Quote from: Sеrgey on January 04, 2024, 09:31:28

Quoteindividual people are modding their 4070/Tis to ~1 Liter

Do they do that by modifying the cooling solution?

The option of turning to water cooling is always there, for those brave enough to go that route...

Yes they install a compatible cooler from another GPU onto it usually, and a fan swap. Sometimes more custom than that.

Watercooling exists but when the PCBs for the GPUs are >200mm to begin with, you can't get a <200mm card. Or they make all 4070 PCB <200mm but then make no <200mm cards lol I don't understand (If you were going to watercool it would be 99% silly to watercool a 4070/Ti instead of a 4080/90). There also needs to be waterblocks accordingly (idk not much demand for <200mm single slot waterblock lol)

The performance graph per 1W of consumption becomes flatter every year. Not surprisingly, the only way to significantly increase performance is through ever-increasing consumption, despite the gradual reduction in transistor size. The silicon dead end is getting closer...

And the new product must somehow continue to be sold so that the life of the company owners is still in chocolate. As long as the masses believe that a new product gives them real benefits, they will continue to pay what they ask (at least those who still belong to the real "middle class"). Once trust is lost (or new technologies simply become unavailable due to increased costs), there will be a dead end for both the mass consumer and the manufacturers who rely on mass demand.

That is why more and more manufacturers are stopping selling goods for the "middle class", focusing primarily on the poor and the very rich.

Quoteindividual people are modding their 4070/Tis to ~1 Liter

Do they do that by modifying the cooling solution?

The option of turning to water cooling is always there, for those brave enough to go that route...

Quoteidiotic power levels and larger coolers at idiotic prices with idiotic memory setups with idiotic naming schemes/segmentation lol

Truth to be told, many people with enough cash to throw at an RTX 4080 also have enough cash to pay their electricity bills. There is a market for those 300-plus watt graphics cards. The question is, how large is that market?

QuoteI think they could do lower TDP just fine since there's no competition.

Yes... AMD is just utterly unable of maintaining an offensive on two fronts at the same time. It's either CPUs or GPUs. Not both. With Ryzens hitting new highs every year, one can't help but feel pity for Radeons.

Thank you so so much for this article, I liked the observation in the trend/pattern of board power over the years


instead of price to performance, we should do "price to performance to power" / normalize for all 3, since you're generally paying for better efficiency as you pay for a better card, and less if you buy an older cheaper card (that still has the same performance as a new midrange card e.g.)













I personally just wanted an ITX / low length/single fan RTX 4070 with 16GB VRAM. Or something like the 1080Ti mini for the 4070Ti (Super/16GB VRAM).

 it's ~200W just like the numerous ITX 1080/2070(S)/3060Ti, all ~1 "Liter" in size, so ~200 "Watt per Liter"

We should be getting better/more efficient/smaller cooler designs over time, not larger


What's so so sad is that individual people are modding their 4070/Tis to ~1 Liter. so so sad how companies with 100s of (paid) ppl don't want to / cannot make a product that already kinda used to exist 6-7 years ago but 1 unpaid passionate person can

There is the RTX 4000 SFF / 4500 if you have double(?) money it's worth lol. People were using A4000 for $500 despite being similar performance to a 3060Ti (but <70%


I should add here that technically there is already a "<=175W 4080/4070", the Laptop 4090/4080 lol

Which in turn is kind of sad because there used to be a 200W 2080 desktop/laptop that didn't need this awful segmentation/naming/power limiting/scamming

More efficient GPUs to just run them at idiotic power levels and larger coolers at idiotic prices with idiotic memory setups with idiotic naming schemes/segmentation lol

AMD dGPUs are seemingly non existent in laptops, which is kinda funny because there's little reason to buy a non AMD CPU laptop other than supply/them existing (Intel = bad battery / bad iGPU until MTL, it's fine in larger laptops I guess). RDNA3 is significantly less efficient than RTX4xxx/Ada Lovelace anyways

Quote from: Fierce.XT on January 03, 2024, 12:23:36Pushing a chip beyond its efficiency limit is a good thing from the company's point of view. It makes a smaller chip do more, and considering costs are growing exponentially, it's also the only way to keep your profit margins healthy.

So I don't believe the TDP will be lower.

Didn't realize nVidia have been struggling keeping their profit margins healthy recently.

Do they need their chips to do more though when AMD is unable to catchup to the 4090 (which wasn't even the fully unlocked chip but a cut down version) and apparent rumors of RDNA4 high end being cancelled, so it'll only target mid tier RTX 4070 Ti - RX 7900 XT perf?

I think they could do lower TDP just fine since there's no competition. Hell, they don't even have to release a new arch considering the stockpile of current RTX 4000 supply. They could rebadge the entire current lineup as RTX 5000, drop the price ever so slightly and it'd still be competitive, wouldn't it?

Pushing a chip beyond its efficiency limit is a good thing from the company's point of view. It makes a smaller chip do more, and considering costs are growing exponentially, it's also the only way to keep your profit margins healthy.

So I don't believe the TDP will be lower.