So, NVIDIA just launched GeForce RTX 4060Ti graphic card. They officially launched 3 models at following MSRP prices which will be gradually released from this point after RTX 4060Ti 8GB model:
- RTX 4060 8GB ($300)
- RTX 4060Ti 8GB ($400)
- RTX 4060Ti 16GB ($500)
The way NVIDIA positioned these models is that you’re heavily steered towards the 500 bucks model because the non-Ti model is just pathetic and has memory problem because a lot of games are showing issues with cards that only have 8GB of VRAM and that’s at only 1080p! In year 2023! Imagine how it’ll fare 2, 3 or more years in the future? Remember, mid range cards buyers don’t swap them every year or two, they hold them for much longer. So, going with Ti model solves nothing, you have to go to Ti 16GB model. For fucking 500 bucks. FIVE HUNDRED for a mid range card that has real world performance advantage over last generation card released almost 3 years ago basically within margin of error. I shit you not, RTX 4060Ti is just single digit percent better than RTX 3060Ti. And percent numbers are always huge and when you translate them into framerate it’s often just 3 or 5 frames per second difference. So, basically margin of error.
And reason for such lackluster showcase is the fact that NVIDIA cut down RTX 4060 so heavily on all ends it basically demolished any performance advantages of newer core architecture based on Ada Lovelace compared to Ampere architecture powering RTX 3000 series.
First they gave zero VRAM upgrade and it’s stuck at 8GB. Like mentioned, there are games that already saturate that at only 1080p. Then they heavily cut down the memory bus from 256bit found on RTX 3060Ti to just 128bit on RTX 4060Ti. NVIDIA’s excuse was increasing L2 cache within GPU core from 4MB to 32MB, but thing with caches is, they only work under ideal conditions. And when you have ideal conditions, it works beautifully, but when it doesn’t, you’re basically running at half a memory bus width and that is pretty severe downgrade.
To add insult to injury, NVIDIA also cut down PCIe bus width too. Instead of running at full x16 rate, PCIe bus only runs at 8x rate. That in itself isn’t an issue on modern PCIe 4.0 slot motherboard, but imagine if you have older system you’re upgrading with only PCIe 3.0 or even 2.0 ? That means PCIe slot runs way way slower. And one may say, you don’t need so much bandwidth on the slot anyway with mid end graphic card, but there is the problem, those 8GB VRAM is not enough and when game can’t fit resources on onboard memory of graphic card, it’ll spill it over PCIe bus into system memory.
Death by a thousand cuts
It may all sound insignificant on the surface, but when you lose a percent here and percent there on bunch of levels, you’re suddenly losing a lot of performance in the end, so much in fact that all the architectural advantages of Ada Lovelace on RTX 4060Ti are entirely negated to a point it’s basically as fast as RTX 3060Ti from years ago because NVIDIA just cut it down so hard. And don’t get me started with Resizable bar (ReBAR), if your system doesn’t support it, too little memory, cut down memory bus and cut down PCIe slot and it’ll just get even worse in this situation.
Did all this cutting of features away bring users any advantages? No, because the damn thing is as fast as RTX 3060Ti in the end, yet still costs the same (or more for 16GB model). You know what it did tho? It increased PROFIT MARGINS for NVIDIA. More VRAM costs money. Memory costs in peanuts for end users these days and for company buying in bulk with heavy discounts, even less. All straight into NVIDIA pocket when you pay $100 more for pathetic 8GB extra and also for the fact NVIDIA doesn’t give you more as standard. Smaller bus width, the wider the memory interface, it costs more because the GPU core needs more connections to the memory, costing more to manufacture in the end. Since NVIDIA cut it down, the difference goes into their pockets, because you as end user sure as hell aren’t benefiting from it. Cut down PCIe slot rate? Same story. Less complexity, less work, more goes into NVIDIA’s pockets. All the cuts weren’t done to make card cheaper for end user, all the cuts were done to ramp up profits for NVIDIA. Greedy bastards.
Are there even any benefits found on RTX 4060Ti ?
Only real benefit of RTX 4060Ti is slightly better power efficiency, AV1 encoder and decoder and DLSS3 which NVIDIA so desperately shills on all ends showcasing it how much better performance it makes, by showing some bullshit X times better graphs, entirely ignoring the fact that a) DLSS3 isn’t really supported in many games so they are selling performance advantage on premise of future updates of games supporting it b) ignoring the fact that inserting fake frames doesn’t make performance better. It makes smoother motion, but adding extra latency and visual artefacting. But they are just selling it as MOAR FRAMES PER SECOND and showing it in graphs with mega differences in framerate as instant winner over RTX 3000 series which don’t support frame generation despite having hardware support for it (optical flow which is supported since Turing architecture as far as I’ve checked).
Conclusion
If you already have RTX 3060Ti card, RTX 4060Ti is absolutely to be avoided. You’ll not be gaining anything unless hardware AV1 support is your absolute requirement. Everything else, including DLSS3 Frame Generation are all worthless benefits. You buy things on what they offer NOW, not on what they may offer in the future. DLSS performance gains are “nice to have”, because if card performs well now “as is”, slapping DLSS on it just means it’ll perform even better. If card performs the same as last generation from years ago and only advantage is “promised” performance gain with fake frames by using DLSS3, do you really want that? For full price? Yeah, I think not. Currently only few games support it and most existing games will never get support unless they are actively being developed by dedicated developers who are into this stuff. Most aren’t and never add such stuff to old games.
And even if you’re running something like RTX 2060, RTX 2060 Super or even older GTX 1600 or GTX 1650 or equivalent Radeon RX 480/570/580/590, I wouldn’t just rush to buy RTX 4060Ti either. AMD’s offerings are just around the corner with new RX 7600 cards and also price adjustments of older generations like RX 6650 XT or even RX 6700 series. Because currently, AMD’s offerings are neck on neck, they just have slight price disadvantage. But you can be sure AMD will adjust them accordingly pretty soon. And AMD also has FSR which is similar to DLSS. It has similar benefits and issues and it also falls into “nice to have” category, but I wouldn’t base my purchase on it either. And lastly, Intel’s ARC cards are pretty aggressive price wise and Intel made quite a lot of improvements to them.
My advice, don’t rush for RTX 4060Ti and wait for what AMD has to say. And then wait a bit for price adjustments depending on how RX 7600 will perform. You can still buy RTX 4060Ti then if numbers won’t be favorable to AMD. I personally think NVIDIA chased profits so hard by cutting things away they entirely fucked up this card that could otherwise be amazing if it had same raw specs as RTX 3060Ti, but with all the benefits of Ada Lovelace architecture. Same class of every new generation always brought pretty significant performance gains where now, it’s literally stagnating on same level as almost 3 years old card. That’s just greedy and bad. And this is the class of card that will be sold the most to the masses. All other higher models had gains, but not 60 series. Pathetic NVIDIA, really pathetic.
Angry Sheep Blog 