• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

NVIDIA GeForce RTX 4090 Founders Edition

Joined
Dec 28, 2012
Messages
4,148 (0.94/day)
System Name Skunkworks 3.0
Processor 5800x3d
Motherboard x570 unify
Cooling Noctua NH-U12A
Memory 32GB 3600 mhz
Video Card(s) asrock 6800xt challenger D
Storage Sabarent rocket 4.0 2TB, MX 500 2TB
Display(s) Asus 1440p144 27"
Case Old arse cooler master 932
Power Supply Corsair 1200w platinum
Mouse *squeak*
Keyboard Some old office thing
Software Manjaro
When you think a PS5 is 500$ and can play ANY game, you kind of start to laugh of those ridiculous prices from nGreedia. I mean, really. In the end is just a game you're playing.
Any price above 800$ for a GPU is not worthing at all.
$500, unless you want to play games off of disk, then its $600, for a console that will likely play many of its games at 30 FPS, with dithered resolution, and a wonderful closed garden environment where you have to subscribe to play online, cant mod, and get to pay full price for years old games, then yeah its great.
 
Joined
May 2, 2017
Messages
7,762 (2.74/day)
Location
Back in Norway
System Name Hotbox
Processor AMD Ryzen 7 5800X, 110/95/110, PBO +150Mhz, CO -7,-7,-20(x6),
Motherboard ASRock Phantom Gaming B550 ITX/ax
Cooling LOBO + Laing DDC 1T Plus PWM + Corsair XR5 280mm + 2x Arctic P14
Memory 32GB G.Skill FlareX 3200c14 @3800c15
Video Card(s) PowerColor Radeon 6900XT Liquid Devil Ultimate, UC@2250MHz max @~200W
Storage 2TB Adata SX8200 Pro
Display(s) Dell U2711 main, AOC 24P2C secondary
Case SSUPD Meshlicious
Audio Device(s) Optoma Nuforce μDAC 3
Power Supply Corsair SF750 Platinum
Mouse Logitech G603
Keyboard Keychron K3/Cooler Master MasterKeys Pro M w/DSA profile caps
Software Windows 10 Pro
No, I disagree. The poorer EU countries have much lower purchase power, so these products are sitting on the shelves collecting dust. Because literally no one buys them.
... exactly. Which drives up prices, because actually running a business selling these things becomes more expensive due to this. You have more money bound up in products that take time to sell, meaning you have less cash flow and more trouble paying your bills, meaning you need more cash reserves and higher margins to cover operating costs relative to the time between you buy and sell the product. What you are describing is literally why this is the way it is. And at the level of distribution, there are less products flowing through the distribution system, forcing distributors to charge more to cover their (otherwise relatively static) business costs.
 
Joined
Apr 13, 2022
Messages
1,268 (1.23/day)
It's very little due to "greedy retailers", and more due to the realities of operating in smaller, less unified markets. The US has massive economics of scale due to being a 300m people, wealthy, single-language market, meaning tons of sales of high margin products with relatively low staff needs and increased opportunity for centralization of various branches of the retail value chain. In contrast, while the EU is nominally a single market, it is still fragmented through language, distribution chains, localized economic factors, and more. This drives up prices simply because there are more people involved, more business costs that need covering in order to deliver the products onto retail shelves.

Germany being the cheapest in Europe just confirms this: it's the largest single market, and has a relatively high level of income, meaning it gets a lot of the same advantages as the US, just to a lesser degree due to less scale. The only major difference would be that many brands have direct distribution in the US (meaning AMD, Nvidia, Asus, and others sell directly to retailers rather than to distributors), which is again a limitation of market size - running your own distribution business is expensive and difficult.

After nearly a decade in electronics retail, even if that's a while back now, I can verify with 100% security that for most retailers, margins (especially on expensive products like GPUs and CPUs) are razor-thin, often below what is actually needed for the company to break even, let alone make any net profit.

Yep this. Way before the founders program Nvidia had direct distribution in the US to places like Best Buy as Nvidia branded cards. The US has a relative few massive retailers here and companies can directly get their products there at a reasonable cost. There are really only a few shipping channels as well that often own their own aiports or own/lease dedicated hubs at airports.

Even among "distributors" that operate on a global scale like say Amazon the scale of their facilities in the US compared to overseas is mind boggling. If you haven't seen a US Amazon facility or say UPS hub you have no idea how stupidly insane things are here. These things have buildings that take up acres, run 24/7 and at times their own air strips. It's nuts.
$500, unless you want to play games off of disk, then its $600, for a console that will likely play many of its games at 30 FPS, with dithered resolution, and a wonderful closed garden environment where you have to subscribe to play online, cant mod, and get to pay full price for years old games, then yeah its great.
This is silly. Diskless is 400 with BR disk is 500. Not all the games are 30 fps and most peoples PCs simply will not run 4k at even 30fps. 4k on PS5 is a reality with performance from 30-60fps which is something that the vast majority of PC gamers simply cannot do. As for subscriptions yeah sure. But I know lots of PC gamers with multiple subscriptions to multiple games or services so really that's not a good argument. You also do not pay full price for much older games that's only the case in stuff like Nintendo's own IPs for the most part they drop fast. You are also generally not playing the games off disk. You're installing part of it to the SSD, and you can add another SSD.

