RTX 3xxx performance speculation

Dayaks said:
Why would they sit on launching a card? Now is the time to protect market share more than ever.
Click to expand...

Possibly because they'd only be competing with themselves in the high end. I would assume RTX and current GTX cards are pretty mature in terms of manufacturing and are turning higher profit per card now than they were a year ago, so why not milk it a bit further ?
AMD is so far behind them on the top end, and they have so much more mindshare.
 
Digital Viper-X- said:
Possibly because they'd only be competing with themselves in the high end. I would assume RTX and current GTX cards are pretty mature in terms of manufacturing and are turning higher profit per card now than they were a year ago, so why not milk it a bit further ?
Click to expand...

When you have your enemy under your heel you don't let up the pressure.
 
DooKey said:
When you have your enemy under your heel you don't let up the pressure.
Click to expand...
You're assuming they have an answer to AMD in the mid range section that doesn't cost more. I'd also assume that right now the market is slower than they'd like for a release :p
 
Digital Viper-X- said:
You're assuming they have an answer to AMD in the mid range section that doesn't cost more. I'd also assume that right now the market is slower than they'd like for a release :p
Click to expand...

No, I'm not. They seem to be selling plenty of cards in the mid range for more than AMD cards sell for. Brand recognition aka mindshare still belongs to Nvidia. They will continue this strategy until they see it doesn't work. Why wouldn't they?
 
There is some benefit launching first if your product beats everything else for the price, the blitz of hungry reviewers desperate for something to show, jump on etc. Anyways if you can get very positive reviews, those will live on into infamy even if your products sucks after your competition blows you away.

Anyways it is not as if it is a facet where you can turn on and off, contracts like with Samsung will have to be fulfilled, board makers and parts for them such as memory bought, marketing . . . Unless Nvidia has not committed to any contracts for items needed, only delays I see is if there is a serious problem with the product(s) or supply chain issues.
 
I'm just curious, about how long before release do GPU manufacturers start producing product? If the 3000 series is scheduled for what, August/Octoberish(?), when would they have to decide when to start making them?
 
DooKey said:
No, I'm not. They seem to be selling plenty of cards in the mid range for more than AMD cards sell for. Brand recognition aka mindshare still belongs to Nvidia. They will continue this strategy until they see it doesn't work. Why wouldn't they?
Click to expand...

Thanks for proving my point.. :)
 
Eshelmen said:
I'm just curious, about how long before release do GPU manufacturers start producing product? If the 3000 series is scheduled for what, August/Octoberish(?), when would they have to decide when to start making them?
Click to expand...
It's different for everything, but I am going to guess at least 8 months to build up enough inventory for release, meaning NVIDIA have probably been making product since the fourth calendar quarter last year.
 
Steam numbers in a few day will bring some data on the table...a fluf debate with no facts are just that...fluf.
 
djoye said:
Is the Ti series usually that more robust?
Click to expand...
It only has meaning for the top-most consumer / gaming model, and, well, usually. Of course, that delta is engineered, as are all of the product levels below it.
 
djoye said:
Another article with rumored specs: https://www.guru3d.com/news-story/rumor-nvidia-geforce-rtx-3070-and-3080-coming-q3-2020-specs.html

Is the Ti series usually that more robust? Regardless, I'm definitely going to buy at least a 3080 since I'm due for an upgrade. I'm considering splurging for a Ti model, but hoping to find a beefed-up AIO water cooled card since my Gigabyte GTX 1080 AIO water has been solid.
Click to expand...

The TI won't come out until a quarter later than the plain 3080 according to that so you'll have to wait longer. I don't want to wait that long. I also really want an AIO water cooled card but it seems like they aren't available until a few months later than the founders cards... and I really don't want to wait any longer. I just want HDMI 2.1 so I can run 4k 120hz gsync on an LG CX.
I'll probably end up buying a founders 3080 direct from nvidia.
 
  • Like
Reactions: Auer
like this
sharknice said:
The TI won't come out until a quarter later than the plain 3080 according to that so you'll have to wait longer. I don't want to wait that long. I also really want an AIO water cooled card but it seems like they aren't available until a few months later than the founders cards... and I really don't want to wait any longer. I just want HDMI 2.1 so I can run 4k 120hz gsync on an LG CX.
I'll probably end up buying a founders 3080 direct from nvidia.
Click to expand...
I want a AIO as well, 3080/3070. Isnt EVGA usually pretty fast getting theirs out?
 
