RTX 4060 Ti review roundup depicts GPU performing 12% better vs RTX 3060 Ti but failing to offer any performance/price advantage

[Redaktion]

NVIDIA released the RTX 4060 Ti on May 24 to a generally cold reception. According to various reports, consumer interest in the GPU was quite low and much of it could be due to the bad launch reviews. 3DCenter has now published a launch review roundup of the board alongside the AMD Radeon RX 7600.

https://www.notebookcheck.net/RTX-4060-Ti-review-roundup-depicts-GPU-performing-12-better-vs-RTX-3060-Ti-but-failing-to-offer-any-performance-price-advantage.722259.0.html

[Ross P]

The mini-itx form factors of 4060 ti offer some interest at least. As much as I would like to put a 3070 into my Aorus eGPU it won't fit and the power consumption would be a concern. However both Gainward and Palit make mini-ITX versions of the 4060 ti which should work a treat. Very niche use-case I must admit.

[S.Yu]

The infamy of the desktop 4060 Ti got me curious about how the mobile 4070 fares in a similar comparison, since they're fundamentally the same chip...and it's just as bad.
Compared to the 3070 Ti, the 4070 is overall heavily cut down, from memory bandwidth to pipelines to even raytracing and tensor cores. Without accounting for architectural differences, AD106 is basically only 75% the chip GA104 was. The only improvement is a ~50% increase in clocks, as a result it's hardly any faster (~10% overall, and ironically, 0.75x1.5=1.125), questionable if any more efficient on a high load(same max TGP given the performance noted), and understandably they refuse to disclose transistor count, since it's no doubt a heavy regression. And they ask for the same if not a higher price.
At the same width, TSMC 4N should be massively more efficient than Samsung 8N, about 2 full nodes' advantage, yet they use nearly the whole of the process advantage to cut down the chip to boost their own margins, instead of actually bringing improvement to the product, releasing this feeble upgrade 2 years after 3070 Ti. The corresponding YoY gain is approaching Intel's 14nm+++++, an industry laugh forever cemented in history.
They're milking the blood out of this cow, purely out of choice, which is to say ill will. Evil. Ngreedia is truly evil.

[Neenyah]

They're milking the blood out of this cow, purely out of choice, which is to say ill will. Evil. Ngreedia is truly evil.

While I agree with you and your comment there is one thing that has to be mentioned - Nvidia is fully in AI now, they don't really care much for regular consumers and making peanuts (even in tens of millions) when compared to insane amounts of money they generate in AI. The whole GeForce is becoming stagnant simply because their main focus is elsewhere these days, yet there is no true competition as long as they fully own CUDA (I highly doubt they will ever sell licence/rights/patent to it even if they sell the whole GeForce business). If you do professional 3D work and animation you simply need CUDA support. Gaming can be done with any brand though and Intel's ARC is becoming really interesting these days as they are very competitive with pricing.

[NikoB]

- Nvidia is fully in AI now

This "boom" for fools, already ridiculed by a bunch of experts, will soon pass, as will the attempt to create an autopilot, which turned into a complete fiasco, despite the fact that more than $ 100 billion was spent on this money of investors for nothing.

And then only sadness will remain, when there will be nothing to sell and no one at prices that are inflated at times.

We'll have to come up with a new "boom". They are already working on it...)))

[Neenyah]

Boom or not, that's irrelevant really. The real boom is on Nvidia's financial side and that's all they care about, GeForce revenue can't come anywhere near those numbers so they don't care much about it currently:

May 24, 2023

• Quarterly revenue of $7.19 billion, up 19% from previous quarter
• Record Data Center revenue of $4.28 billion
• Second quarter fiscal 2024 revenue outlook of $11.00 billion

""The computer industry is going through two simultaneous transitions — accelerated computing and generative AI," said Jensen Huang, founder and CEO of NVIDIA.

"A trillion dollars of installed global data center infrastructure will transition from general purpose to accelerated computing as companies race to apply generative AI into every product, service and business process.

"Our entire data center family of products — H100, Grace CPU, Grace Hopper Superchip, NVLink, Quantum 400 InfiniBand and BlueField-3 DPU — is in production. We are significantly increasing our supply to meet surging demand for them," he said."

nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2024

[S.Yu]

Boom or not, that's irrelevant really. The real boom is on Nvidia's financial side and that's all they care about, GeForce revenue can't come anywhere near those numbers so they don't care much about it currently:

May 24, 2023

• Quarterly revenue of $7.19 billion, up 19% from previous quarter
• Record Data Center revenue of $4.28 billion
• Second quarter fiscal 2024 revenue outlook of $11.00 billion

""The computer industry is going through two simultaneous transitions — accelerated computing and generative AI," said Jensen Huang, founder and CEO of NVIDIA.

"A trillion dollars of installed global data center infrastructure will transition from general purpose to accelerated computing as companies race to apply generative AI into every product, service and business process.

