Nvidia Ampere Discussion [2020-05-14]

Alleged leaked specs as well:
https://www.notebookcheck.net/NVIDI...10-GB-VRAM-and-4-325-CUDA-cores.469098.0.html

Doesn't seem to leave room for a 3080 Ti SKU with specs like these. GA102 for 3080? 3080 Ti will *not* be using GA100, and NV has never used 101 naming for a chip before, to my knowledge.

My guess is this is fake, and I hope that's the case. Don't get me wrong, specs like this for 3080 would be fine by me, but again, where's the Ti?

Apparently this is the source of the rumors:

Faaaaaaaaaake
GDDR6x?
Please.
 
You don't need room for a 3080Ti in this lineup as the 3080ti already exists. It's just a renaming of the 3080ti to 3090. They should've done this with Turing already, then only half the people would've complained about the "RTX2090", as it would've been a new product class.

kopite7kimi was very reliable so far and german Igorslab, which has very good contacts to AIBs is telling the same story with RAM.

PCBs are changing with Ampere, because the new Ram (GDDR6X) needs to be nearer to the gpu than before. Speed over 20Gbps... Manufacturers need to change their cooling solutions for ampere, becaue GPU and RAM are so close and the GPU shouldn't heat up the Ram.
 
I remain skeptical. We had how many generations of products with GDDR5 before 5x came along, you're telling me that we get 1 generation of GDDR6 products before 6x?
 
I remain skeptical. We had how many generations of products with GDDR5 before 5x came along, you're telling me that we get 1 generation of GDDR6 products before 6x?
I don't buy the rumour, but if any vendor could deliver faster memory technology at reasonable price, AMD and Nvidia would use it as soon as possible.
 
Both Samsung and Micron list 16Gbps modules as their fastest memory in production on their websites. Nothing faster than that is in sampling. And SK Hynix must have fallen asleep or something.
The offer on their official websites is not exactly the same as the offer they can present to big clients like AMD and Nvidia.
More suspicious is that JEDEC never mentioned any further improvements to GDDR6. All previous standards were announced months before the first products with new memories were released to the market.
If this was true it could be some kind of innovation outside of GDDR6 modules or some properiaty solution based on GDDR6.
 
"6x" is completely fake. Not only do ram providers publicly announce new product lines well in advance, it's not some super top secret thing they spring on people, JEDEC, the engineering standards body for random access memory, announces standards even before that with accompanying papers. And 6x is nowhere to be mentioned.

The only advancement in GDDR ram coming up is a bump to 18gbps, announced quite a while ago.
 
cooling consept edited by me over the chiphell image, the memory arrangement and pcb detail is stand in no conformation of 256bit memory
174239vr70d9ct7c0rg69t5jqc.png

Would this setup be SLI friendly? Maybe checkerboard rendering will be a thing on Ampere.
 
someone photoshopped the back of the card without foils
mvqsa8atv9351ahj40.png


cooling consept edited by me over the chiphell image, the memory arrangement and pcb detail is stand in no conformation of 256bit memory
174239vr70d9ct7c0rg69t5jqc.png
I'm not expert on these, but apparently the fan geometry confirms that both fans push "in". Based on what can be seen in pictures of the card, the most likely explanation is that the "rear fan" is in fact completely independent and separate from the card, and instead heat gets transferred there via heatpipes.
 
I'm not expert on these, but apparently the fan geometry confirms that both fans push "in". Based on what can be seen in pictures of the card, the most likely explanation is that the "rear fan" is in fact completely independent and separate from the card, and instead heat gets transferred there via heatpipes.

Yeah it looks like 2 intake fans with hot air exhausted out the case. Maybe the offset fans better help direct airflow.
 
Yeah it looks like 2 intake fans with hot air exhausted out the case. Maybe the offset fans better help direct airflow.
The end fan can't be an intake as it wouldn't be cooling anything at all.
 
Presumably the heatsink extends through the full length of the card. Either way the fans in the pics are definitely oriented for intake.
Well they're all fake concepts, the orientation of the fan isn't necessarily correct. Good point on the heatsink extending though, that's possible.
 
Back
Top