8GB VRAM not enough (and 10 / 12)

[BFG10K]

8GB

• Dying Light 2, Battlefield 2042, Avatar Frontiers of Pandora, Call of Duty Modern Warfare 3, Starfield, Forza Horizon 5, Horizon Forbidden West, Alan Wake 2, The Last of Us, Tekken 8, FC 24, Detroit Become Human, Death Stranding, Alone in The Dark, Overwatch 2
  3060 12GB higher 1% low than 4060 8GB https://www.youtube.com/watch?v=gIfz57wmywg
• Cyberpunk 2077, Diablo 4, Hogwarts Legacy, Horizon Zero Dawn, Microsoft Flight Sim 2020, Witcher 3, Watch Dogs Legion skip to 4K + DLSS for each game https://www.youtube.com/watch?v=tPiaWWd0xGc
• Ratchet & Clank, Plague Tale Requiem, Last Of Us, Jedi Survivor, Far Cry 6, Hogwarts Legacy, Forza Horizon 5, Cyberpunk 2077 https://www.youtube.com/watch?v=Wr3MckDxMfE&pp=ygUTMTZnYiB2cyA4Z2IgNDA2MCB0aQ==
• Far Cry 6, Godfall, Hogwarts Legacy, Forza Horizon 5, Forspoken, Doom Eternal, Watchdogs Legion https://youtu.be/fJc--C01P90?t=268
• Last Of Us, Resident Evil 4, Callisto Protocol, Plague Tale Requiem, Halo Infinite, Forspoken https://youtu.be/2_Y3E631ro8?t=429
• Last Of Us, Hogwarts Legacy, Resident Evil 4, Forspoken, Plague Tale Requiem, Callisto Protocol https://youtu.be/Rh7kFgHe21k?t=288
• Hogwarts Legacy, Last Of Us, Spider Man Remastered, Doom Eternal, Forza Horizon 5 https://youtu.be/R943CbDTq_s?t=1473
• Horizon Forbidden West 4060 has issues @ 1080p + DLSS unless you switch off frame generation and reduce settings to high https://youtu.be/3AyKvI23VGw?t=233
• Horizon Forbidden West 3070 requires reducing textures 2 notches https://youtu.be/xTwEGy6HHKo?t=351
• Deathloop https://youtu.be/lu_XHyO-6zY?t=918
• Ghost Recon Breakpoint https://youtu.be/iUGppQVpXzU?t=529
• Resident Evil 2 https://youtu.be/Aa-gsgRjMII?t=796
• Talos Principle 2 https://youtu.be/K6FATQAuEwI?t=337
• Spider Man Remastered https://youtu.be/vRVR_plVSlc?t=1060
• Witcher 3 https://youtu.be/vRVR_plVSlc?t=1240

Horizon Forbidden West 3060 is faster than the 2080 Super despite the former usually competing with the 2070. Also 3060 has a better 1% low than 4060 and 4060Ti 8GB.

Spoiler

Resident Evil Village 3060TI/3070 tanks at 4K and is slower than the 3060/6700XT when ray tracing:

Spoiler

Company Of Heroes 3060 has a higher minimum than the 3070TI:

Spoiler

10GB / 12GB

• Plague Tale Requiem 3080 10GB tanks if you enable ray tracing, would be fast enough if it had more VRAM because if you stop moving, the framerate stabilizes to >60FPS https://youtu.be/7zjrMww3fAc?t=496
• Hogwarts Legacy 4K + RT, 4060Ti 16GB is faster than 4070 12GB https://youtu.be/X-h_R488Mq8?t=191

Reasons why still shipping 8GB since 2014 isn't NV's fault.