"Our entire data center family of products — H100, Grace CPU, Grace Hopper Superchip, NVLink, Quantum 400 InfiniBand and BlueField-3 DPU — is in production. We are significantly increasing our supply to meet surging demand for them," he said."

nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2024

That's unfortunately true. Cryptomining or ML it's businesses competing with retail for production capacity, and retail always loses. Now the question is whether AMD can up their game because whether Intel remains committed to discrete GPUs for consumers is in doubt, and they don't have a mobile line anyway.

[NikoB]

When short-sighted businessmen (as earlier at Intel) don't care about anything on the horizon, they run the risk of falling into the dust of oblivion.

If there is capitalism in the West, why do private companies, like Intel, save taxpayers' money, and not go bankrupt? Like the filthy banksters who stole trillions? As well as f*** officials and politicians? They must be bankrupt both in business and in service and in politics.

In the West, according to statistics, more than 50% of companies are "zombies", they do not receive real profit and are unprofitable. They are supported by taxpayers' money, although they are private through various indirect schemes.

If stop the money's printing press - the collapse of the system. If don't stop (as grandfather Biden and Co. are now getting out before the end of 2024), it's still a slightly delayed collapse. But politicians (like most of the population, businesses) prefer to shift their problems onto the shoulders of future generations. "After us, at least a flood" - that's their motto and creed of life.

[NikoB]

fucked Google Translate with fake "AI" )))), correct version:

When short-sighted businessmen (as earlier at Intel) don't care about anything on the horizon, they run the risk of falling into the dust of oblivion.

If there is capitalism in the West, why do private companies, like Intel, saved by taxpayers' money, and not go bankrupt? Like the filthy banksters who stole trillions? As well as f*** officials and politicians? They must be bankrupt in business and in service and in politics and in society.

In the West, according to economist's investigations more than 50% of companies are "zombies", they do not receive real profit and are unprofitable. They are kept bytaxpayers' money through various indirect schemes (including the stupid printing of money for loans that dilute the purchasing power of people who honestly earned their savings and lose them  because bank deposit rates are ALWAYS lower than real inflation, so , money is simply stolen from investors through a real negative interest rate on deposits, as well as on annual dividends on shares), although they are private companies and are obliged to get out on their own or go bankrupt ruthlessly (which will naturally lead to severe punishment for those who allowed this), as well as required under capitalism.

If stop the printing press of FRS - the collapse of the system. If don't stop (as grandfather Biden and Co. are now getting out before the end of 2024), it's still a slightly delayed collapse. But politicians (like most of the population, businesses) prefer to shift their problems onto the shoulders of future generations. "After us, at least a flood" - that's their motto and creed of life.

[Ry]

The mini-itx form factors of 4060 ti offer some interest at least.

Absolutely. This is really it's main strength. I can take the 3060 12GB out of a MSI Trident 3 and drop in a 4060Ti for signficant gains and lower temps/noise profile depending on the card.

I would hope the 16GB versions come with faster memory to ofset the lack of memory bandwidth the 4060Ti suffers from but it doesn't appear that's the case, making a $100 extra asking price for the 16GB card laughable. If it came with the new Samsung 24GBps GDDR6 that would be good.. Maybe they'll do a 4060 Super with GDDR6X.

[SPH]

Boom or not, that's irrelevant really. The real boom is on Nvidia's financial side and that's all they care about, GeForce revenue can't come anywhere near those numbers so they don't care much about it currently:

May 24, 2023

• Quarterly revenue of $7.19 billion, up 19% from previous quarter
• Record Data Center revenue of $4.28 billion
• Second quarter fiscal 2024 revenue outlook of $11.00 billion

""The computer industry is going through two simultaneous transitions — accelerated computing and generative AI," said Jensen Huang, founder and CEO of NVIDIA.

"A trillion dollars of installed global data center infrastructure will transition from general purpose to accelerated computing as companies race to apply generative AI into every product, service and business process.

"Our entire data center family of products — H100, Grace CPU, Grace Hopper Superchip, NVLink, Quantum 400 InfiniBand and BlueField-3 DPU — is in production. We are significantly increasing our supply to meet surging demand for them," he said."

nvidianews.nvidia.com/news/nvidia-announces-financial-results-for-first-quarter-fiscal-2024

Any idea how much they're making from just RTX 4090 and the Tegra SoC used in Switch? I guess not much for RTX 4090 as it's a low volume / higher margin product but the Switch has sold 100+ million, so I think they've made quite a bit from that even if it's a low margin product.

[Neenyah]

Any idea how much they're making from just RTX 4090 and the Tegra SoC used in Switch? I guess not much for RTX 4090 as it's a low volume / higher margin product but the Switch has sold 100+ million, so I think they've made quite a bit from that even if it's a low margin product.

I honestly wonder the same but can't find any kind of such full-breakdown on their site, they just divide different fields (Data center/AI, gaming etc.) into different categories and that's it. Maybe they will release that info in about 7-9 months, as a 2023 recap 🤔