Anyone planning on getting the 3090 for VR?

[imacken]

That really doesn't have anything to do with my post here saying that I'm going for the 3090 based on my understanding that VR benefits from additional VRAM (I've seen it talked about elsewhere with real-world testing being done on DCS, with the Titan RTX). What this boils down to is that the 3090 isn't for you, and there's nothing wrong with that. Not trying to sell anybody on anything, just sticking to the subject.

 

Sorry, but you are wrong there. My intention is almost certainly that I will try and buy a 3090 next week. Also, my VRAM point is very relevant to the 3090 discussion as I feel there are a lot of urban myths kicking around this whole topic.

 

However, until we see some VR benchmarks next week, no one really knows what effect 24Gb of VRAM will have on DCS VR performance.

[eaglecash867]

 

However, until we see some VR benchmarks next week, no one really knows what effect 24Gb of VRAM will have on DCS VR performance.

 

:thumbup:

 

That I can agree with. So, I guess we can stop with all of the theorizing based on incomplete data? :D

[imacken]

:thumbup:

 

That I can agree with. So, I guess we can stop with all of the theorizing based on incomplete data? :D

 

:thumbup:

[Supmua]

I agree with this, i was ready to spend about that much on an old 2080ti, when i held because of the launch dates. Im still ready to spend it on a 3090.

 

The question, is it worth while holding for the 20Gb version, and as yet nobody has posted on actual VRAM experiences... it will be fascinating because if correct the 20gb will have the same number of Cuda cores as the 10Gb version an d we can actually do some perf comparisons.

 

But we dont really have any data on memory impact except with lower end cards in 2d. A Reverb is essentially a 4k+ display, so the 4k numbers are super relevant, but none of the games tested are running on DX11 (Rainbow 6 maybe) and none in VR ... we lack hard data ... so we have a bunch of people speculating...

 

Price/performance wise, the 3080Ti with 20GB VRAM will be better I have no doubt--but who knows when that is going to happen, not to mention availability. Knowing NVIDIA, they'll want to deplete the existing 3080/3070 stocks first then release the Ti variants later. I also understand perfectly why they don't want to just release GPUs with higher VRAM--to keep the price low enough to be competitive with AMD.

 

If you're undecided the best thing to do is just wait as I'm pretty sure several people here will buy the card on release day. Then we'll get to see some DCS benchmarks. I'm not super excited about the 3090 but it is something that has been on my to do list (always wanted a Titan class GPU, almost bought one last year). If it is widely available on release day I'll pull the trigger and move on. If not I'll simply wait.

[Supmua]

Yet another leaked benchmarks.

 

It is encouraging that this is the only card that achieved >60 fps in MSFS at 4K high in this testing, 64 vs 45 compared to the 2080Ti.

[Waxer]

Yet another leaked benchmarks.

 

It is encouraging that this is the only card that achieved >60 fps in MSFS at 4K high in this testing, 64 vs 45 compared to the 2080Ti.

 

The video is 95% unless. All he does is multiply 3080 results by 20% on the basis of synthetic benchmark results. Meanwhile a Chinese site has leaked in game results at 4K in a Z490 / OCed 10900K system. Bottom line is 4-10% better fps in a selection of AAA games. Not 20%.

[wormeaten]

Pay 114% more for just 20% gain?!

 

 

 

Do I have to comment this at all?

[imacken]

Pay 114% more for just 20% gain?!

 

 

 

Do I have to comment this at all?

 

Well, yes. For lots of people - myself included - it's about performance not cost.

[Waxer]

Pay 114% more for just 20% gain?!

 

No. Your maths is wrong:

Pay 114% more for 4-11% gain in games. :megalol:

 

https://videocardz.com/newz/nvidia-geforce-rtx-3090-gaming-performance-review-leaks-out

 

20% is just for synthetic benchmarks.

 

And, no. You don't need to say anything.

[Gunnars Driver]

Well, yes. For lots of people - myself included - it's about performance not cost.

 

Same here.

 

Actually 3080 coming from a 2080ti feels like a lot of money for a small gain, and probably ”need” to upgrade earlier next time compared to a 3090.

[BeastyBaiter]

One could simply wait 6 months for the 3080 Ti or Super or whatever they are going to call it. Nvidia always does it that way. Probably going to just replace the 3080 entirely, much like the 2080 Super did to the 2080. I can say that pretty confidently given that the 3xxx pricing is basically identical to the 2xxx pricing, so I expect we'll see the same thing again. There is also AMD floating around, I doubt they'll compete at the 3080 level given their current GPU situation, but we don't really know yet.

[Harlikwin]

No. Your maths is wrong:

Pay 114% more for 4-11% gain in games. :megalol:

 

https://videocardz.com/newz/nvidia-geforce-rtx-3090-gaming-performance-review-leaks-out

 

20% is just for synthetic benchmarks.

 

And, no. You don't need to say anything.

 

So can someone explain to me why going from 4352 cuda cores, on a 2080ti, to like 10k cores on the 3090 basically has no impact on actual FPS performance? Or doubling the RAM? Is it a CPU limit at that point for DCS or other games?

[BeastyBaiter]

Nvidia redesigned their cores, so 10k of the new cores are not equal to 10k of the old ones. What they've done reminds me of the old AMD FX CPU's where they in theory had 8 cores but had a built in bottleneck so it was really only 4 at best.