I dont like waiting for anything usually you cant see any graphical improvement in new cards games but you could with Ray Tracing. Maybe Nvidia will add some super effect like PhysX 3.
 
Auer said:
I want a AIO as well, 3080/3070. Isnt EVGA usually pretty fast getting theirs out?
Click to expand...

I bought a 1080ti founders edition a month after the initial release and no water cooled models were available. I don't think they even started selling water cooled versions yet and all of the air cooled 3rd party models were sold out. I was lucky to even get a founders edition because it was usually sold out too.

I eventually bought an EVGA hybrid AIO diy kit and the end result is great but I do not want to do that myself again. There were like 100 of the tiniest screws several of which i stripped. It was several hours of tedious work not really knowing what I was doing and hoping I don't brick a $1000.
 
sharknice said:
I bought a 1080ti founders edition a month after the initial release and no water cooled models were available. I don't think they even started selling water cooled versions yet and all of the air cooled 3rd party models were sold out. I was lucky to even get a founders edition because it was usually sold out too.

I eventually bought an EVGA hybrid AIO diy kit and the end result is great but I do not want to do that myself again. There were like 100 of the tiniest screws several of which i stripped. It was several hours of tedious work not really knowing what I was doing and hoping I don't brick a $1000.
Click to expand...

I always wait for AIBs. I'd only buy FE if I was open loop and fitting them with blocks.
 
sharknice said:
I eventually bought an EVGA hybrid AIO diy kit and the end result is great but I do not want to do that myself again.
Click to expand...
Furious_Styles said:
I always wait for AIBs.
Click to expand...
It's worth the wait, IMO. One vendor to ensure QA and warranty and so on.

I also don't want to buy another top-end card without an AIO. It's not the absolute quietest solution, but it's the least worry when it comes to keeping it stable under load.
 
If the specs are true then I may upgrade my EVGA Hybrid FTW 2080 Super
But only if the performance gains are there for VR. It's worth waiting for the AIO ones. So much quieter. No worry on heat
 
djoye said:
Another article with rumored specs: https://www.guru3d.com/news-story/rumor-nvidia-geforce-rtx-3070-and-3080-coming-q3-2020-specs.html

Is the Ti series usually that more robust? Regardless, I'm definitely going to buy at least a 3080 since I'm due for an upgrade. I'm considering splurging for a Ti model, but hoping to find a beefed-up AIO water cooled card since my Gigabyte GTX 1080 AIO water has been solid.
Click to expand...

Hmmm that's a pretty big jump between 3080 and 3080Ti.
Looks to me like Titan.
I won't have money for that. Unless it's <€2000 (not likely).
I can however see Nvidia doing this to keep 3080Ti price sky high.
Something like €700 3080 and €1500 3080Ti

3080 performance could be compared to big Navi, while 3080Ti is needed to be the absolute king in both price and performance.
But still, where does Titan fit in there? Probably the top specs, max everything from that chart while 3080ti is somewhere in between.

But regardless, I'll wait on AIO 3080Ti, so probably no card for me until Christmas :)
 
Last edited:
IdiotInCharge said:
It's worth the wait, IMO. One vendor to ensure QA and warranty and so on.

I also don't want to buy another top-end card without an AIO. It's not the absolute quietest solution, but it's the least worry when it comes to keeping it stable under load.
Click to expand...

I like the AIOs, have never owned one yet for the GPU. The extra mark-up on them sometimes makes me reconsider. I might get one with a 3080 ti, right now I have a kraken modded 2080 ti and it's nice.
 
IdiotInCharge said:
It's worth the wait, IMO. One vendor to ensure QA and warranty and so on.

I also don't want to buy another top-end card without an AIO. It's not the absolute quietest solution, but it's the least worry when it comes to keeping it stable under load.
Click to expand...
Another decent option.
Accelero Xtreme air coolers are very quiet and give a halfway house on clockspeed vs full on water cooling.
They are also very cheap.

When the 1080ti was released I bought a Founders Edition at launch and whacked the Xtreme III on it long before aftermarket cards surfaced.
Max temp is around 55C (at the new max overclock) usually lower temp, and its rock stable. I'm still using it.
Its so quiet the fans are on full speed permanently, 24/7, in the living room.
A great performing solution needing almost no maintenance, only an occasional hoover clean.

The only downside is its a hefty beast.
2 friends PCs had to have a notch cut in a 3.5" drive bay to let it slide in.
My mobo is horizontally mounted inside a table without any PC case so it mounts easily.
 
imsirovic5 said:
Interesting rumor - ray tracing up to 4x faster on Ampere, potentially placing 3060 ray tracing capabilities to those of 2080 Ti:

https://www.pcgamer.com/nvidia-ampere-turing-not-aging-well/
Click to expand...

