Thank you for the link.
I have not read it yet, I will, but let me guess.
Not many games use RTX.
It is still early days.
It's just another PhysX.
The cards are way too expensive.
You should not pre-order until you see benchmarks.
These cards are probably a placeholder until Nvidia can move to 7nm.
Nvidia are using the RTX and AI cores as an excuse to add value for gamers that isn't there.
Claims that Nvidia are hiding the fact that these new 2080Ti cards do not give much of a performance uplift (15% / Maybe 10% in games) versus GTX1080Ti.
Yet the fact that AI and RTX has been added to Consumer GPU seems very important to me indeed.
Not just for gaming but also for other compute / applications such as Blender Rendering, and AI for pattern recognition etc.
People should very definitely wait for benchmarks.
Probably better for most to stay with GTX1080 or 1080Ti as prices continue to drop rather than pre-order these 2080Ti's yet.
I have seen articles claiming that Nvidia got hit hard by Cryptocurrency crash, and have lots of GPU's returned to inventory.
This 2080Ti etc 'launch' - actually a pre-order, until 20 September, at these high prices - could be a way to get many people to throw in the towel and buy the older Nvidia cards, at lower prices, therefore reduce their inventory.
I do not like to see the current situation. I own many AMD cards, far more than the number of Nvidia GPU's I own.
I bought into the RX Vega message, and it overpromised and underdelivered, the cards were late, and I still stayed with AMD cards.
I have an interest in AMD cards being a success. But last time I looked they dropped as low as 8% of Steam Users just after Christmas in January 2018.
I want to see AMD do better than this. Look at what they did with Ryzen and Threadripper.
OK, thank you, very interesting article.
Based on what they seem to say, my estimate of performance improvement does not seem unreasonable then, and that would make these GTX2080Ti cards over 50% faster than an RX Vega 64 Liquid card. Vega 64 on 7nm is reportedly 1.35x (35%) faster than Vega 64 today. That means AMD would still be behind if they released Vega 64 on 7nm for gamers today.
I do disagree with some things said in the Article. This in particular.
"True, DX12 has been kinder to Team Red than Team Green, but if you bought a 2013 Radeon thinking Mantle was going to take over the gaming industry, you didn’t get a lot of shipping titles before it was retired in favor of other APIs. If you bought a Radeon in 2013 thinking you were getting in on the dawn of a new age of gaming, well, you were wrong."
Regarding the DX12 versus DX11 story. Why would Nvidia want to move their cards and games from DX11 to DX12 when their cards run so much better on DX11 than AMD cards do? Isn't it clear that Nvidia would do everything possible to prevent encouraging a move to DX12? Is there no way AMD can go back and do something to improve their newer card DX11 performance?
I still seem to be able to run Battlefield 4 in Mantle. I was running it just the other day. Battlefield Hardline does crash - I recently posted about that.
The author seems to forget that AMD Mantle was a starting point for Vulkan. My radeon card I bought in 2013 runs Doom really well on Vulkan.
Another point they don't seem to make, announced just recently is the new Steam Play Beta for Linux , which would be nowhere without Vulkan & AMD.
Here is the information: Steam for Linux :: Introducing a new version of Steam Play
To cut a long story short, you can now run Doom 2016 on Linux and other Windows based games will be easily ported to Steam on Linux because of Vulkan.
This is a major change for Linux Users. This will also be a major change for people who have had enough of Windows.
I just tested it today. Steam Plays Beta for Linux today.
I will add another post about it next, with links to videos showing the performance.
FYI, Doom at least, worked first time and it runs really well.
"Regarding the DX12 versus DX11 story. Why would Nvidia want to move their cards and games from DX11 to DX12 when their cards run so much better on DX11 than AMD cards do? Isn't it clear that Nvidia would do everything possible to prevent encouraging a move to DX12?"
Exactly. DX12, being a low level API has the ability to run far more efficiently than DX11 ever could. But due to NVidia low levels of compute hardware on gaming GPUs, they instead build Gameworks solutions to try and achieve those effects. You can see my lengthy blurb about it above. The crux of the issue is that NVidia would rather maintain the differentiation of their product stack than really push gaming forward. That is the primary reason I don't support them.
"The author seems to forget that AMD Mantle was a starting point for Vulkan. My radeon card I bought in 2013 runs Doom really well on Vulkan.
Another point they don't seem to make, announced just recently is the new Steam Play Beta for Linux , which would be nowhere without Vulkan & AMD."
