AMD Threadripper: 2 of the 4 dies are 'basically rocks' says exec

Quote:

---

So there is "no path to operation", and no way to convert a Threadripper to an Epyc CPU.

---

Read more.

I do like a good flow chart/diagram :D

if you need a 32 core ThreadRipper SIMPLY buy the 32 core Epyc.

Quote:

---

Originally Posted by lumireleon

if you need a 32 core ThreadRipper SIMPLY buy the 32 core Epyc.

---

Quite. I guess people really wanted a 360W TDP :D

Quote:

---

Originally Posted by DanceswithUnix

Quite. I guess people really wanted a 360W TDP :D

---

I rather suspect people wanted 16 free cores, despite the platform clearly not being able to support 4-die MCMs (can you imagine the NUMA mess of having four dies, but only two of them having memory access...?!).

As I've said previously, it's a sensible use of AMD's limited resources to keep the same physical socket and packaging; which inevitably means having to stick something in their to balance the heatspreader. Why not use dies that failed qualification - they'd end up in the bin otherwise...

Quote:

---

Originally Posted by scaryjim

I rather suspect people wanted 16 free cores, despite the platform clearly not being able to support 4-die MCMs (can you imagine the NUMA mess of having four dies, but only two of them having memory access...?!).

---

Would suck for SuperPi, but might work OK for other tasks.

In fact as someone who used to have to use 286 DOS PCs back in the day with their upper memory, high memory, low memory and extended memory it sounds fine :D

edit: Oh and expanded memory, which on a 386 could be emulated using extended memory. *shudder*

Quote:

---

Originally Posted by DanceswithUnix

In fact as someone who used to have to use 286 DOS PCs back in the day with their upper memory, high memory, low memory and extended memory it sounds fine :D

edit: Oh and expanded memory, which on a 386 could be emulated using extended memory. *shudder*

---

And that arcitechture (or an emulation of it for legacy purposes) was carried forwards from there to almost the present day (protected mode anyone?). Is that still the case or has it been killed off now?

Quote:

---

Originally Posted by DanceswithUnix

... In fact as someone who used to have to use 286 DOS PCs back in the day with their upper memory, high memory, low memory and extended memory it sounds fine :D ...

---

Thankfully I didn't do anything complex enough on my 286 PC to need to worry about the memory management ;)

Quote:

---

Originally Posted by scaryjim

Thankfully I didn't do anything complex enough on my 286 PC to need to worry about the memory management ;)

---

Like boot? ;)

In the 386 era I used to have a recurring conversation:

Them: "Linux is complicated"
Me: "What is in your config.sys file?"
Them: ramble for 10 minutes flat on hand crafted memory layouts optimising to put things like mouse and network packet drivers into upper and high memory to get as much of the 640K base memory available as possible
Me: "Yeah, and you say typing 'rm *.txt' rather than 'del *.txt' is complicated?"

Quote:

---

Originally Posted by DanceswithUnix

Like boot? ;) ...

---

That's what autoexec.bat and config.sys were for :p I just left them as they were and let them get on with it!

I love the way he describes them as rocks, i know what he means but it made me smile as even working cores are technically (afaik) rocks what with them being made up of silicon.

please explain how yall did not know this? why has this become a topic to re discuss.

Quote:

---

Originally Posted by elites2012

please explain how yall did not know this? why has this become a topic to re discuss.

---

I think some people had it in their head that AMD was telling fibs about the dies, and that they actually had real CPUs that could be enabled in them. Once the "placeholders" were ground down to reveal real silicone, I guess people had flashbacks from the old Phenom days where cores could be unlocked and got a bit hopeful.

I'll be honest in that I always thought they were just failed components. As others have said, it would make no sense to built entirely separate elements to fill up space when you have duff objects that are already built to fit in there.

Quote:

---

Originally Posted by elites2012

please explain how yall did not know this? why has this become a topic to re discuss.

---

Part wishful thinking to get a relative bargain, part a need to find a lie to debunk.
Seems to be a thrill from catching someone out, proving they are lying.

dummy's save money in design and structual integrity for handling the force of the heatsink, probably

Quote:

---

Originally Posted by me-yeah

dummy's save money in design and structual integrity for handling the force of the heatsink, probably

---

Yeah, it's a known fact that Threadripper uses the same physical socket as Epyc, which does have four functional dies per package, so it's fairly obvious that to maintain the same physical properties you need four somethings under that heat spreader.

But the point here is that only the best 5% of working dies go into Threadripper, so unless you have yields of over 95%, you're always going to have more defective dies than functional Threadripper-binned dies. The only things you can do with those defective dies are landfill or tech-hipster jewellery, so why not repurpose them as spacers?