Yet another nonsense rumor. The source is another YT nonsense channel.

More than half the cost of RT effects come from traditional compute units. For a 3060 to have 2080Ti RT capability, it would also have to have 2080Ti Raster capability as well.
 
I was looking into an AIO at the time of the 1080ti launch as well, real bummer with delays, and eventually limited quantity releases with major price gouging (IMHO) on the AIO. I learned my lesson there that unless there's high availability, competitive prices and better builds I'd be better with an air cooled card at a decent price/perf ratio, and potentially installing a cooler myself.
 
Snowdog said:
Yet another nonsense rumor. The source is another YT nonsense channel.

More than half the cost of RT effects come from traditional compute units. For a 3060 to have 2080Ti RT capability, it would also have to have 2080Ti Raster capability as well.
Click to expand...

Just curious, why would it be outside of realm of possibilities for Nvidia to engineer better / more efficient RT cores that need less traditional compute units for the ray tracing?
 
imsirovic5 said:
Just curious, why would it be outside of realm of possibilities for Nvidia to engineer better / more efficient RT cores that need less compute units for the ray tracing?
Click to expand...
Because according to devs, the raytracing portion of frame times is minimal. It's the shading that eats up extra performance as it does its passes.
 
GoldenTiger said:
Because according to devs, the raytracing portion of frame times is minimal. It's the shading that eats up extra performance as it does its passes.
Click to expand...

That is interesting, so if this is a fact (not arguing that it is not) I wonder why there are so many different rumors (from a lot of sources) stating that ray tracing performance will be way more improved / faster in Ampere.
 
geforce-rtx-gtx-dxr-one-metro-exodus-frame.png

x-dxr-metro-exodus-rtx-rt-core-dlss-frame-expanded.png

I would imagine that the values would vary depending on the scene/game/engine/optimizations used.
 
imsirovic5 said:
That is interesting, so if this is a fact (not arguing that it is not) I wonder why there are so many different rumors (from a lot of sources) stating that ray tracing performance will be way more improved / faster in Ampere.
Click to expand...

It probably will be, but not 400% more faster.

Even if Nvidia could do something better I doubt they will as they'd like to keep ray tracing a premium feature for as long as possible. They'll cut down specs and keep pricing high accordingly to stretch the profits out as long as possible.
 
imsirovic5 said:
That is interesting, so if this is a fact (not arguing that it is not) I wonder why there are so many different rumors (from a lot of sources) stating that ray tracing performance will be way more improved / faster in Ampere.
Click to expand...

Because lots of people want to hear that, and rumors tend to tell people what they want, to attract more attention.

People want to believe huge gains in RT performance are right around the corner, or that RT won't even cause a performance hit...

Ever notice a rumor cycle dominated by rumors of tiny incremental performance gains?
 
jologskyblues said:
View attachment 243092

View attachment 243094

I would imagine that the values would vary depending on the scene/game/engine/optimizations used.
Click to expand...

I can't find it right now, but there is an even better example with multiple RT effects highlighted.

The anatomy of a typical RT effect is one part Intersection Testing (RT cores - the actual tracing of rays) one part shading those results (traditional cores) and one part denoising (contrary to NVidia arguments, usually traditional cores). More of that RT effect frame time was spent dealing with the output of RT cores, than actually in the RT cores.

So most RT effects spend most of their time in traditional HW, and that's just the RT effects. Hybrid rendering is here for the foreseeable future, so all the non RT effects are on traditional cores as well. Making the traditional cores much more of a bottleneck than RT cores are.

RT HW is a small part of frame time. For example it might only be 10% of frame time.

Forget 4 times faster. Even if you made RT cores 100X faster, you would still increase frame rate by less than 10% in this example.

This is just a bogus rumor made up by someone that doesn't understand how Ray Tracing works in games.

It also highlights why spending precious die space, mostly on RT HW boosting has poor payback. A balanced approach makes much more sense given how much more traditional HW affects frame rates. You want enough RT HW so it isn't the bottleneck. That has already been accomplished, to the point that Traditional cores are more the bottleneck now.

That said, I can see NVidia improving the relative balance of RT performance, if they can squeeze in a bit more efficiency, but it wouldn't be sensible to spend die space heavily on RT cores, when the bottleneck is elsewhere. They need a balanced performance increase.
 
Last edited:
You must log in or register to reply here.
Back
Top