Spoiler

1. It's the player's fault.
2. It's the reviewer's fault.
3. It's the developer's fault.
4. It's AMD's fault.
5. It's the game's fault.
6. It's the driver's fault.
7. It's a system configuration issue.
8. Wrong settings were tested.
9. Wrong area was tested.
10. Wrong games were tested.
11. 4K is irrelevant.
12. Texture quality is irrelevant as long as it matches a console's.
13. Detail levels are irrelevant as long as they match a console's.
14. There's no reason a game should use more than 8GB, because a random forum user said so.
15. It's completely acceptable for the more expensive 3070/3070TI/3080 to turn down settings while the cheaper 3060/6700XT has no issue.
16. It's an anomaly.
17. It's a console port.
18. It's a conspiracy against NV.
19. 8GB cards aren't meant for 4K / 1440p / 1080p / 720p gaming.
20. It's completely acceptable to disable ray tracing on NV while AMD has no issue.
21. Polls, hardware market share, and game title count are evidence 8GB is enough, but are totally ignored when they don't suit the ray tracing agenda.

According to some people here, 8GB is neeeevaaaaah NV's fault and objective evidence "doesn't count" because of reasons(tm). If you have others please let me know and I'll add them to the list. Cheers!

 

[poke01]

Sorry I just have to, because I was able to take #1 from a 4080S

View attachment 103514View attachment 103516

Why doesn’t anyone pair a 4090 with a fx 8350?

Probably cause they don’t want their house to blow up. 🤪

Great score btw, what was the power consumption of the 7900xtx in that bench?

 

[marees]

Just watched the video

Testing the 8gb card on a Pci 2.0/3.0 bus effectively turned it into an IGP class card

 

[marees]

Just watched the video

Testing the 8gb card on a Pci 2.0/3.0 bus effectively turned it into an IGP class card

Hope all this activism leads to a 5060 12gb/16gb

 

[igor_kavinski]

Hope all this activism leads to a 5060 12gb/16gb

Even 5050 should have 10GB minimum. I'm sick of 8!

Jensen can put 6GB/8GB on 5030/5040.

 

[ryanjagtap]

Even 5050 should have 10GB minimum. I'm sick of 8!

So 160-bit bus minimum.......
What about PCIE lanes? x8 or x16 minimum?

 

[igor_kavinski]

What about PCIE lanes? x8 or x16 minimum?

x8 should be OK. x4 should be reserved for stuff used to play DX11 games.

 

[SolidQ]

Even 5050 should have 10GB minimum.

GB 206 128 bit, seems to fight vs N44

Jensen can put 6GB/8GB on 5030/5040.

200 and 300$

 

[mikeymikec]

I'd like to know whether limiting lanes / size of interface actually results in saving money or whether it's something they like to do purely to drive upgrades to the higher end.

 

[Iron Woode]

I'd like to know whether limiting lanes / size of interface actually results in saving money or whether it's something they like to do purely to drive upgrades to the higher end.

probably more of the latter than the former.

 

[SolidQ]

esting the 8gb card on a Pci 2.0/3.0 bus effectively turned it into an IGP class card

Some people will pair with B840

 

[Aapje]

I'd like to know whether limiting lanes / size of interface actually results in saving money or whether it's something they like to do purely to drive upgrades to the higher end.

It saves silicon space and having to add the connections from the silicon to the outside of the chip. It saves a bit of money.

 

[DAPUNISHER]

It saves silicon space and having to add the connections from the silicon to the outside of the chip. It saves a bit of money.

In your considered opinion, do you think that is the main motivator, or is it planned obsolescence?

 

[Aapje]

In your considered opinion, do you think that is the main motivator, or is it planned obsolescence?

Just saving money. The cards typically have enough bandwidth to have minimal loss in performance with the previous generation PCIe at half bandwidth. This current issue is more of a confluence of factors, with them running out of VRAM and then depending on the bus to swap in all the textures that won't fit in the VRAM. Such a confluence of factors is hard to predict or to plan, so it is probably not something that they intended or even saw coming. Even the limited VRAM is probably just about Jensen trying to save money by keeping the bus size small, rather than trying to create planned obsolescence.

My impression is that Jensen got in a mindset where the good stuff is for business and he believes that we should be grateful for the gimped stuff he is sending to consumers. And with their market dominance, he probably just assumes that developers will optimize to Nvidia hardware, rather than abandon the 8 GB card owners.

