Is Ray Tracing good for anything non-gaming?

Peat Moss

Gawd
Joined
Oct 6, 2009
Messages
543
I'm upgrading and trying to decide between a 1660ti and a 2060. I don't game, but I occasionally use navisworks for some light 3D modeling (but not nearly enough to justify buying a Quadro card).

If I get the 2060, would the ray tracing feature be good for anything non-gaming related? e.g. Will it help with CAD or 3D modeling?
 
Another advantage of 2060 over 1660Ti for non-gaming that I can see is machine learning.
Addibng hardware accelerated RT support in to productivity software might take some time, especially since this feature requires low level APIs like DX12 or Vulkan.

Personally I would suggest getting GPU that supports newest features since price difference between 16600Ti and 2060 is not very high and you get about 20% higher performance. I am not sure if you will be able to utilize it in Navisworks though.
 
I'm upgrading and trying to decide between a 1660ti and a 2060. I don't game, but I occasionally use navisworks for some light 3D modeling (but not nearly enough to justify buying a Quadro card).

If I get the 2060, would the ray tracing feature be good for anything non-gaming related? e.g. Will it help with CAD or 3D modeling?

Eventually expect support. Since 3D rendering is essentially Ray Tracing, eventually Ray Tracing HW will accelerate these operations, see above where there is early alpha support in Blender.

The other advantage of the RTX cards is Tensor cores. Potential use cases for AI processing is wider, but largely untapped outside of gaming right now.

If it isn't much of price delta, it can give some additional technology to dabble with moving from 1660Ti to RTX 2060, though the 6GB frame buffer could be something of a limitation with the usage of Ray Tracing hardware as this tends to be memory intensive...
 
Thanks. So it sounds like a 2060 would be useful for a few other applications, and for future proofing.

Are 2060 prices expected to go down this year? What would be a good price to look for?
 
In terms of CAD/3D Modeling utilizing pure ray-tracing (path-tracing), NVIDIA recently launched Omniverse Platform [1] [2] to early adopters in their developer program.

Though to answer your question more directly, ray-tracing hardware in RTX doesn't speed up anything not related to ray-tracing.
 
Thanks. So it sounds like a 2060 would be useful for a few other applications, and for future proofing.

Are 2060 prices expected to go down this year? What would be a good price to look for?

I wouldn't expect any more falls in price until around October. Right now, Q1 demand is higher than it's ever been (because of Covid bored gamers are building new systems.)

If you're willing to wait for October, they should have Ampere based cards available. They should be cheaper for similar performance (or you can catch Black Friday clearance of the 2060?)
 
I wouldn't expect any more falls in price until around October. Right now, Q1 demand is higher than it's ever been (because of Covid bored gamers are building new systems.)

If you're willing to wait for October, they should have Ampere based cards available. They should be cheaper for similar performance (or you can catch Black Friday clearance of the 2060?)

Ampere cards will be cheaper than current Turing cards?
 
Ampere cards will be cheaper than current Turing cards?


Why wouldn't they be? Last time I checked they were 7nm. You don't see AMD charging a inordinate amount of money for Navi cards.

Expect similar launch prices to Pascal ($500 for RTX 3070, $700 for RTX 3080,) then expect them to drop hard once yields improve.

The 30-40% improved performance should be enough to drop prices on Turning immediately, and if you can wait a few months (February usually,) they should fill-out the rest of the product range (3060.).
 
Last edited:
Because Turing cards were more expensive than Pascals when they came out. And the RTX Quadro cards were more expensive than the Pascal Quadro cards when they came out.

That's why I expect the Ampere cards to be more expensive than the Turing cards when they come out.
 
Because Turing cards were more expensive than Pascals when they came out. And the RTX Quadro cards were more expensive than the Pascal Quadro cards when they came out.

That's why I expect the Ampere cards to be more expensive than the Turing cards when they come out.


When you get the record Consumer die sizes of Turing, you get a bit expensive!

I would expect Ampere to shrink the 3080 die size down to around 400-450 mm² This should mean similar launch prices to Pascal

In addition, the GDDR6 memory is no-longer exotic, so that should drop prices.

I would expect $100 lower launch p[rice for both the 3070 (now at $500) and 3080 (now at $700) versus Turing. And after thye release the3 3080 Ti, that should drop the two cards another $50-100.

The smaller die sizes should mean we go back to sourcing the 3070 card from cut 3080.

the 106 card will continue to off3er raytracing, as well as the 107 added RT units (bus-powered)
 
Last edited:
Streaming and video encoding you get a notch up in quality.
The specific tools you use will have varying degrees of Turing optimizations.
If you analyze datasets on a local workstation, Turing will reduce your training time.
I’d argue that brute force dataset reductions are a sign of laziness or lack of experience but people love throwing resources at jobs.
 
You don't see AMD charging a inordinate amount of money for Navi cards.
Actually Navi cards cost almost as much as Turing cards for the same class of performance.
There is small price difference but it is insignificant and definitely shows that AMD adopted higher pricing...

I would be really surprised if prices of new series from both AMD and Nvidia were cheaper. I do not think prices will be higher either.
 
I would expect $100 lower launch p[rice for both the 3070 (now at $500) and 3080 (now at $700) versus Turing. And after thye release the3 3080 Ti, that should drop the two cards another $50-100.

It's a dream world to just expect prices to drop. They pretty much never do that.

Perf/$ will improves, but I expect each series (x60, x70, x80) will maintain, if not increase their price.
 
Back
Top