Der8auer takes die shots of Threadripper to prove that it contains four Ryzen dies

[WYP]

A full Threadripper 1950X was sacrificed to test AMD's "dummy dies" claim.



Read more on der8auer's Threadripper "dummy dies" research.

 

[NeverBackDown]

His research doesn't mean much. He can't get the equipment to figure out if they worked or not anyway. It's all just speculation. Could just be failed Epyc cores and they Lazer them out completely which is why you can't BIOS hack and add more. Because they are dummies. Like AMD said. They don't work.

 

[Bartacus]

Ok, I must be missing something. Why would he rip apart the top of the line TR looking for blank dies? Aren't the blanks only used on lower end models? Isn't the 1950x the top part, with all dies intact? What key piece of info am I missing here?

 

[AlienALX]

Ok, I must be missing something. Why would he rip apart the top of the line TR looking for blank dies? Aren't the blanks only used on lower end models? Isn't the 1950x the top part, with all dies intact? What key piece of info am I missing here?

No the 1950x is obviously not the top end part because there are a further 16 cores and 32 threads on the die that have been disabled, or are faulty.

Seriously, expect a 32/64 TR at some point after Intel finally launch all of theirs.

His research doesn't mean much. He can't get the equipment to figure out if they worked or not anyway. It's all just speculation. Could just be failed Epyc cores and they Lazer them out completely which is why you can't BIOS hack and add more. Because they are dummies. Like AMD said. They don't work.

I think they are there by design dude I really do. Yeah they may be there to stabilise the IHS but the design is obviously to have four. Oh, and the IF must work on them too, or they wouldn't be there, IMO*

*note, most of that is just me speculating to form an opinion....

 

[KTM_RULZ]

So do they have terrible yields so the have lots of bad dies to place in there or do they have extremely low costs, so they can afford to disable good dies?

 

[AlienALX]

So do they have terrible yields so the have lots of bad dies to place in there or do they have extremely low costs, so they can afford to disable good dies?

It's not a yield problem. I've not seen Ryzen go out of stock once yet !

 

[WYP]

So do they have terrible yields so the have lots of bad dies to place in there or do they have extremely low costs, so they can afford to disable good dies?

Remember that the same Ryzen die covers EPYC, Threadripper and Ryzen 7,5 and 3 products. An absolute tonne of these dies are being produced to meet demand.

Threadripper is not exactly a high sales product, so AMD doesn't exactly need an obscene number of broken dies to use on Threadripper anyway. Only a minority of PC users needs that much CPU grunt.

It isn't really a matter of "terrible yields", it is just the simple fact of life when you create thousands of the same part. Even a yield rate of 99% produces a hundred broken parts when 10,000 components are made.

Looking at the recent Mindfactory data that retailer alone sold over 7,000 Ryzen CPU units in August alone. Add that to retailers globally and the prior sales of Ryzen and that amounts to some huge levels of production and a lot of broken dies.

 

[AlienALX]

AMD said that yields were awesome and that they were cheap to make, IIRC. I guess because Keller looked after them good and proper and made sure that manufacturing wouldn't be an issue.

If only he made GPUs