Uh-huh. How many shipping titles are there with Vulkan support? The author is making the point that when you buy a graphics card with a new feature Mantle, Vulkan, Ray tracing, whatever it is, by the time that feature is widely implemented the original graphics generation that supported it is obsolete. I can't really argue that fact.
"Based on what they seem to say, my estimate of performance improvement does not seem unreasonable then, and that would make these GTX2080Ti cards over 50% faster than an RX Vega 64 Liquid card"
Tempted to put a check minus in the reading for comprehension section, but no! I will distill the essence of the extremetech article here.
The author indicates that NVidia cards gains in previous generations were tied very closely to the fill rate gains made. They even supply this handy dandy chart.
As you can see, the author points out that despite other gains the performance in existing games seems to be tied very closely to the fill rate. The GTX 980 gained 1.50X compared to the the GTX 780 in fill rate and saw an improvement of 46% across 14 titles despite actually losing memory bandwidth. The GTX 1080 then, saw a 1.60X fill rate increase over the GTX 980 and saw a 65% increase across the same 14 titles. The GTX 2080 actually has a slightly slower fill rate than the GTX 1080, which means the gains will likely be limited.
Similarly, the GTX 1080 Ti has a fill rate of 130.2 (Gpix/s), while the the GTX 2080 Ti has a fill rate of 135.96 (Gpix/s), an increase of 1.044X. So if the trend holds, it would be 5% faster in those same 14 games.
The author does qualify that with the following.
"And those facts alone suggest that unless Nvidia managed to deliver the mother of all IPC improvements via rearchitecting its GPU core, the RTX 2080 family is unlikely to deliver a huge improvement in current games."
So you may see some additional gains, but I think even expecting a 20% increase is generous.
RE: Tempted to put a check minus in the reading for comprehension section, but no!
This forum is a laugh, it really is.
I make some comments about the RTX2080Ti series, based on some GTX1080Ti data I have and a few simple assumptions, pretty much that Nvidia know what they are doing and don't mess up the GPU core.
I get trolled and report it.
The thread disappears.
The guy in 11th place on the forum (I have no problem with him at all, he is a really nice person) gets 1000 point increase in one day from the person I reported so I vanish from your Drivers page.
Keep sticking your heads in the sand.
Herd mentality in action.
A herd of ostriches
If your talking about kingfish giving me a badge. That was his way of rewarding a User for helping out another User solve their problem. He has done that with several other Users in the past. I really don't expect anything from this Forum. I am not in competition with anyone. I just enjoy troubleshooting.
I am sure he didn't do it so just to get back at you. He is more professional then that.
EDIT: By the way, Any User can give a Badge to another User from what I gathered. Just need to click on the User's Profile and click "Give a Badge". If you want I'll give you a Badge if that well make you feel better. I have no idea how many points it will be worth.
If my comment offend you, then I do apologize. I was merely pointing out that author indicated that the 2080 Ti only has a modest (5%) fill rate increase over the previous model (1080 Ti) and historically speaking the fill rate seems to correlate almost directly with the gains seen in games.
You had said "Based on what they seem to say, my estimate of performance improvement does not seem unreasonable then, and that would make these GTX2080Ti cards over 50% faster than an RX Vega 64 Liquid card."
That really doesn't seem to be the case, the author is indicating based on fill rates (which historically have been a good predictor of gaming performance) we really shouldn't expect much in the way of gains in current games. So I'm not exactly sure how you read the author's data as supporting your conclusion.
I'm not sure what other situation you are referring to, so I will leave that alone. And the truth is we won't know how fast the GTX 2000 series is until benchmarks launch in September. But it may be a worse case scenario in which the cards are only 10-15% fast in existing titles but are significantly more expensive.
Been busy testing Steam Play Beta.
1. I do not want to fight with anyone. I am interested in what AMD response will be to these Nvidia cards.
2. Can you please tell me, how did the author calculate that fill rate of "109?" for the RTX2080? What assumptions did he make?
3. Here is some idea of how many Game titles now run using Vulkan, on Linux admittedly:
Steam for Linux :: Introducing a new version of Steam Play. If you run Linux, can you please test ...
Like I said earlier, we will see what the benchmark numbers come out like.
I think the pixel fill rates are reported by the card manufacturers in there own specification sheets, I don't think they are being calculated.
The stream Linux stuff is pretty interesting. Nice to see the gaming community outside of Windows continues to grow.