• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Possible RTX 3080 Ti FE?

Yeah this is some unintended absurdity right now. I also think the Green/Red fanboy thing ends up being academic and mostly a thing of the online discussion world, because people quickly stop seeing colors when the performance is there from both sides.

In other words I suspect a lot of people that have been waiting on a 3080, and may have even bought Nvidia for the past several gens, would gladly buy a 6800XT if one was more easily available.

3080 or 6800XT, I'll personally buy whatever becomes available to me on Amazon first.

I'm pretty much in the same boat with a slight leaning toward a 6800xt this refresh.
 
Were you around 8 years ago saying the same thing about 4gb of ram?

4 years ago saying the same thing about 6gb of ram?

2 years ago saying the same thing about 8gb of ram....

And so on... Yea sure by the time there are several titles that can use more than 10gb of video memory new cards may be out. But for those that WANT to use that... that will run MODS in whatever fallout/starfield/whatever game then they will care.

And clearly vram helps sell video cards.
I've been PC gaming since before I built, from salvaged parts, a 386 DX.

So yes, I've been around. I've seen people pissing and moaning about it. Good for them for wanting more, and good for people who try and help the majority of users make an informed decision.

To the minority that don't care about evidence, or tools that can now show them the facts - good bye.


Back on topic - I really hope that there will be a 3080 Ti with 20GB of VRAM for $999. Competition is good.

Let me say that again for the cheap seats: COMPETITION IS GOOD FOR THE CONSUMER.
 
I got my 1080 ti with 11gb of vram in December 2017. This was the Evga SC2 Hydro Copper card (pre installed waterblock) and it was $840. Three years later and I'm looking to pay more for less vram, it's not a great situation for me. I was really hoping they would reconfigure the card to allow for a 16gb s / ti version, but at 1k likely msrp + a likely big waterblock tax, I may just keep waiting. Not looking to enter 4k, but my 1440p display can run up to 165hz, so I really would prefer not to downgrade. The display being gsync also locks me into nvidia so I'll just keep on waiting.
 
So a 3080 Ti with 20gb that has the same or less performance of a 6800 XT with 16gb that cost $350 more. Gee, if I was solely a gamer, I wonder which one I would buy? Duh. I can see if one needs it for Cuda, must have that extra RT mostly for professional programs that use it but for a purely gaming card the price is stupid. $899 to me would be max reasonable price. Then again all these graphics cards will come down to availability and the winner is going to the one that make the most cards so that people can actually put their hands on one.
 
Valid point that if you want to move from a 1070Ti to something you can keep for a couple years, then you might be better off waiting for Radeon 6700/6800. However, I think the idea you can be "future proof" when it comes to something as rapidly evolving as GPU's.. is a fallacy. It's just always going to be a compromise.

Now that 10GB 3080's are out in the wild, AAA devs will be catering to them for their highest settings, ensuring 10GB doesn't cause performance problems. However this gen will definitely be the last time we see 10GB for Nvidia xx80 and in future it will not be enough. 4080/Hopper will get somewhere between 12-20, I reckon, depending on how the bus width math and GDDR7 supply pricing shakes out.

Well the second challenge I have is that I have a G-Sync monitor lol!

But yes, I don’t think it’s unreasonable to expect some degree of longevity out of a $700 investment. I’m not saying the 3080 is garbage or will deliver a bad experience or anything like that. I’m saying that if the trend in gaming right now is that textures at 4K will be demanding more than 10GB of VRAM in the relatively near future, which seems to be the case, then Nvidia should have packed more than 10GB on a flagship card. If you’re the type of person who upgrades every two years, then who cares. I’m not that type of person though.

What you’re saying about future proofing is true, which is why I don’t typically like that term, but there is something to be said about designing a flagship product with foreseeable trends in mind. 8GB was the standard for years. We’re seeing that that is no longer sufficient for ultra settings at 4K, but 10GB is right on the edge for many titles. It seems strange to step back here given that Nvidia’s had 11GB on their top cards for several years now.

In any case, if AMD’s RDNA2 benchmarks are accurate, and that remains to be seen, then it would be impossible to escape the conclusion that they’re offering more for less on the high end.
 
What you’re saying about future proofing is true, which is why I don’t typically like that term, but there is something to be said about designing a flagship product with foreseeable trends in mind. 8GB was the standard for years. We’re seeing that that is no longer sufficient for ultra settings at 4K, but 10GB is right on the edge for many titles.
Looking quickly (maybe too quickly) 8 gig seem to be more than enough for 2 of the new really hard to run game Assassin Creed Valhalla and Watch Dogs Legion at 4K ultra setting (and could indicate that game are really focused to run on the new console what the future could look like), 3-4 gb of ram and some but not all of the 6 gig card having issue with Watch Dogs and 6 gig card Ok with Creed, even Flight Simulator 4K ultra setting we do not see the 11+ gig card jump ahead of the 8 gig one (at least not with any consistency).
 