[Harlikwin]

Nvidia redesigned their cores, so 10k of the new cores are not equal to 10k of the old ones. What they've done reminds me of the old AMD FX CPU's where they in theory had 8 cores but had a built in bottleneck so it was really only 4 at best.

 

Ah I see, marketing Gamesmanship. Greeeeat.

[Waxer]

It is not quite as simple as that. In DCS there will always be a bottleneck somewhere. The simplistic statement that many people make is "the CPU is the bottleneck" or "the GPU is..." or whatever.

 

In some people's system - where components are not well optimised - it might well be one component of the PC that is consistently the bottleneck. In a well optimised system the bottleneck will be moving from component to component depending on what you are doing at any particular time. And it is generally true that on most people's systems are more often towards the CPU part of the spectrum, or something related to it like memory bandwidth (feeding information to the CPU from memory). But the more you travel from 1080p, to 1440p, to 4K, to 5K (my setup) to 8K the more you shift the bottleneck towards the GPU more of the time.

 

(There is some guy on here that keeps droning on about CPU bottlenecks, but while he is not wrong in general terms, many recent threads have been focusing on the GPU bottleneck. So I feel his repeated comments about the CPU are singing a song that a lot of people already know the words to). I digress.

 

If your CPU is 100% of the time the bottleneck (at your monitor resolution and settings) you can move from a 980 Ti to a 3090 and it will literally do nothing for you. That is one possible reason why could be seeing no GPU scaling.

 

On the thing that the OP just mentioned - Ampere CUDA cores are not equal to Turing CUDA cores - this is true. But Nvidia have a very good and legitimate reason for counting CUDA cores as they do, independent of the marketing spin. For details listen to this, but be warned that it gets rather technical. The long preamble is partially necessary to listen to understand the part where he starts to explain the difference with Ampere. Otherwise you will not understand what he is talking about.

 

[dburne]

One could simply wait 6 months for the 3080 Ti or Super or whatever they are going to call it. Nvidia always does it that way. Probably going to just replace the 3080 entirely, much like the 2080 Super did to the 2080. I can say that pretty confidently given that the 3xxx pricing is basically identical to the 2xxx pricing, so I expect we'll see the same thing again. There is also AMD floating around, I doubt they'll compete at the 3080 level given their current GPU situation, but we don't really know yet.

 

Interestingly for the 20xx series of cards they introduced the 2080 Ti at the same launch of the 2080 cards. Obviously they are not taking that route this time, but they have left a gap seemingly for it so I am sure it will come.

 

I am half tempted to hold off on the 3090 and wait for it, being as how cards are so hard to get at launch this go around.

[Waxer]

In terms of CUDA cores / the GPU chip itself, I don't think Nvidia have anything between the 3090 and 10GB 3080. I think the only thing in between will be the 20GB 3080. And I don't think that will be a six month wait. More like a 1-2 month wait. Probably either just before the 6900 XT launch (to spoil the launch) or soon after once they know the pricing of the AMD part.

 

And we already know there is something over and above the 3090, except that it will cost >$4,000 and not be supported by game drivers.

[BeastyBaiter]

Interestingly for the 20xx series of cards they introduced the 2080 Ti at the same launch of the 2080 cards. Obviously they are not taking that route this time, but they have left a gap seemingly for it so I am sure it will come.

 

I am half tempted to hold off on the 3090 and wait for it, being as how cards are so hard to get at launch this go around.

 

 

For the 20xx series I'm thinking of the super line. So they came out with the 2060, 2070, 2080 and 2080 Ti. After a few months, they released the "Super" versions of each. These followed the typical approach of releasing a "Ti" card but they named them "Super" instead. The only outlier being the 2080 Ti instead of being called a Titan. Given Nvidia has been doing this pretty much since Nvidia was founded, I expect we'll see Ti or Super versions of all these cards in the not too distant future. Given the current price brackets chosen, I also think they'll do like they did with the 20xx series and simply replace the old cards with them at the same price.

[Snake122]

I too bet that the increased VRAM versions are coming much sooner than the Supers and probably why there is no Ti versions at launch (although lots of people thought the 3090 naming wasn't going to be a thing too). It seems that the main reason for the low VRAM compared to the previous generation at launch is the GDDR6X isn't in enough production yet. Several articles are saying as soon as the end of next month, which seems too soon to me also.

 

But I can see where they are coming from, as long as the demand for the standard versions isn't so great they can't get enough GDDR6X to make those versions or decide they have enough of a good thing going with the demand.

 

Sent from my LM-G850 using Tapatalk

[Gunnars Driver]

3090 on order, since 15:00:30 today.

[Waxer]

That is great. Congrats. Glad the actual humans got in on this one.

[imacken]

Nor for me they didn't. I had 6 sites open, and all went to pre-order instantly, and that was with refreshing every few seconds from 2pm on the dot!

[Waxer]

Ahh, sucks. OP is in Sweden, not UK. I guess we are a bigger market for the bots to target. Never mind... you get to wait until all the information gets in to make your choice. I remember you saying that you were more focused on ultimate performance than on price, but even so you will be armed with the data and hopefully have DCS benchmark data to make you decision. Even if you still go for the BFGPU in the end! Besides the 2080 Ti still no slouch.

[imacken]

I've got a pre order in with Scan for expected delivery 5 October though. Not too long to wait if that is correct, of course!

[Waxer]

There you go: enjoy! And remember your mates on the forum... let us know how much of an improvement from the 2080 Ti in DCS.