I have a PC, a Switch, and a PS5 (along with my older stuff) and the PS5 is really better at 4k in general than the PC for most people. I live in a very afluent area and the general trend among the people who were PC gamers has been moving to a macbook pro and getting a console for a while now. Simply because even though these tesla/bmw owning assholes could afford a 4090 but it's sort of ludicrus when a PS5 does what it should, is less hassle, and really just works. The PC gaming side of things that is still around is stupidly highrefresh 1080p monitors for stuff like FPS.

All these things have their own place and purpose and I enjoy all of them. Really I enjoy the PC for the keyboard and mouse on FPS and I love me some strategy games. But now that I have a 120hz OLED in living room and consoles are getting there for stuff like Dark Souls, Elden Ring, and other stuff it's sort of a wash.
 
Joined
May 2, 2017
Messages
7,762 (2.74/day)
Location
Back in Norway
System Name Hotbox
Processor AMD Ryzen 7 5800X, 110/95/110, PBO +150Mhz, CO -7,-7,-20(x6),
Motherboard ASRock Phantom Gaming B550 ITX/ax
Cooling LOBO + Laing DDC 1T Plus PWM + Corsair XR5 280mm + 2x Arctic P14
Memory 32GB G.Skill FlareX 3200c14 @3800c15
Video Card(s) PowerColor Radeon 6900XT Liquid Devil Ultimate, UC@2250MHz max @~200W
Storage 2TB Adata SX8200 Pro
Display(s) Dell U2711 main, AOC 24P2C secondary
Case SSUPD Meshlicious
Audio Device(s) Optoma Nuforce μDAC 3
Power Supply Corsair SF750 Platinum
Mouse Logitech G603
Keyboard Keychron K3/Cooler Master MasterKeys Pro M w/DSA profile caps
Software Windows 10 Pro
Lol, the "consoles need subscriptions" argument is ... well, growing thinner by the day. How many PC gamers have Game Pass PC or Ultimate? I sure do, and I love it. Xbox All Access is possibly the best gaming deal available, with a zero-interest downpayment plan for a great console including Game Pass Ultimate? Awesome deal. Playstation doesn't have the same value proposition, but they still give out a ton of good free games through PS+. PCs have the advantage of backwards compatibility with a huge library of free games, but if you want to play new games, it takes a long time for the ~$10 advantage of PC game pricing to surpass the cost difference vs. a console at similar graphics/performance levels.
 
Joined
Apr 13, 2022
Messages
1,268 (1.23/day)
... exactly. Which drives up prices, because actually running a business selling these things becomes more expensive due to this. You have more money bound up in products that take time to sell, meaning you have less cash flow and more trouble paying your bills, meaning you need more cash reserves and higher margins to cover operating costs relative to the time between you buy and sell the product. What you are describing is literally why this is the way it is. And at the level of distribution, there are less products flowing through the distribution system, forcing distributors to charge more to cover their (otherwise relatively static) business costs.

To get back to this as the US is massive stuff is routed quickly to the places that have the income. Case in point we have COSTCO which is a US thing. This is a warehouse full of stuff and it's a national chain. It's a "discount" store in the sense that it's cheaper but they stell stuff in massive amounts ie by the crate, or the entire rib rack, or multiple whole chickens. I've included some pictures but in the US these things aren't rare. There are dozens in areas. To call it a store is a bit silly. It requires fork lifts to get you some things. And we have these in the middle of downtown cities! They also have pharmacies, food courts, and auto centers.

The funny thing though is due to the stupidly insane ability of the US to move stuff about they don't carry the same things. The COSTCO where I live sells diamond jewlrey in the 10s of thousands range, Rolex watches, LG OLEDs, 1600 dollar giant slabs of wagyu beef and other stuff. Drive 30 mins away and they don't have any of that. It has jewlrey, watches, and TVs, but it's not the same stuff.

So in the US the higher end GPUs show in the market areas that can afford them, the other areas get lower end ones. All this stuff is done by AI systems and it runs itself. And if you want to buy a gift for a relative in another state where their COSTCO doesn't have it you buy it at yours and they route the shipping from their distrocenter to the other store and they come and pick it up.

You can't get the scale of the US if you haven't seen how bonkers it gets.
 

