• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD 7nm GPU News and Rumours 2018/2019

Welcome to 3 years ago.

I think it is fairly safe to say that unfortunately HBM hasn't revolutionised the industry. That is not to say there is anything wrong with it particularly, but GDDR with 5x and now 6 hasn't been left behind as expected.:)

The Geforce Titan cards use HBM2, so there is obviously some value in it. Only with AMD you don't need to spend well over a grand for that compute power :p
 
HBM allows this sweet :eek:




Hbm's strengths were less power usage, smaller pcb's and more bandwidth. The smaller card thing died off right after the fury series. Remember that "vega nano" that they "gave" to Tim Sweeney? According to him after they presented it to him onstage and put it in the box he never seen it again. Probably a dummy card of a line of cards that were cancelled.
 
Zotac did a bunch of Pascal mini cards with ~6inch PCBs (around 8inch including the cooling) without HBM in sight (with a single fan design they'd have probably been 6 inch cards).

You mean 970 mini which is a fair bit slower in performance...

I want short like Nano cards because there is no space in the cases.
 
You mean 970 mini which is a fair bit slower in performance...

I want short like Nano cards because there is no space in the cases.

Full range of Pascal cards and they even do a 2070 mini though that one looks a little more chunky and I can't find dimensions off the top of my head - all as fast or faster as any GPU AMD has out currently.

https://www.zotac.com/se/product/graphics_card/zotac-geforce-gtx-1080-ti-mini#spec

Short cards in the long run are really a gimmick, look at the fury cards that had elongated heatsinks on a small pcb to cool the card. Furyx also wasn't really shorter if you take into consideration the aio on it.

Still a mini card but 8inch PCB on that one - some of the other Pascal cards had smaller PCB - but yeah proves the HBM advantage isn't all that massive over plain old GDDR in that regard.
 
The Nano is still shorter - it is 6", while that^^ is 8.31".

The pcb's are about the same length, had they went with a single fan cooler it could have been the same size. Makes little difference really, when you go that small cooling is bound to suffer to some extent.
 
Still a mini card but 8inch PCB on that one - some of the other Pascal cards had smaller PCB - but yeah proves the HBM advantage isn't all that massive over plain old GDDR in that regard.

The PCI-E slot length is the limit, they can't go shorter than the golden connectors.
 
The PCI-E slot length is the limit, they can't go shorter than the golden connectors.

PCI-e works fine at shorter lengths aslong as you go in multiples (a x1 PCI-e card works fine in a x16 slot) and aslong as your application isn't limited by the restricted bandwidth - and a nano isn't really going to be impacted by PCI-e 2.0 x8 for instance.
 
There is not much difference in power usage or performance between a Titan V and a RTX Titan except one uses HBM2 and the other GDDR6 memory.

If HBM uses less power the difference must be very small.


It's well documented that hbm uses less power than gddr, gamers nexus has an entire video on why vega had to make use of it and power was one main reason.
 
PCI-e works fine at shorter lengths aslong as you go in multiples (a x1 PCI-e card works fine in a x16 slot) and aslong as your application isn't limited by the restricted bandwidth - and a nano isn't really going to be impacted by PCI-e 2.0 x8 for instance.

Very good idea - never thought about it - would they do such graphics cards? :eek:
 
It's well documented that hbm uses less power than gddr, gamers nexus has an entire video on why vega had to make use of it and power was one main reason.

Ah yes the myth that we all accept but no one challenges.

What is interesting about Turing cards when gaming is nearly all the silicon gets used, SP cores, Tensor cores, RTX cores and the card packs 24gb of GDDR6.

When a Titan V is gaming it just uses SP cores and 12gb of HBM2.

If HBM uses so little power why do the two Titans above have pretty close power usage even despite the fact that the Turing card is using a lot more of the available silicon?
 
Ah yes the myth that we all accept but no one challenges.

What is interesting about Turing cards when gaming is nearly all the silicon gets used, SP cores, Tensor cores, RTX cores and the card packs 24gb of GDDR6.

When a Titan V is gaming it just uses SP cores and 12gb of HBM2.

If HBM uses so little power why do the two Titans above have pretty close power usage even despite the fact that the Turing card is using a lot more of the available silicon?

Like I said, the gamers Nexus video has a break down of it. Go watch that and it's explained. Afaik it was only around 30 or so watts difference between the 2, but it's less power regardless.


 
Like I said, the gamers Nexus video has a break down of it. Go watch that and it's explained. Afaik it was only around 30 or so watts difference between the 2, bib it's less power regardless.

Not really interested in Gamers Nexus, I am still laughing at them when they made a big thing about putting an AIO on a Titan V and tweaked their setup to set some benching records. It took me one go to beat their scores with a standard air cooled Titan V because they did not know enough about tweaking the CPU memory.

You should not believe everything some of these tech sites claim, it is far better to test for yourself..
 
Back
Top Bottom