I can see that descreet ray tracing cards would definitely have a market considering in order to have viable on board ray tracing you have to shell out a lot of money for a top tier card which is enthusiast only territory and still take a performance hit with it on. A descreet card could bring ray tracing to the masses as it wouldn't mean a GPU upgrade was required for those happy with their current cards and those buying 3070 or below wouldn't have to choose between frame rates or ray tracing. Even if it was a short term solution until on board ray tracing was viable for all new cards there would be plenty of demand.
-
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.
NVIDIA ‘Ampere’ 8nm Graphics Cards
- Thread starter LoadsaMoney
- Start date
Soldato
- Joined
- 20 Apr 2004
- Posts
- 4,444
- Location
- Rugby
I think a separate card would not work on a piratical level as there be the added latency and increase frame times from the GPU waiting on the Ray trace card or "RTPU"
Was also thinking this... it would just be a terrible idea vs putting RT into a GPU.
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,802
- Location
- Planet Earth
AMD and Nvidia are both trying to work on "chiplet like" implementations for GPUs,so if it was multiple dies on a single substrate it might be workable. But at some point we will need to go that way,as manufacturing becomes harder and harder.
Last edited:
Separate ray tracing card.......not going to happen, unless all the other leaks and rumours are wrong. As only the top tier card is suppose to have an NVLink connector, so what is the point of an add in card that you can only add to the top tier card.
The co processor idea.....maybe, but I doubt it.
The co processor idea.....maybe, but I doubt it.
Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,802
- Location
- Planet Earth
It depends. I would imagine it would need a high speed interconnect,and some way of syncing all the different chips together. It might be worth going the monolithic way for the immediate future,but I would argue there would be manufacturing advantages going with smaller chips. The question is who will get there first?
- Joined
- 8 Jul 2003
- Posts
- 30,080
- Location
- In a house
They won't need NVLink, they're on about them being like the old PhysX cards, that you just had to plug in another slot.
Last edited:
Permabanned
- Joined
- 8 Dec 2015
- Posts
- 1,485
Ive been watercooling long enough Andy to know that watercooling decreases temps which allows you to increase the voltage which can help sustain a higher overclock. Overclocking is a skill and to find prime 95 stability takes alongtime to do this when your pushing your cpu/gpu threshold. Maybe you lack the skills or just wasnt doing it right. Im only going by my own experiences here and anybody that thinks or says that air cooling gives the same performance over watercooling (custom not AIO) are simply deluded.
Permabanned
- Joined
- 8 Dec 2015
- Posts
- 1,485
Not used nowadays to know if they are anygood and tbh i dont really care. I also dont like that you cannot enforce MSAA+SGSSAA for AMD cards. I will always buy the premium which is Nvidia, thats my choice although i know i will pay more for it.
And I don't like that I cannot control the level of Tessellation on an Nvidia card or that all their so called premium features come with a premium frame drops as well making them useless to me personally. You are right, you can do with your money as you see fit and I don't care which card you buy. Buy Nvidia if it makes you sleep better at night
but to be honest, and I seriously don't mean this as an insult or an attack on your person but I find your way of thinking very single-minded. By this I mean, stating that you will always buy from a certain brand, or ignore another, without scrutiny and without weighing the pros and cons is going to bite you one day.One Vision! Only see’s greenAnd I don't like that I cannot control the level of Tessellation on an Nvidia card or that all their so called premium features come with a premium frame drops as well making them useless to me personally. You are right, you can do with your money as you see fit and I don't care which card you buy. Buy Nvidia if it makes you sleep better at night
They won't need NVLink, they're on about them being like the old PhysX cards, that you just had to plug in another slot.
Ok so it plugs into another slot.
Just how does it send and receive the data fast enough to be able to calculate the BVH info?
NVlink on a 2080ti is 100GB/s
PCIe gen 3.0 is 16GB/s
PCie gen 4.0 is 32GB/s
Not really a contest is it.
But hey maybe the info that is needed to be passed back and forth isn't that large so it can go across the PCIe bus no problem.
Any way in the corteks video, which is where this idea came from.
He clearly states the add in card would use be using NVlink.
Not that it matter as it isn't going to happen anyway. Just my opinion, and as we know opinions are like backsides, every bodies got one and very often they are full crap.
How much bandwidth would a RT card need? That would tell us if Nvlink would be essential or not. If less than 32GB/s PCIE 4.0 would be sufficient. It's a Shane the industry can't get together more and agree on an approach then just market their version of it.
- Joined
- 8 Jul 2003
- Posts
- 30,080
- Location
- In a house
If the NVLink is the case, then they may have NVLink on all the Ampere cards, and not just like now, only the high-ends.
We just don't know anything, until the cards are released, its all just speculation
We just don't know anything, until the cards are released, its all just speculation
I wouldn't go as far as to say it's a skill, more a matter of trial and error and patience.
