NVIDIA Computex 2025 Keynote Address Liveblog

[Onasi]

It's very easy to look at "haha, look at this guy. he runs defense for a company I don't like and LMAO he plays Genshin, opinion discarded (heard that one before lol)"

That’s an actual fair take though, Genshin has the worst waifus. Real degenerates play NIKKE. Or Azur Lane, I suppose.

It is a bit personal when a company is lying while making the whole industry put along with the brakes on and not moving the needle. DLSS is garbage. Fake frames are fake frames. Fake frames create latency. It is all garbage. