This should settle the NV memory gimping issue.

No, it is more complex than that, but what would be nice when GPU that play well at high but not at ultra happen, screenshots of how much quality is lost, in many game it is hard to see a difference, a bit like we do for GPU not able to run raytracing in some title, how much it is lost.

It just focus the issue on the 128 bits memory bandwith gimping more than the raw total. a $500 new 2023 GPU with 16GB of VRAM should not lose to 2020 "$500" 3070 so much .
 
It's worth noting they don't specify game settings for any of the games. It seems like they may be only testing raster perf, given their benchmarks for various games fairly closely match prior review framerates for non-RT settings (eg: RE4R).
 
They claim to test at max settings and RT results were done also. Still believe mem BW is king when comparing a 3080 to a 4060Ti.
 
It's worth noting they don't specify game settings for any of the games. It seems like they may be only testing raster perf, given their benchmarks for various games fairly closely match prior review framerates for non-RT settings (eg: RE4R).
If not specified it is maximum according to the
https://www.techpowerup.com/review/nvidia-geforce-rtx-4060-ti-16-gb/4.html
  • All games are set to their highest quality setting unless indicated otherwise.
TPU always test raster alone and then RT :
https://www.techpowerup.com/review/nvidia-geforce-rtx-4060-ti-16-gb/33.html

relative-performance-rt-2560-1440.png
 
Last edited:
I still believe we focus too much on specifications and measurements or what I call "measurbrators". People are moaning about the VRAM limits as if their inheritance or financial livelihood was stolen by NVIDIA.

Most in game FPS differ by a small or statistically irrelevant amount. People need to but more pressure on developers who half ass or just port a game from console to PC to make a profit.
 
I still believe we focus too much on specifications and measurements or what I call "measurbrators". People are moaning about the VRAM limits as if their inheritance or financial livelihood was stolen by NVIDIA.

Most in game FPS differ by a small or statistically irrelevant amount. People need to but more pressure on developers who half ass or just port a game from console to PC to make a profit.
That's because nvidia is attempting to create more market segmentation by artificially limiting the amount of VRAM certain cards have. It's dumb and they should be ridiculed for it especially with the stupid prices they are charging for the 40 series cards.
 
That's because nvidia is attempting to create more market segmentation by artificially limiting the amount of VRAM certain cards have. It's dumb and they should be ridiculed for it especially with the stupid prices they are charging for the 40 series cards.
Nope. Now we just deflect and claim reviews don't matter. Up until now everyone has been fine with X is 5% faster than Y, but now that's becoming inconvenient...
 
Need to read the review if they has some frame time analysis as like SLI benchmarks you actually need to watch the benchmark or have frame time over time graphs. I miss [H] GPU reviews as they really gave you a good idea what you'll see subjectively in regards to smoothness/stutter. The biggest issue for me with any CPU/memory limitation is it breaks VRR so you tend to see stutter when hitting a CPU or GPU memory bottelneck.
 
4060 Ti 8GB, 1080P high setting + RT (NOT ultra) has massive FPS drops when panning the camera. I get that folks want 8GB to be good enough for the next decade, but the signs have been there for awhile now - including game devs stating that they will stop optimizing for 8GB cards in the near future:

 
  • Like
Reactions: Kzoak
like this
Look at the hardware unboxed review and you see for a FACT that 8gb is in issue in some games RIGHT NOW even at 1080p in some cases. It is ridiculous that it has to be said over and over that it is not just about average fps in some cases either as there are cases where huge stutters can occur or proper textures do not even load due to lack of VRAM.


 
Last edited:
and you see for a FACT that 8gb is in issue in some games RIGHT NOW even at 1080p in some cases
Has anyone doubt that at some settings, some games in some section of them it is an issue ? (if you include RT ultra has one of those possible setting, there is not many GPU that has no issues in some games at 1080p right now)

The issue with the video (and maybe youtube compression does not allow it), but after showing issue at ultra, does not look how different the graphic look at the level (say high) when vram issues disappears, i.e. is it a real issue or one that does not matter because the visual difference is small (like a lot of case of RT ultra vs high)
 
Last edited:
HBU review is much better review for anyone actually going to buy either 4060ti. Frame time graphs and more in depth real world gameplay testing. It's clear to me if someone is buying any new mid range graphics card should be getting one with at least 12/16GB or probably buying another video card in a year or two. Curious to see similar frame time graphs for 10GB 3080s at 1440p.
 
If not specified it is maximum according to the
https://www.techpowerup.com/review/nvidia-geforce-rtx-4060-ti-16-gb/4.html
  • All games are set to their highest quality setting unless indicated otherwise.
TPU always test raster alone and then RT :
https://www.techpowerup.com/review/nvidia-geforce-rtx-4060-ti-16-gb/33.html
Thanks. Somehow missed that line despite seeing the page.

Strange the RT benchmarks are aggregated on a separate page. In per-game reviews from their site they're placed immediately after the raster benchmarks.
 
That's because nvidia is attempting to create more market segmentation by artificially limiting the amount of VRAM certain cards have. It's dumb and they should be ridiculed for it especially with the stupid prices they are charging for the 40 series cards.
Show me the last time a 60-class card got the same amount of memory as the 80-class card. They're not "artificially" limiting anything. Market segmentation is done in every industry.
All I got out of that is how $150 looking boards are now $500.

View attachment 585720
It's not that much smaller compared to the 4090. The 4090 is only bigger because of the extra power phases required (and more room for the VRAM). My poor MS Paint skills with the 4060 on top scaled to the 4090:

1690393698152.png


More compact PCB does not mean it's cheap.
 
Show me the last time a 60-class card got the same amount of memory as the 80-class card. They're not "artificially" limiting anything. Market segmentation is done in every industry.

It's not that much smaller compared to the 4090. The 4090 is only bigger because of the extra power phases required (and more room for the VRAM). My poor MS Paint skills with the 4060 on top scaled to the 4090:

View attachment 585733

More compact PCB does not mean it's cheap.
I wasn't just commenting on the size of the board. Look at the components and cheapened out memory interface and x8 PCI-E. Not that long ago that was considered low end to maybe the low end of mid-range. (i.e. at best a 50 branded card for Nvidia).

As for 8GB vs 16GB, I don't see the point of the 16GB 4060 Ti. At 128-bits, that's just wasted.
 
Last edited:
Back
Top