies is something we can only speculate atm, perhaps AMD can clear this bit of confusion later on.

 

[Vya Domus]

Or , there might be a higher core count version in the future after all. Using just 2 dies would have yielded no issue with the heat spreader , weren't C2Q built the same way with 2 dies as well ?
Also it was always more cost effective to manufacture 1 design and then disable stuff. This could be the case for Epyc and TR.

 

[R0H1T]

Or , there might be a higher core count version in the future. Using just 2 dies would have yielded no issue with the heat spreader , weren't C2Q built the same way with 2 dies as well?

With lots of empty space, I don;t remember C2Q delidded but for TR that'd be a huge waste of space inside the package. Whether that hints at more than 16 TR cores in the future is entirely plausible, atm it's all speculation I feel as to whether the dies were ever functional or not & also if this was done just to save assembly costs, presumably EPYC would use the same.

 

[Vya Domus]

With lots of empty space, I don;t remember C2Q delidded but for TR that'd be a huge waste of space inside the package. Whether that hints at more than 16 TR cores in the future is entirely plausible, atm it's all speculation I feel as to whether the dies were ever functional or not & also if this was done just to save assembly costs, presumably EPYC would use the same.

Which tells us that AMD can in fact manufacture these with little cost.

 

[Mirkoskji]

Well as I understood these dies cost amd about 50$ each, when working. So, four dies on epyc are roughly 200$. The dummy die seems a good idea. And if I was amd I wouldn't throw away 100$ of functional dies in TR. Talking about value, the markets were worried about returns of profit on ryzen 5 and 7. With Threadripper they make about 8 times the cost of the two working dies. With Epyc they will make about ten times the cost of the CPU for each piece.