But of course, I don't know what Jensen is thinking. This is mostly just a psychological analysis from how I see him act.

 

[DAPUNISHER]

Thanks for the detailed reply. I accept that explanation as the primary motivator.

I do think planned obsolescence is an ancillary motivator though. The fact AMD and Intel upscaling and frame generation work on everything is evidence enough for me, that the reason Nvidia did not provide such a solution, was to force users off their old cards. Anemic vram too. Recently saw the 980ti vs new games and the conclusion was it has held up surprisingly well, but would have been significantly better had it been given 8GB of vram.

 

[Aapje]

It's hard to distinguish between planned obsolescence and Jensen trying to make the new gen more attractive by withholding software improvements from the last gen. In a way it's the same thing, but also, it's not.

 

[DaaQ]

Why doesn’t anyone pair a 4090 with a fx 8350?

Probably cause they don’t want their house to blow up. 🤪

Great score btw, what was the power consumption of the 7900xtx in that bench?

Unfortunately I had to turn off HWinfo to baing the extra points to make 1st, I can tell you im not on a flashed VBIOS, it is the Nitro+ OC BIOS. I can give a screen of SN where I think it is glitched but not the 8500point bad glitch, I didn't have a valid result showing from this one. It was from 3 days earlier.

here is the valid result but no Hwinfo captured.

Thanks for the detailed reply. I accept that explanation as the primary motivator.

I do think planned obsolescence is an ancillary motivator though. The fact AMD and Intel upscaling and frame generation work on everything is evidence enough for me, that the reason Nvidia did not provide such a solution, was to force users off their old cards. Anemic vram too. Recently saw the 980ti vs new games and the conclusion was it has held up surprisingly well, but would have been significantly better had it been given 8GB of vram.

I ran D4 on a GTX 780 without much issues using acceptable settings to me, looked better than PS4 OG version. I did use the FSR balanced option in game setting option to maintain higher framerates.
Here is a screen capture from that. I will add I never went over the 3GB frame buffer. Zoom in to see the RT stats. (riva tuner)

 

[DaaQ]

Why doesn’t anyone pair a 4090 with a fx 8350?

Probably cause they don’t want their house to blow up. 🤪

Great score btw, what was the power consumption of the 7900xtx in that bench?

Here is a quick 24/7 settings run just to grab the HWinfo screen.

 

[marees]

Just watched the video

Testing the 8gb card on a Pci 2.0/3.0 bus effectively turned it into an IGP class card

Below is from a year ago (maybe someone already posted this, but I found it relevant to the discussion here


The RTX 4060ti utilizes only 8 PCIe lanes, which does not make a difference for PCIe 4.0 systems. However, for older PCIe 3.0 systems, this acts as a limitation to the card's performance.

In practice, the performance of the RTX 4060ti on these systems is even closer to that of the RTX 3060ti.

Der8auer did a video about it.

RTX 4060ti is even worse on PCIe 3.0

The RTX 4060ti utilizes only 8 PCIe lanes, which does not make a difference for PCIe 4.0 systems. However, for older PCIe 3.0 systems, this acts as a limitation to the card's performance. In practice, the performance of the RTX 4060ti on these systems is even closer to that of the RTX 3060ti...

www.neogaf.com

 

[marees]

Below is from a year ago (maybe someone already posted this, but I found it relevant to the discussion here


The RTX 4060ti utilizes only 8 PCIe lanes, which does not make a difference for PCIe 4.0 systems. However, for older PCIe 3.0 systems, this acts as a limitation to the card's performance.

In practice, the performance of the RTX 4060ti on these systems is even closer to that of the RTX 3060ti.

Der8auer did a video about it.