Attachments

  • costco.JPG
    costco.JPG
    129.5 KB · Views: 97
  • costco2.JPG
    costco2.JPG
    150.2 KB · Views: 96
Joined
May 2, 2017
Messages
7,762 (2.74/day)
Location
Back in Norway
System Name Hotbox
Processor AMD Ryzen 7 5800X, 110/95/110, PBO +150Mhz, CO -7,-7,-20(x6),
Motherboard ASRock Phantom Gaming B550 ITX/ax
Cooling LOBO + Laing DDC 1T Plus PWM + Corsair XR5 280mm + 2x Arctic P14
Memory 32GB G.Skill FlareX 3200c14 @3800c15
Video Card(s) PowerColor Radeon 6900XT Liquid Devil Ultimate, UC@2250MHz max @~200W
Storage 2TB Adata SX8200 Pro
Display(s) Dell U2711 main, AOC 24P2C secondary
Case SSUPD Meshlicious
Audio Device(s) Optoma Nuforce μDAC 3
Power Supply Corsair SF750 Platinum
Mouse Logitech G603
Keyboard Keychron K3/Cooler Master MasterKeys Pro M w/DSA profile caps
Software Windows 10 Pro
@W1zzard Looking more closely at the efficiency measurements and CP2077 benchmarks here, it seems to me that you've run into a problem: CP2077 @1440p is quite clearly CPU limited on this GPU. No, it isn't quite at the average FPS of 1080p, but the difference is negligible, meaning that for a significant proportion of the test sequence, the game is likely CPU bound rather than GPU bound, with some sections being heavier on the GPU and thus drawing the average FPS down by a couple of percentage points. To me this would seem to align with other media outlets reporting power draws in the ~450W range for gaming loads. IMO, this makes your conclusions on efficiency essentially invalid - it seems impossible that the GPU is actually being loaded 100% during this workload, making it rather unrepresentative outside of that specific game and system configuration.

Of course that also makes the performance increase in CP2077 all the more impressive, but as Der8auer showed in a recent video, the 4090 seems to lose only ~5% performance with a power limit reduction down to 60%!
 

ARF

Joined
Jan 28, 2020
Messages
4,670 (2.55/day)
Location
Ex-usa | slava the trolls
... exactly. Which drives up prices, because actually running a business selling these things becomes more expensive due to this. You have more money bound up in products that take time to sell, meaning you have less cash flow and more trouble paying your bills, meaning you need more cash reserves and higher margins to cover operating costs relative to the time between you buy and sell the product. What you are describing is literally why this is the way it is. And at the level of distribution, there are less products flowing through the distribution system, forcing distributors to charge more to cover their (otherwise relatively static) business costs.

I want to ask you the following - when do the prices decrease?
1. when the demand is low, supply is fine, and no one buys?
or
2. when the demand is high, supply is plenty, and everyone buys?
 
Joined
Aug 28, 2019
Messages
190 (0.10/day)
System Name 4K 144hz gaming PC :)
Processor Ryzen 9 7950X
Motherboard ASrock X670E Steel Legend
Cooling NZXT Kraken Z73
Memory 2x16GB Corsair Dominator Platinum RGB DDR5 6400 CL38 @ 6200 CL32
Video Card(s) Gainward RTX 4090 Phantom @ GS
Storage WB Black 1TB Pcie4
Display(s) iiyama G-MASTER G4380UHSU-B1
Case Lian Li O11D Evo
Audio Device(s) Realtek ALC1220
Power Supply EVGA Supernova 1300 G2
Mouse Razer Deathadder Essential
Keyboard Razer Ornata Chroma
Software Win 11 PRO
Benchmark Scores Time Spy: 34857 Time Spy Extreme: 19629 Port Royal: 28542 Speed Way: 11257 (Hall of Fame)
Soo, i just went home at 15:00 and didn't check NBB and stuff. Did the FE drop and how much did it cost in EU? Did 1599$ turn into 1949€ this time, although the 3080 FE was 699$ or 710€ at launch in 2020.
 
Joined
Aug 21, 2015
Messages
1,788 (0.52/day)
Location
North Dakota
System Name Office
Processor Ryzen 5600G
Motherboard ASUS B450M-A II
Cooling be quiet! Shadow Rock LP
Memory 16GB Patriot Viper Steel DDR4-3200
Video Card(s) Gigabyte RX 5600 XT
Storage PNY CS1030 250GB, Crucial MX500 2TB
Display(s) Dell S2719DGF
Case Fractal Define 7 Compact
Power Supply EVGA 550 G3
Mouse Logitech M705 Marthon
Keyboard Logitech G410
Software Windows 10 Pro 22H2
Hard to say; depends on what else and how much of that else you have in your system.

I would recommend an 850W Gold / Plat minimum PSU to be safe. I think if you are going to slot in an RTX4090 on a 750W PSU, your system probably won't post or if it does, you will have to power limit the card in MSI afterburner before use, and even if you power limit the card, what else might happen is, during load, your machine will "dip" out and reboot.