LOL, I mean thats great that they planned to launch this at $999 to go against the 6900XT. But, does it even matter? It will have the same terrible stock at the 3080 and 3090.....Whats the point?

Nvidia just needs to stop releasing new card with names and try to up the stock of the old stock lol....
 
Since so many people, myself included, figured the 3080 20GB model would be priced somewhere between $900-$1,000+, having the TI come in at $999 means that it's the 20GB model with some extra CUDA cores for the higher end of the contemplated price range. Though, I'd of course expect these chips to be even worse overclockers than the 3090 since there must be some binning going on.

Hopefully this will make AMD reconsider the pricing of the 6900XT which only seemed "reasonable" in light of Nvidia massive overcharge of $1,500 for the 3090. Comparing the expectations of both cards, AMD probably has a slight edge in raster performance, overclocking (maybe), and total power draw. However, the 6900XT loses against the 3090 in onboard memory, ray tracing, DLSS fakery, and presumably driver support. Yes AMD has its Smart Access feature, but because it turns out that is seemingly baked into the PCIe spec, Nvidia says its coming out with a competing version of its own.

We all know that the 6900XT is just the 6800XT with 8 extra CUs enabled... unless all the 6900s are golden samples as well in power efficiency and the 6900 boards are built with significantly pricier and more capable components.

When looking at probable the wins and losses, the 6900xt really not worth the money compared to the expected 3080ti if it really does come with all of the 3090's CUDA and 20GB of VRAM and the prospective buyers are at all interested in ray tracing this gpu generation.
 
Sweet. I'll wait for the 20GB version and see how it compares. $999 is exactly the price I was expecting given the cost of GDDR6X.

I assume Nvidia is sticking with Samsung for fab?
 
When looking at probable the wins and losses, the 6900xt really not worth the money compared to the expected 3080ti if it really does come with all of the 3090's CUDA and 20GB of VRAM and the prospective buyers are at all interested in ray tracing this gpu generation.
Yeah, the 3080 Ti would be an easy choice for me if it's priced the same as 6900 XT.

This competition is great for consumers. Once availability stabilizes (not for a few months I'm sure) we may even see competition start to drive prices down!
 
June 2018 Interview with Phil Specter where Hardware development was well under way. You say below that gpus are currently a multi year process, it would be impossible to react to RTX in Fall of 2018 and get both MS and Sony to agree to redesign their consoles that late in the game
https://www.theverge.com/2018/6/12/17453174/microsoft-xbox-next-genreation-2020-release-date-rumors
Of course it was. Doesn't mean set in stone, point stands.
Again, wrong. The Desktop version of adaptive sync was finalized in March of 2014, but even in the white paper announcing it they bring up the fact that

"Variable refresh rate technology has been available to Notebook PC makers for quite some time as a system power saving feature for embedded notebook panels.

https://www.vesa.org/wp-content/uploads/2014/07/VESA-Adaptive-Sync-Whitepaper-140620.pdf

The idea was absolutely available in laptops long before Nvidia debuted the idea in desktops. Yes, it's intention was power savings, but the idea certainly wasn't an original Nvidia one.
You are just proving my point for me. From the whitepaper you linked, regarding the older refresh rate control functionality used in laptops:
Helps improve battery life by reducing the refresh rate of the panel when the screen is static

Yep, nothing to do with gaming, with frame time syncing.

Goal post move. GPU is synonymous with video card, 3dfx was the first company to pair multiple video cards together for extra performance, and buying the company that did it first isn't the same as inventing it.
You said GPU, now who is moving the goalposts. My original point: "multi-gpu tech", again, point stands, but nice try.
More goal post moving. You said it was important that Nvidia had the first T&L acceleration in a GPU because it showed leadership. It was the next logical path, and Ati was on the same path at the same time. Now it's about who did it faster.

Release times were the same, approx every year. It's only in the last generation that release times have slipped to two years.

https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units

https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units
We can give AMD the benefit of the doubt, and assume that hardware T&L was already planned. But I think it was tacked on after Nvidia's card came out that included it. What other explanation is there for it's absolutely horrible performance? It took ATI over a year to catch up... so sure, they planned it all along..

mm hmm. Both of our points of view can be considered speculation. I'm either correct, or you are and ATI's product was shit. Pick the lesser of the evils and believe what you want. Nvidia was first with this tech, and it performed quite well. Almost like they planned it in their GPU... (and ATI copied it, poorly)
How about you tell me what use Tensors have on desptop GPUS? GPU development costs are high, and dev times are long, it only makes sense to reuse what you can.
Speculation. It doesn't really matter if they are re-purposing some processing units for a new function. The point was that they implemented something new and innovative, and it has a real impact in moving GPU tech forward. Again, point still stands. Nvidia did this first, AMD is following.
I'm glad that you have insight into nvidia that non-employees don't, feel free to fill us in on your particular set of facts, or are they just the opinions? We've ben held rapt here with your historical handjob on Nvidia, it's practically "just buy it" levels of love from your corner.

