AMD Clarifies Why Threadripper Uses 4 Silicon Dies

[2FA]

7 minutes ago, The Benjamins said:

also isn't the biggest cost to a wafer the printing of the transistors then the preparation of the wafer it self?

Wafers cost around $500 each.

[Kukielka]

1 minute ago, DeadEyePsycho said:

Wafers cost around $500 each.

Fuck me, I'm getting myself a silicon wafer and use it as wallart or some shit.

[The Benjamins]

1 minute ago, DeadEyePsycho said:

Wafers cost around $500 each.

so like $10 or less per chip I don't think that is a bit deal for a $550+ CPU. that cost the same as the heat sink that comes with the 1200.

[The Benjamins]

6 minutes ago, Kukielka said:

Fuck me, I'm getting myself a silicon wafer and use it as wallart or some shit.

I got a Wafer of a old Chip on my desk at work for $20 on ebay. 

 

 

 

oh and just outside the picture I have a E5-2620 on the sticky notes.

[2FA]

2 minutes ago, Kukielka said:

Fuck me, I'm getting myself a silicon wafer and use it as wallart or some shit.

That's brand new, unused. You can buy "bad" ones for a lot less.

 

1 minute ago, The Benjamins said:

so like $10 or less per chip I don't think that is a bit deal for a $550+ CPU. that cost the same as the heat sink that comes with the 1200.

Let's say at most $50 of cost for each TR CPU.

[DocSwag]

8 hours ago, Lord Nicoll said:

The dies are cut from the water, the wafer is just each die all on a disk.

I know, I was saying who knows if there's even silicon under there. Others have pointed out that there may be transistors, though. 

[Lord Nicoll]

1 hour ago, DocSwag said:

I know, I was saying who knows if there's even silicon under there. Others have pointed out that there may be transistors, though. 

there is silicon under there, the whole die and wafer are made silicon, it has to be BGA soldered to the CPU package and then go through the soldering process, it can't be done any other way as the machines can only be retooled so much without changing a lot on the production line.

[Lord Nicoll]

2 hours ago, M.Yurizaki said:

Who said they were even from silicon wafers?

Because they have to be, there isn't other way to do it reliably. 

[DocSwag]

8 hours ago, Lord Nicoll said:

there is silicon under there, the whole die and wafer are made silicon, it has to be BGA soldered to the CPU package and then go through the soldering process, it can't be done any other way as the machines can only be retooled so much without changing a lot on the production line.

But why does there even have to be silicon under there?!?

[Juanitology]

22 hours ago, Lord Nicoll said:

Those dies cost money, the wafers they're cut from are not cheap, the indium soldering process is also expensive, indium only wets to gold, gold won't wet to silicon either, you have to vapour deposit I think it's vanadium oxide then titanium and then gold to get the indium to stick to the silicon die, and the IHS needs a gold plating, and indium isn't cheap either, that is a lot of expense for two dummy dies, I'm not totally convinced. 

I remember reading on an analysis of the shitty Intel TIM material that the indium solder + gold deposition process cost about 5 usd per die. So they're using about 10USD to guarantee good cooling and structural integrity on CPUs that cost several hundred dollars. I think they'll be fine... if only intel cared that much about their customers

 

[Coaxialgamer]

23 hours ago, Dash Lambda said:

I think what they mean by 'they don't have working transistors' is that they're nonfunctional, not that they're blank.

I can't imagine that they would avoid using failed chips, since spacers on something like this is the perfect way to actually use nonfunctional parts, but at the same time I can't imagine they're only using failed Ryzen dies if their yields are good. So I'd guess they're both using failed dies and some sort of dummies, but that also seems convoluted...

Really, what I'd love for AMD to clear up, is exactly how they sort out and differentiate their dies.

The post says "no transistors", not broken. They are most likely blank wafers. 

Considering the yields, i doubt amd even has many chips  that cannot be salvaged as r3 parts... 

It's possible those are just old or failed wafers. As process nodes shrink, wafer purity needs to increase, and so does cost. 

An old wafer designed for an old process would provide the same functionality in this case. 

 

[Coaxialgamer]

1 hour ago, Terodius said:

I remember reading on an analysis of the shitty Intel TIM material that the indium solder + gold deposition process cost about 5 usd per die. So they're using about 10USD to guarantee good cooling and structural integrity on CPUs that cost several hundred dollars. I think they'll be fine... if only intel cared that much about their customers

 

More like a couple cents to a dollar of gold per die tbh. It only needs to be there for good contact. It's likely to be an extremely thin sheet

[Juanitology]

36 minutes ago, Coaxialgamer said:

More like a couple cents to a dollar of gold per die tbh. It only needs to be there for good contact. It's likely to be an extremely thin sheet

it's mostly the indium and the process of gold vapor deposition that costs money. the gold itself yeah it's likely to be cents.

[bcredeur97]

so which two dies is it? 

It's like a roulette with your CPU...

 

"hmmm which two can I chop off and it keep working"

 

 

LOL

[Lord Nicoll]

6 hours ago, DocSwag said:

But why does there even have to be silicon under there?!?

For the indium to stick to and because der8auer removed some of the indium, it was silicon. 

[ravenshrike]

5 hours ago, Coaxialgamer said:

An old wafer designed for an old process would provide the same functionality in this case. 

It would be hilarious if just for shits and giggles AMD released a limited edition run that used failed Intel wafers for the blank Threadripper dies.

[leadeater]

55 minutes ago, ravenshrike said:

It would be hilarious if just for shits and giggles AMD released a limited edition run that used failed Intel wafers for the blank Threadripper dies.

With the 'Intel Inside' sticker on the box.

[Exotath]

So.... could AMD potentially replace these blanks with working silicon in a refresh later on then? Or would that be too much for this particular platform?

[The Benjamins]

7 minutes ago, Valermos said:

So.... could AMD potentially replace these blanks with working silicon in a refresh later on then? Or would that be too much for this particular platform?

well x399 only has quad channel memory for the ram balance would be broken, so to me it sound like a bad decision.

[Loot]

11 hours ago, The Benjamins said:

well x399 only has quad channel memory for the ram balance would be broken, so to me it sound like a bad decision.

so with that theory ...would their be a massive performance hit for running dual channel ram in a TR rig ?

[Loot]

11 hours ago, Valermos said:

So.... could AMD potentially replace these blanks with working silicon in a refresh later on then? Or would that be too much for this particular platform?

Hmm maybe a CPU w/ GPUs on package A20-9870x APU :))

[TheRandomness]

12 hours ago, Valermos said:

So.... could AMD potentially replace these blanks with working silicon in a refresh later on then? Or would that be too much for this particular platform?

VRMs would probably die trying to power 4 Zeppelin dies at once. 

[Drak3]

32 minutes ago, Loot said:

Hmm maybe a CPU w/ GPUs on package A20-9870x APU :))

With the following statement

1 minute ago, TheRandomness said:

VRMs would probably die trying to power 4 Zeppelin dies at once. 

I'd feel like an X399 APU would murder the system when you plug it into the wall.

[The Benjamins]

1 hour ago, Loot said:

so with that theory ...would their be a massive performance hit for running dual channel ram in a TR rig ?

IDK but I would think AMD would want to avoid it to not have people complain about it having inconsistent ram preformance.

 

In that setup half of the cores will not have direct access to ram so it would have to go through the infinity fabric.

[Jito463]

5 hours ago, Loot said:

so with that theory ...would their be a massive performance hit for running dual channel ram in a TR rig ?

That's a good question.  Unfortunately, we'll have to wait a little bit longer to find out for sure.

 

Personally, I don't think it would be a massive hit, but likely noticeable.