Er, why wouldn't it POST? I'm not going to claim that 750W isn't riding kinda close to the edge, but I can't think of a reason it wouldn't even run.
 

chowow

New Member
Joined
Oct 11, 2022
Messages
22 (0.03/day)
You'll get your 13900K with RTX 4090 results soon enough, just be a bit more patient
THANK FOR PUTTING IN THAT 5800X now we'll see if there's any difference AT 4K MAYBE if there's not much different like 5% I GOING TO GET 4090 WITH MY 5600 don't listen to these goofballs again thanks for all your hard work.
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
28,149 (3.72/day)
Processor Ryzen 7 5700X
Memory 48 GB
Video Card(s) RTX 4080
Storage 2x HDD RAID 1, 3x M.2 NVMe
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
THANK FOR PUTTING IN THAT 5800X now we'll see if there's any difference AT 4K MAYBE if there's not much different like 5% I GOING TO GET 4090 WITH MY 5600 don't listen to these goofballs again thanks for all your hard work.
+1 for 5600X + 4090, it's probably not even 5%, but we'll know soon enough, very interesting question
 

HTC

Joined
Apr 1, 2008
Messages
4,668 (0.76/day)
Location
Portugal
System Name HTC's System
Processor Ryzen 5 5800X3D
Motherboard Asrock Taichi X370
Cooling NH-C14, with the AM4 mounting kit
Memory G.Skill Kit 16GB DDR4 F4 - 3200 C16D - 16 GTZB
Video Card(s) Sapphire Pulse 6600 8 GB
Storage 1 Samsung NVMe 960 EVO 250 GB + 1 3.5" Seagate IronWolf Pro 6TB 7200RPM 256MB SATA III
Display(s) LG 27UD58
Case Fractal Design Define R6 USB-C
Audio Device(s) Onboard
Power Supply Corsair TX 850M 80+ Gold
Mouse Razer Deathadder Elite
Software Ubuntu 20.04.6 LTS
After checking the reviews of the non reference cards, i see why EVGA opted out: 1 to 3% higher performance only @ 4K, for $100 to $400 more ...
 
Joined
Nov 26, 2021
Messages
1,804 (1.55/day)
Location
Mississauga, Canada
Processor Ryzen 7 5700X
Motherboard ASUS TUF Gaming X570-PRO (WiFi 6)
Cooling Noctua NH-C14S (two fans)
Memory 2x16GB DDR4 3200
Video Card(s) Reference Vega 64
Storage Intel 665p 1TB, WD Black SN850X 2TB, Crucial MX300 1TB SATA, Samsung 830 256 GB SATA
Display(s) Nixeus NX-EDG27, and Samsung S23A700
Case Fractal Design R5
Power Supply Seasonic PRIME TITANIUM 850W
Mouse Logitech
VR HMD Oculus Rift
Software Windows 11 Pro, and Ubuntu 20.04
I don't see how it is possible to generate an intermediary frame between two frames using temporal (time-based) data and not require that the second frame be rendered before the interpose. Frame1 - Interframe - Frame2, in order to create interframe you must have the already rendered Frame1 and Frame2, and by the time you push Frame1 to the screen it is behind by a frame, giving the 1/3 extra latency. Is this incorrect?
Latency varies; In Spider-Man, it is similar to native rendering without Reflex. In Cyberpunk, it is slightly less than native with Reflex. In both of these, it is worse than DLSS 2 with Reflex.

Marvel's Spider-Man Feast HQPerf DifferentialReflex OffReflex On
Native 4K100%39ms36ms
DLSS 2 Performance136%24ms23ms
DLSS 3 Frame Generation219%-38ms


Cyberpunk 2077 MarketPerf DifferentialReflex OffReflex On
Native 4K100%108ms62ms
DLSS 2 Performance258%42ms31ms
DLSS 3 Frame Generation399%-54ms
 
