Here is an alleged photo of AMD Radeon Big Navi Radeon RX 6000

by Hilbert Hagedoorn on: 09/10/2020 08:53 AM | source: 搞机猛男 (bilibili) (via ChipHell Forums) | 32 comment(s)

Yesterday Lisa Su from AMD announced that big Navi is going to be announced next month (and there have been far too many announcements about announcements). But meanwhile, in Asia, there's always a guy that has something to show. And as it turns out, that would be a photo of big Navi.

The photo shows the backside' of the allegedly a Big Navi prototype PCB. You can see that Big Navi indeed is big as well as eight memory SMT traces with paper labels "Typical XT ASIC" references for a "16 Gb Samsung GDDR6 memory."

Also, excellent cooler :) (again prototype sample)

That's a 256-bit memory interface then. The chip thus should be paired with 16 GB of GDDR6 memory; we don't think AMD would be silly enough to go with 8GB. It's an engineering sample of some kinds, hence all the paper labels and such. Well, we'll see it at the October 28 launch.

Related Stories

Update Your NAS Server - There is a Vunerability in Samba - 05/26/2017 08:14 AM
There is a new a new vunerability discovered with Samba it affects versions 3.5 (released March 1, 2010) and onwards, the defacto standard for providing Windows-based file and print services on Unix ...

AMD: There is no such thing as full support for DX12 today - 09/01/2015 09:03 AM
After yesterdays turmoil on a lacking DX12 feature for Nvidia AMD&rsquo;s Robert Hallock shares that FuryX is also missing a number of DX12 features. ...

7 pages « < 4 5 6 7

CPC_RedDawn
Senior Member

Posts: 9311
Joined: 2008-01-06

#5829281 Posted on: 09/13/2020 11:41 PM

It's not like AMD, which was a part of the HBM design team, wouldn't know the significance of memory bandwidth. Why would they suddenly starve their cards on purpose? Unless they came up with some phenomenal, previously unimaginable way to double the bandwidth use efficiency.

New rumours floating around that Big Navi could have a 128MB cache using some form of infinity fabric. In order to make up for using GDDR6 instead of GDDR6X (which they can't use).

Not sure how this would make a difference, I think its just more people throwing rumours around.

If they really wanted to tackle the high end then 16GB HBM2 would of easily solved the bandwidth issues. If it is an 80CU monster then wouldn't a 256bit bus starve it of bandwidth? And trying to create a caching system like on Ryzen CPU's would be more trouble than its worth and create memory latency too. Going HBM2 would allow for lower memory latency, more bandwidth, smaller PCB, cheaper PCB thanks to less traces needed. Sort of like mixing Radeon VII with Fury but this time with a gaming focused architecture in RDNA2.

I don't know, the different news stories of standard GDDR6 just doesn't fit well with me with the other rumours of an 80CU chip. It kinda makes me think that AMD really are not going for the high end at all and that 80CU chip is probably a CDNA card.

itpro
Senior Member

Posts: 1359
Joined: 2020-02-20

#5829283 Posted on: 09/13/2020 11:57 PM

Cdna does not exist. At least for gaming. Rdna will get a 3.0 version after 2.0. It is not finished, yet. So, they should focus more in "gaming" performance.

7 pages « < 4 5 6 7

Post New Comment

Click here to post a comment for this news story on the message forum.