Also, check the sig. My last system was an intel build, and I'm running an Nvidia GPU. I'm just not sitting here waiting for what they're looking to give me.
Could care less that you have an Nvidia GPU.

At this point you are just being a troll. Good bye.
 
I got my 1080 ti with 11gb of vram in December 2017. This was the Evga SC2 Hydro Copper card (pre installed waterblock) and it was $840. Three years later and I'm looking to pay more for less vram, it's not a great situation for me. I was really hoping they would reconfigure the card to allow for a 16gb s / ti version, but at 1k likely msrp + a likely big waterblock tax, I may just keep waiting. Not looking to enter 4k, but my 1440p display can run up to 165hz, so I really would prefer not to downgrade. The display being gsync also locks me into nvidia so I'll just keep on waiting.
Just wait for reviews of 4k games and decide from that. There's been little evidence that 16Gb is any better in game performance than 10Gb. (Workstation workloads or GPU virtualization is another story). Radeon VII had 16Gb vram, how much future proofing did that provide? Future proofing doesn't really exist. Whatever you buy, in a year something faster is going to come out, that is just the nature of the beast. With the ever increasing prices, I can see why anyone would want to stretch the life of their investment.
Well the second challenge I have is that I have a G-Sync monitor lol!

But yes, I don’t think it’s unreasonable to expect some degree of longevity out of a $700 investment. I’m not saying the 3080 is garbage or will deliver a bad experience or anything like that. I’m saying that if the trend in gaming right now is that textures at 4K will be demanding more than 10GB of VRAM in the relatively near future, which seems to be the case, then Nvidia should have packed more than 10GB on a flagship card. If you’re the type of person who upgrades every two years, then who cares. I’m not that type of person though.

What you’re saying about future proofing is true, which is why I don’t typically like that term, but there is something to be said about designing a flagship product with foreseeable trends in mind. 8GB was the standard for years. We’re seeing that that is no longer sufficient for ultra settings at 4K, but 10GB is right on the edge for many titles. It seems strange to step back here given that Nvidia’s had 11GB on their top cards for several years now.

In any case, if AMD’s RDNA2 benchmarks are accurate, and that remains to be seen, then it would be impossible to escape the conclusion that they’re offering more for less on the high end.
Same advice. Just wait for reviews of all these new cards, and if you think 4K gaming is in your near future, look at the reviews covering 4k resolutions. But there is nothing in pc gaming that is future proof, save for maybe a 1500W psu. If for you 4k gaming is more than 2 years out, even when that time comes and you keep the GPU you buy now, you will have to make some compromise for performance. Not a big deal, just turn down one or two sliders in the game settings, and you will have acceptable performance. Future proofing, especially in regards to video cards, is a unicorn. Just make adjustments to keep FPS reasonable. People are still gaming on 1080Ti's just fine, and those are what, 4 years old? That GPU is the closest thing to a unicorn I think I have ever seen in GPU's, but you still need to turn some settings to medium to play on it. Or just stick to 1080p resolution, then everything else can be maxed out.

As fast as the PCIe bus is, having to swap textures into VRAM more often (due to size) isn't as big of an issue as it was when video cards had 512Mb vram. Even with 16Gb vram, or 24Gb, there are going to be instances when the data in the VRAM has to be loaded or swapped. It's what 1 or 2% performance? These days game engines are really good at memory management, which is why even a 3080 with "only" 10Gb vram still runs amazingly fast, even at 4k. With enough memory bandwidth, you are not going to notice if it has to swap textures every 5 seconds vs every 20 seconds. The swaps happen faster than ever on today's cards.
 
Whelp, looks like Godfall doesn't really need that 12GB after all: https://www.techpowerup.com/review/godfall-benchmark-test-performance-analysis/4.html

vram.png
 


Considering that it is a game where the 8 gig 3070 is clearly ahead of the 2080ti (well by 10%) of the 11 gig 2080TI and the 2070 8 gig ahead of the 1080 GTX with 11 gig and that the 6 gig 1660 TI stay has close to the 1070-1080 than at the other resolution and that the 6 gig 2060 do not move much from the 8 gig Vega 64, i am starting to wonder if:
Average FPS do not fall over a cliff except if you are missing a significant amount and 1%-0.1% would be needed in those type of analysis, a bit like compressing what we put in memory or using swap is less dramatic with fast large cache HDDor ssd/many cores, missing VRAM has talking to the GPU ram get faster and regular ram get faster is less of an issue maybe.

Or those VRAM usage are not fully actual necessitate usage, according to this only 3-4 gig VRAM card got particularly bad performance with that game, nothing special with the 6-8 gig cards (average FPS that is, worst frame could reveal something here).

Edit: I Just read the message just above, that explain it better.

Looking at this video:


It seem the game never really goes above much above 6 gig of dedicated VRAM at 4K Epic Settings (in that video it peak at 6.5, but with HDR and other extra maybe it achieve those numbers above), maybe that is more why only the 3-4 gig video cards get a performance hit on that title.
 
Last edited:
Back
Top