AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion

Too many messages to quote, but it's 99.99% sure that Vega uses Samsung HBM, not SK Hynix, at least for the 16 GB models
 
Too many messages to quote, but it's 99.99% sure that Vega uses Samsung HBM, not SK Hynix, at least for the 16 GB models
Anandtech just launched a newspiece claiming SK Hynix still hasn't started mass production of HBM2, so RX Vega is definitely coming with Samsung HBM.
 
To bring Project 47 to life, AMD worked closely with Samsung Electronics with respect to the HBM2 memory used across the “Vega”-based product lines including the Radeon Instinct MI25 accelerators. Samsung also provided high-performance NVMe SSD storage and high-speed DDR4 memory to enable the 1 PetaFLOPS of performance. AMD also collaborated with Mellanox Technologies, leveraging their InfiniBand solution to deliver 100Gb connectivity through the rack.
Little more context on that quote. So looks like Samsung HBM2.
 
So, I guess that puts the "AMD is getting exclusive access to 1GHz chips" rumour to rest.

Well I think all rumors of which HBM2 AMD were using were based off Hynix as Samsung didn't release their products so we have no idea what base speed or base voltage right?
 
Unfortunately, Anton messed up the source link - or is that his way of telling us there is no source but original content?
Sorry about that. The URL to the earnings call transcript was screwed up.

https://seekingalpha.com/article/40...-results-earnings-call-transcript?part=single

The complete relevant passage, which was in response to an institutional investor question:

"And compared to the GDDR5, it is clear that both products would have price premium over it. And although I cannot specify the exact pricing that we expect, I can tell you that the customers are willing to pay at least double or, let's say, 2.5 times of the HBM."
 
Sorry about that. The URL to the earnings call transcript was screwed up.

https://seekingalpha.com/article/40...-results-earnings-call-transcript?part=single

The complete relevant passage, which was in response to an institutional investor question:

"And compared to the GDDR5, it is clear that both products would have price premium over it. And although I cannot specify the exact pricing that we expect, I can tell you that the customers are willing to pay at least double or, let's say, 2.5 times of the HBM."

Ohh ok, so i was think the comparaison was directly aimed to GDDR5, but it seems ( if im not too drunk ), that they speak about a direct comparaison between HBM and HBM2 ?
 
I find the mellanox-part more intriguing - so they're not leveraging their own infinity farbric beyond MCM?
AMD's fabric so far has demonstrated going as far as a second CPU socket. Since it's based on Hypertransport, this seems consistent with it being an interconnect within components of the same board.

InfiniBand is a standard that works within a server board,and then has the range and routing capability to go throughout a rack between blades, network, and storage.
AMD hasn't indicated that its interconnect has been given a longer reach than it did the first time, and there's nobody else using AMD's fabric for it to link to in a rack.
 
Sorry about that. The URL to the earnings call transcript was screwed up.

https://seekingalpha.com/article/40...-results-earnings-call-transcript?part=single

The complete relevant passage, which was in response to an institutional investor question:

"And compared to the GDDR5, it is clear that both products would have price premium over it. And although I cannot specify the exact pricing that we expect, I can tell you that the customers are willing to pay at least double or, let's say, 2.5 times of the HBM."
Thanks for looking into it, Ryan and thanks for reporting on it in the first place as well!

But i must say, the wording from seeking alpha is a little unclear to me - not being a native speaker, i would expect him to say "for the HBM" if he meant to compare to GDDR5 and "of the HBM(2)" if he meant to compare HBM gen2 to HBM gen1. Maybe someone with a better grasp of the english language can help me out here?
 
Thanks for looking into it, Ryan and thanks for reporting on it in the first place as well!

But i must say, the wording from seeking alpha is a little unclear to me - not being a native speaker, i would expect him to say "for the HBM" if he meant to compare to GDDR5 and "of the HBM(2)" if he meant to compare HBM gen2 to HBM gen1. Maybe someone with a better grasp of the english language can help me out here?

The whole sentence has small grammatical errors that require some parsing, but I read it as initially stating that both variants of HBM are able to be sold at a premium over GDDR5, with HBM2 able to be sold at an additional 2.5x premium over HBM.
 
Thanks for looking into it, Ryan and thanks for reporting on it in the first place as well!

But i must say, the wording from seeking alpha is a little unclear to me - not being a native speaker, i would expect him to say "for the HBM" if he meant to compare to GDDR5 and "of the HBM(2)" if he meant to compare HBM gen2 to HBM gen1. Maybe someone with a better grasp of the english language can help me out here?
He doesn't appear to be a native English speaker either, which coupled with the on-the-fly nature of earnings call Q&A responses, is what leads to the odd structure of the statement.
 
But i must say, the wording from seeking alpha is a little unclear to me - not being a native speaker, i would expect him to say "for the HBM" if he meant to compare to GDDR5 and "of the HBM(2)" if he meant to compare HBM gen2 to HBM gen1. Maybe someone with a better grasp of the english language can help me out here?
It reads like it was translated or spoken by a Korean, which isn't unreasonable, so you likely understand as much as everyone else here.
 
Native English speaker here, and the last sentence is rather awkward.

it is not akward, this is just financial talk for investors.. they say as much than nothing for content them..

But in financial words, this mean: we are really really confident to increase our growth in Q3 .. ( SK Hynix is second behind Samsung in the memory market ( with a 3 time differences ), they have capture around 11.5% of the market, when Samsung is still holding 44.7% of the market ( Q2 and Q3 should goes around 46.3% for Samsung as a result it seems, but Hynix should close the quarter at 11.8-... )
 
Last edited:
He doesn't appear to be a native English speaker either, which coupled with the on-the-fly nature of earnings call Q&A responses, is what leads to the odd structure of the statement.

Taiwanese.. not Korean.. as i have write before, this statements are made for financial market, and if sometimes we find good information in there, they are not intended for give us any indication than for the financial market, referring to other financial statements, than any ... ( well choose you term behind this )... why i say that, is just to be really careful on this statement.. ( this said, you can sue a company for push wrong information during an cast, so in general they are really careful to reproduce, their vision , of the truth in line with the market .. ( Ask Toshiba how they feel now about it )
 
Last edited:
Back
Top