Joined
May 2, 2017
Messages
7,762 (2.74/day)
Location
Back in Norway
System Name Hotbox
Processor AMD Ryzen 7 5800X, 110/95/110, PBO +150Mhz, CO -7,-7,-20(x6),
Motherboard ASRock Phantom Gaming B550 ITX/ax
Cooling LOBO + Laing DDC 1T Plus PWM + Corsair XR5 280mm + 2x Arctic P14
Memory 32GB G.Skill FlareX 3200c14 @3800c15
Video Card(s) PowerColor Radeon 6900XT Liquid Devil Ultimate, UC@2250MHz max @~200W
Storage 2TB Adata SX8200 Pro
Display(s) Dell U2711 main, AOC 24P2C secondary
Case SSUPD Meshlicious
Audio Device(s) Optoma Nuforce μDAC 3
Power Supply Corsair SF750 Platinum
Mouse Logitech G603
Keyboard Keychron K3/Cooler Master MasterKeys Pro M w/DSA profile caps
Software Windows 10 Pro
I want to ask you the following - when do the prices decrease?
1. when the demand is low, supply is fine, and no one buys?
or
2. when the demand is high, supply is plenty, and everyone buys?
That simplistic reasoning can only be used for comparing absolute pricing across different markets if all other factors are equal - otherwise you're just ignoring a heap of variables in how differently businesses operate in different regions. A distributor covering a market of 5 million people like Norway will have much higher costs per product sold than one covering 80 million people like in Germany, even if the latter necessarily has a lot more employees and other costs. Why? Because the market is 15 times larger, but they won't need 15x the employees, warehouse space, or other operating costs. A lot of those costs (assuming similar price levels in each location) will be very, very similar, despite the latter potentially selling 15x more products. They will also have a much more substantial cash flow, making it easier for them to pay creditors and thus having less credit interest to pay at any given time. The larger market also allows for thinner stock margins as you can make better predictive models for a larger population, making it possible to have less stock sitting around unsold for significant periods of time, further improving cash flow. All this means that they need much lower margins in order to break even or make any given profit, which again drives down prices compared to smaller markets. The lack of a market basis for any real competition in smaller markets also severely affects pricing.

I don't see how it is possible to generate an intermediary frame between two frames using temporal (time-based) data and not require that the second frame be rendered before the interpose. Frame1 - Interframe - Frame2, in order to create interframe you must have the already rendered Frame1 and Frame2, and by the time you push Frame1 to the screen it is behind by a frame, giving the 1/3 extra latency. Is this incorrect?
That's not what they're doing. Rather, let's say we have three frames, 0, 1 and 2. They use the data from frames 0 and 1 to make a guess at what will happen between frame 1 and the yet-to-be-rendered frame 2. Interframe 1, between frames 1 and 2, has no actual relation to frame 2 beyond coming before it in the order of frames displayed.
 
Joined
Oct 5, 2022
Messages
12 (0.01/day)
That simplistic reasoning can only be used for comparing absolute pricing across different markets if all other factors are equal - otherwise you're just ignoring a heap of variables in how differently businesses operate in different regions. A distributor covering a market of 5 million people like Norway will have much higher costs per product sold than one covering 80 million people like in Germany, even if the latter necessarily has a lot more employees and other costs. Why? Because the market is 15 times larger, but they won't need 15x the employees, warehouse space, or other operating costs. A lot of those costs (assuming similar price levels in each location) will be very, very similar, despite the latter potentially selling 15x more products. They will also have a much more substantial cash flow, making it easier for them to pay creditors and thus having less credit interest to pay at any given time. The larger market also allows for thinner stock margins as you can make better predictive models for a larger population, making it possible to have less stock sitting around unsold for significant periods of time, further improving cash flow. All this means that they need much lower margins in order to break even or make any given profit, which again drives down prices compared to smaller markets. The lack of a market basis for any real competition in smaller markets also severely affects pricing.


That's not what they're doing. Rather, let's say we have three frames, 0, 1 and 2. They use the data from frames 0 and 1 to make a guess at what will happen between frame 1 and the yet-to-be-rendered frame 2. Interframe 1, between frames 1 and 2, has no actual relation to frame 2 beyond coming before it in the order of frames displayed.
So:

Frame 0 (player frame) - render, queue
Frame 1 (ai frame), render, queue
Frame 0 push
Frame 2 (player frame), render, queue
Frame 1 push
Frame 2 push
repeat

Is this correct? If so, how does Frame 1 get generated temporally if it does not have data from Frame 2? It would be like seeking in an mpeg video only having access to already played video frames, it wouldn't work? What am I missing?
 

3x0

Joined
Oct 6, 2022
Messages
974 (1.14/day)
Processor AMD Ryzen 7 5800X3D
Motherboard MSI MPG B550I Gaming Edge Wi-Fi ITX
Cooling Scythe Fuma 2 rev. B Noctua NF-A12x25 Edition
Memory 2x16GiB G.Skill TridentZ DDR4 3200Mb/s CL14 F4-3200C14D-32GTZKW
Video Card(s) PowerColor Radeon RX7800 XT Hellhound 16GiB
Storage Western Digital Black SN850 WDS100T1X0E-00AFY0 1TiB, Western Digital Blue 3D WDS200T2B0A 2TiB
Display(s) Dell G2724D 27" IPS 1440P 165Hz, ASUS VG259QM 25” IPS 1080P 240Hz
Case Cooler Master NR200P ITX
Audio Device(s) Altec Lansing 220, HyperX Cloud II
Power Supply Corsair SF750 Platinum 750W SFX
Mouse Lamzu Atlantis Mini Wireless
Keyboard HyperX Alloy Origins Aqua
It's like this:
Frame 1, Frame 2 get rendered but not pushed yet, Frame 1.5 AI gets generated based on 1 and 2, and then pushed out Frame 1, AI 1.5, 2.
 