RTX 4060ti is even worse on PCIe 3.0

The RTX 4060ti utilizes only 8 PCIe lanes, which does not make a difference for PCIe 4.0 systems. However, for older PCIe 3.0 systems, this acts as a limitation to the card's performance. In practice, the performance of the RTX 4060ti on these systems is even closer to that of the RTX 3060ti...

www.neogaf.com

Tl:dr when GPU is VRAM limited then pcie bandwidth makes a huge difference

RTX 4060 Ti Should Be Avoided On PCIe 3.0 Boards Due To Bottlenecks

GeForce RTX 4060 Ti, due to its limited X8 interface, is bandwidth limited on the PCIe 3.0 interface, negatively affecting framerates.

tech4gamers.com

 

[Mopetar]

It's hard to distinguish between planned obsolescence and Jensen trying to make the new gen more attractive by withholding software improvements from the last gen. In a way it's the same thing, but also, it's not.

PHY is actually expensive area wise as it doesn't scale at all with new nodes so if you were to do a node shrink on some existing die you'd wind up with a smaller chip, but a larger percentage of it would by taken up by the physical interconnects.

Reducing the size (either for the memory bus size or the PCI lanes) is usually offset by increased speed resulting in more bandwidth. As other posts have pointed out it's not usually a problem unless you're limited to older generation speeds and even then it's usually not going to be noticeable.

Of course when you're VRAM limited the card needs to use the PCI bus in order to load data from RAM that couldn't fit in VRAM and that can cause noticeable performance drops.

Cutting out 8 PCI lanes or a 32 bits of memory bus saves a lot of area and when the newest nodes are expensive as they are it's easy to understand why both Nvidia and AMD will go that route. For the smaller dies it probably lets them squeeze out a few dozen additional dies per wafer.

 

[DAPUNISHER]

I ran D4 on a GTX 780 without much issues using acceptable settings to me, looked better than PS4 OG version. I did use the FSR balanced option in game setting option to maintain higher framerates.
Here is a screen capture from that. I will add I never went over the 3GB frame buffer. Zoom in to see the RT stats. (riva tuner)

Exactly. Instead of being forced to turn down native resolution you are able to use another vendor's upscaling tech. If FSR was not vendor agnostic you would have no recourse since Nvidia walled it off. They could have taken the same approach Intel has of offering a version that works on other vendors, and a version that uses their dedicated hardware.

Kepler is the poster child for badly aging Nvidia cards, and yes it's partially due to cheaping out on vram. Hardware Lab used the 780 TIE vs 290X to demonstrate why the TPU relative performance chart is a terrible idea.The 3GB of vram was one of the factors in why it aged so much worse. Would not start Plague Tale Requiem, there is no redemption for it in Red Dead, getting demolished by Hawaii in other tested games as the years go by through 2022. It'd be even worse now. 4GB may only be 1GB extra, but HL shows it makes a difference anyways. There are other reasons of course, but vram is on the list.

It's a long vid but skim through the games and commentary and it is brutal. Not only for that old 3GB card, but for the idea that the TPU charts have any relevance. They have the ARC A380 with the same performance as your 780 and IMO it is beyond wrong to even have them on the same chart. One has modern features and drivers, the other is the 780.

The vid will start with Shadow of the Tomb Raider where much more than fps is seriously wrong on Kepler. BTW there is no equivalent to R.ID for Nvidia on windows is there?

 

[poke01]

BTW there is no equivalent to R.ID for Nvidia on windows is there?

No there isn’t. That being AMD dropped the ball when it comes to offical driver support for older GPUs pre-RDNA. Nvidia at least supports the Maxwell cards.

It's a long vid but skim through the games and commentary and it is brutal. Not only for that old 3GB card, but for the idea that the TPU charts have any relevance. They have the ARC A380 with the same performance as your 780 and IMO it is beyond wrong to even have them on the same chart. One has modern features and drivers, the other is the 780.

The vid will start with Shadow of the Tomb Raider where much more than fps is seriously wrong on Kepler.

TPU charts are not for making comparisons and do a bad job at conveying all the characteristics of a card.


Kepler was a mid architecture but it was efficient. It wasn’t till Pascal that Nvidia cards aged more gracefully. I say Turing will age the best over time and then the upper end Ada cards.

Ampere is another Kepler in terms of vram and it doesn’t help it’s made a 8nm Samsung node. Its tensor cores and RT cores are good but VRAM will hold it back. Mainly the 8GB and 10gb cards. 3090 is an exception.

 

[DAPUNISHER]

No there isn’t. That being AMD dropped the ball when it comes to offical driver support for older GPUs pre-RDNA. Nvidia at least supports the Maxwell cards.

Including Maxwell 2.0 so games don't throw up a warning screen about out of date drivers is a good thing of course. But performance and compatibility issues do not reflect any real attention being paid to them. Some games are broken on any given driver release with those cards. Guardians of the Galaxy and Halo Infinite among them. Some releases work ok others the performance is completely nerfed. There are other games, those are the one off the cuff.

R.ID gives old Radeon owners recourse to this day. And those games tested by HL were all done on the last drivers for both cards. Both of which had their driver support ended that year of 2022. Intro mentions all the accusations that Nvidia stopped optimizing for Kepler long before that, and the games tested support the contention.

TPU charts are not for making comparisons and do a bad job at conveying all the characteristics of a card.

I must be misunderstanding the bold. The relative performance chart is - "Based on TPU review data: "Performance Summary" at 1920x1080, 4K for 2080 Ti and faster." that is exactly what it is used for by tech tubers using it to show a "general idea of how cards stack up" all the time. It is thrown around in forums that way quite often too.

 

[poke01]

I must be misunderstanding the bold. The relative performance chart is - "Based on TPU review data: "Performance Summary" at 1920x1080, 4K for 2080 Ti and faster." that is exactly what it is used for by tech tubers using it to show a "general idea of how cards stack up" all the time. It is thrown around in forums that way quite often too.

My bad, should have said it shouldn't be the definitive way to get performance numbers.

 

[DaaQ]

Exactly. Instead of being forced to turn down native resolution you are able to use another vendor's upscaling tech. If FSR was not vendor agnostic you would have no recourse since Nvidia walled it off. They could have taken the same approach Intel has of offering a version that works on other vendors, and a version that uses their dedicated hardware.

Kepler is the poster child for badly aging Nvidia cards, and yes it's partially due to cheaping out on vram. Hardware Lab used the 780 TIE vs 290X to demonstrate why the TPU relative performance chart is a terrible idea.The 3GB of vram was one of the factors in why it aged so much worse. Would not start Plague Tale Requiem, there is no redemption for it in Red Dead, getting demolished by Hawaii in other tested games as the years go by through 2022. It'd be even worse now. 4GB may only be 1GB extra, but HL shows it makes a difference anyways. There are other reasons of course, but vram is on the list.

It's a long vid but skim through the games and commentary and it is brutal. Not only for that old 3GB card, but for the idea that the TPU charts have any relevance. They have the ARC A380 with the same performance as your 780 and IMO it is beyond wrong to even have them on the same chart. One has modern features and drivers, the other is the 780.

The vid will start with Shadow of the Tomb Raider where much more than fps is seriously wrong on Kepler. BTW there is no equivalent to R.ID for Nvidia on windows is there?

No disagreements with you here, I really wanted a 290X but the msrp and availability I just, couldn't IIRC the mining had begun.
Don't get me wrong, I bought that 780 at the time it was only one available at msrp with waterblock compatibility. EVGA ACX.

I did run some native tests too, was 45-60 which 60Hz is max the card would link over HDMI, moving to DP it allowed me to select 100Hz, but had some really weird post issues, like no post screen until windows loaded, couldn't see BIOS.

I did grab an OG Titan for the 6GB VRAM to compare, but I never got around to putting the water block on it. I did OC the snot out of the 780. Seemed like the legacy drivers might have removed the greenlight stuff from back then. But I am not sure. I was able to apply higher voltages than the old threads from the time were reporting.

Anyway, waiting on Z53D and new motherboards to see if the memory speeds improve.

I am keeping my PCIE soundcard until it dies though.