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
28,149 (3.72/day)
Processor Ryzen 7 5700X
Memory 48 GB
Video Card(s) RTX 4080
Storage 2x HDD RAID 1, 3x M.2 NVMe
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
It's like this:
Frame 1, Frame 2 get rendered but not pushed yet, Frame 1.5 AI gets generated based on 1 and 2, and then pushed out Frame 1, AI 1.5, 2.
This is what NVIDIA told me is happening.

To expand slightly, you have a render queue in which rendered frames are saved before they go out to screen, also called "frames render ahead" or similar. Typically the queue is 3 frames

DLSS 3 (just like Reflex) removes that queue, buffers the frames internally and sends them out at the right time, with generated frames in-between.

If this conflicts with other sources let me know so I can double-check with them
 
Joined
Oct 5, 2022
Messages
12 (0.01/day)
It's like this:
Frame 1, Frame 2 get rendered but not pushed yet, Frame 1.5 AI gets generated based on 1 and 2, and then pushed out Frame 1, AI 1.5, 2.
That's what I originally wrote though...

I don't see how it is possible to generate an intermediary frame between two frames using temporal (time-based) data and not require that the second frame be rendered before the interpose. Frame1 - Interframe - Frame2, in order to create interframe you must have the already rendered Frame1 and Frame2, and by the time you push Frame1 to the screen it is behind by a frame, giving the 1/3 extra latency. Is this incorrect?
But then it was said that I was wrong...
That's not what they're doing. Rather, let's say we have three frames, 0, 1 and 2. They use the data from frames 0 and 1 to make a guess at what will happen between frame 1 and the yet-to-be-rendered frame 2. Interframe 1, between frames 1 and 2, has no actual relation to frame 2 beyond coming before it in the order of frames displayed.

This is what NVIDIA told me is happening.

To expand slightly, you have a render queue in which rendered frames are saved before they go out to screen, also called "frames render ahead" or similar. Typically the queue is 3 frames

DLSS 3 (just like Reflex) removes that queue, buffers the frames internally and sends them out at the right time, with generated frames in-between.

If this conflicts with other sources let me know so I can double-check with them
What does 'removes that queue and buffers the frames internally' mean? Does it mean that it gets rid of the queue and then frames would be real time 'out of order', so it is pushing frames when ever they are ready, and using a buffer somewhere to hold the frames and generate the AI frame and push them in whatever order is most performant?
 
  • Like
Reactions: 3x0
Joined
Dec 28, 2012
Messages
4,148 (0.94/day)
System Name Skunkworks 3.0
Processor 5800x3d
Motherboard x570 unify
Cooling Noctua NH-U12A
Memory 32GB 3600 mhz
Video Card(s) asrock 6800xt challenger D
Storage Sabarent rocket 4.0 2TB, MX 500 2TB
Display(s) Asus 1440p144 27"
Case Old arse cooler master 932
Power Supply Corsair 1200w platinum
Mouse *squeak*
Keyboard Some old office thing
Software Manjaro
This is silly. Diskless is 400 with BR disk is 500.
If you can find them. Wal mart, for instance, is selling the diskless for 658 and the BR for 739.
Not all the games are 30 fps and most peoples PCs simply will not run 4k at even 30fps
4k on PS5 is a reality with performance from 30-60fps which is something that the vast majority of PC gamers simply cannot do.
4k on ps5 is dithered/checkerboarded form lower resolutions. You can easily do the same with most gaming PCs. You can also, you know, choose what settings are frame rate you would like on PC, instead of being stuck on 4k30 of a console.
As for subscriptions yeah sure. But I know lots of PC gamers with multiple subscriptions to multiple games or services so really that's not a good argument.
Whataboutism at its finest. Come back with an actual argument.
You also do not pay full price for much older games that's only the case in stuff like Nintendo's own IPs for the most part they drop fast.
You clearly do not pay attention tot he price of console VS PC games if you think the console prices are anywhere close.
You are also generally not playing the games off disk. You're installing part of it to the SSD, and you can add another SSD.
This has absolutely nothing to do with the argument, which is that consoles are a walled garden and not the same experience as a (relatively) open and customizable PC experience.
I have a PC, a Switch, and a PS5 (along with my older stuff) and the PS5 is really better at 4k in general than the PC for most people. I live in a very afluent area and the general trend among the people who were PC gamers has been moving to a macbook pro and getting a console for a while now. Simply because even though these tesla/bmw owning assholes could afford a 4090 but it's sort of ludicrus when a PS5 does what it should, is less hassle, and really just works. The PC gaming side of things that is still around is stupidly highrefresh 1080p monitors for stuff like FPS.

All these things have their own place and purpose and I enjoy all of them. Really I enjoy the PC for the keyboard and mouse on FPS and I love me some strategy games. But now that I have a 120hz OLED in living room and consoles are getting there for stuff like Dark Souls, Elden Ring, and other stuff it's sort of a wash.
Literally this entire section is annecdotal gibberish.
 
Joined
Dec 25, 2020
Messages
7,458 (4.97/day)
Location
São Paulo, Brazil
System Name "Icy Resurrection"
Processor 13th Gen Intel Core i9-13900KS
Motherboard ASUS ROG Maximus Z790 Apex Encore
Cooling Noctua NH-D15S upgraded with 2x NF-F12 iPPC-3000 fans and Honeywell PTM7950 TIM
Memory 32 GB G.SKILL Trident Z5 RGB F5-6800J3445G16GX2-TZ5RK @ 7600 MT/s 36-44-44-52-96 1.4V
Video Card(s) NVIDIA RTX A2000
Storage 500 GB WD Black SN750 SE NVMe SSD + 4 TB WD Red Plus WD40EFPX HDD
Display(s) 55-inch LG G3 OLED
Case Pichau Mancer CV500 White Edition
Power Supply EVGA 1300 G2 1.3kW 80+ Gold
Mouse Microsoft Classic IntelliMouse (2017)
Keyboard IBM Model M type 1391405
Software Windows 10 Pro 22H2
Benchmark Scores I pulled a Qiqi~
does RTX 4000 series have LHR or similar workload limiters?

I doubt it. The 3090 series didn't either, besides, mining is dead.
 
Joined
Aug 28, 2019
Messages
190 (0.10/day)
System Name 4K 144hz gaming PC :)
Processor Ryzen 9 7950X
Motherboard ASrock X670E Steel Legend
Cooling NZXT Kraken Z73
Memory 2x16GB Corsair Dominator Platinum RGB DDR5 6400 CL38 @ 6200 CL32
Video Card(s) Gainward RTX 4090 Phantom @ GS
Storage WB Black 1TB Pcie4
Display(s) iiyama G-MASTER G4380UHSU-B1
Case Lian Li O11D Evo
Audio Device(s) Realtek ALC1220
Power Supply EVGA Supernova 1300 G2
Mouse Razer Deathadder Essential
Keyboard Razer Ornata Chroma
Software Win 11 PRO
Benchmark Scores Time Spy: 34857 Time Spy Extreme: 19629 Port Royal: 28542 Speed Way: 11257 (Hall of Fame)
I doubt it. The 3090 series didn't either, besides, mining is dead.

Which is why scalping now is very risky compared to 2020, when buying a scalped 3090 meant 2 months longer to ROI and they could've mined for 23 months anyway, so it didn't matter.

I see ebay is filling up with 4090s for 3000-4000 EUR by the hour :kookoo:
 
Joined
Dec 25, 2020
Messages
7,458 (4.97/day)
Location
São Paulo, Brazil
System Name "Icy Resurrection"
Processor 13th Gen Intel Core i9-13900KS
Motherboard ASUS ROG Maximus Z790 Apex Encore
Cooling Noctua NH-D15S upgraded with 2x NF-F12 iPPC-3000 fans and Honeywell PTM7950 TIM
Memory 32 GB G.SKILL Trident Z5 RGB F5-6800J3445G16GX2-TZ5RK @ 7600 MT/s 36-44-44-52-96 1.4V
Video Card(s) NVIDIA RTX A2000
Storage 500 GB WD Black SN750 SE NVMe SSD + 4 TB WD Red Plus WD40EFPX HDD
Display(s) 55-inch LG G3 OLED
Case Pichau Mancer CV500 White Edition
Power Supply EVGA 1300 G2 1.3kW 80+ Gold
Mouse Microsoft Classic IntelliMouse (2017)
Keyboard IBM Model M type 1391405
Software Windows 10 Pro 22H2
Benchmark Scores I pulled a Qiqi~
Which is why scalping now is very risky compared to 2020, when buying a scalped 3090 meant 2 months longer to ROI and they could've mined for 23 months anyway, so it didn't matter.

I see ebay is filling up with 4090s for 3000-4000 EUR by the hour :kookoo:

Yeah, I saw this coming. But if my hunch is right, it should self-correct fast, this scalping is mostly to profit on people who are hasty to get the new thing day one no matter what. This kind of always happened. But I have to be honest with you: early adopting these GPUs have "sucked" for a while now, because they made it a habit of releasing the full processor (coupled with some decent project improvements) as a mid-gen refresh, so you pay flagship money for a slightly cut down card these days.

Honestly, the market was significantly different back then. This time around I feel comfortable waiting for AMD's answer and the mid-generational refresh, it's not like my 3090 sucks :D
 
Joined
Jan 29, 2021
Messages
1,917 (1.31/day)
Location
Alaska USA
What's the EU MSRP for those? If you're thinking of US$ MSRPs, those don't include sales tax/VAT/GST, so you need to add however much your country charges on top of the US MSRP. Blame the silly Americans with their wildly variable sales tax rates.
No sales tax in the US state where I live. This silly American can get that card for MSRP. :)
 

3x0

Joined
Oct 6, 2022
Messages
974 (1.14/day)
Processor AMD Ryzen 7 5800X3D
Motherboard MSI MPG B550I Gaming Edge Wi-Fi ITX
Cooling Scythe Fuma 2 rev. B Noctua NF-A12x25 Edition
Memory 2x16GiB G.Skill TridentZ DDR4 3200Mb/s CL14 F4-3200C14D-32GTZKW
Video Card(s) PowerColor Radeon RX7800 XT Hellhound 16GiB
Storage Western Digital Black SN850 WDS100T1X0E-00AFY0 1TiB, Western Digital Blue 3D WDS200T2B0A 2TiB
Display(s) Dell G2724D 27" IPS 1440P 165Hz, ASUS VG259QM 25” IPS 1080P 240Hz
Case Cooler Master NR200P ITX
Audio Device(s) Altec Lansing 220, HyperX Cloud II
Power Supply Corsair SF750 Platinum 750W SFX
Mouse Lamzu Atlantis Mini Wireless
Keyboard HyperX Alloy Origins Aqua
And here are the performance jumps from the top flagship card in each generation to the next, based on TPU's own data:

Tesla 2.0 (GTX 285) --> Fermi (GTX 580) = 67% performance jump gen on gen
Fermi (GTX 580) --> Kepler (GTX 780 Ti) = 104% performance jump gen on gen
Kepler (GTX 780 Ti) --> Maxwell (Titan X) = 45% performance jump gen on gen
Maxwell (Titan X) --> Pascal (Titan Xp) = 72% performance jump gen on gen [Titan Xp and 1080 Ti were very similar, 1080 Ti a couple points faster]
Pascal (GTX 1080 Ti/Titan Xp) --> Turing (RTX 2080 Ti) = 39% performance jump gen on gen
Turing (RTX 2080 Ti) --> Ampere (RTX 3090 Ti) = 56% performance jump gen on gen
Ampere (RTX 3090 Ti) --> Ada (RTX 4090) = 45% performance jump gen on gen

If you assume the "true" Ada flagship will be a 4090 Ti, and that card will be 10% faster than the 4090, then Ada is a 59% performance jump gen on gen.
Thanks for summarizing the architectural improvements. I have added release dates and performance improvement per month into the calculation, which can provide an additional data point although the data takes into account manufacturing problems (delays) and lack of competition (ie nV sandbagging with the Kepler generation)

Tesla 2.0 (GTX 285) January 15, 2009--> Fermi (GTX 580) November 9, 2010 = 67% performance jump gen on gen (~22 months between releases, 3.05% perf jump per month)
Fermi (GTX 580) November 9, 2010 --> Kepler (GTX 780 Ti) February 19, 2013 = 104% performance jump gen on gen (~27 months between releases, 3.85% perf jump per month)
Kepler (GTX 780 Ti) February 19, 2013 --> Maxwell (Titan X) March 17, 2015 = 45% performance jump gen on gen (~25 months between releases, 1.8% perf jump per month)
Maxwell (Titan X) March 17, 2015 --> Pascal (Titan Xp) April 6, 2017 = 72% performance jump gen on gen (~25 months between releases, 2.88% perf jump per month)
Pascal (GTX 1080 Ti/Titan Xp) April 6, 2017 --> Turing (RTX 2080 Ti) September 27, 2018 = 39% performance jump gen on gen (~18 months between releases, 2.16% perf jump per month)
Turing (RTX 2080 Ti) September 27, 2018 --> Ampere (RTX 3090 Ti) March 29, 2022 = 56% performance jump gen on gen (~42 months between releases, 1.33% perf jump per month)
Comparing RTX 2080 Ti vs. the 3090 September 24, 2020 = 43% performance jump gen on gen (~24 months between releases, 1.79% perf jump per month)
Ampere (RTX 3090 Ti) March 29, 2022 --> Ada (RTX 4090) October 12, 2022 = 45% performance jump gen on gen (~7 months between releases, 6.42% perf jump per month)
Comparing RTX 3090 September 24, 2020 vs. the 4090 October 12, 2022 = 64% performance jump gen on gen (~25 months between releases, 2.56% perf jump per month)

There were outliers in the perf. per month improvement calculation Turing (RTX 2080 Ti) September 27, 2018 --> Ampere (RTX 3090 Ti) March 29, 2022 and so I added the 3090 into the equation with its Sep. 24th 2020 release date.

Your hierarchy is:
Kepler
Pascal
Fermi
Ada
Ampere
Maxwell
Turing

While the perf. per month hierarchy is:
Kepler
Fermi
Pascal
Ada
Turing
Maxwell
Ampere
 
Top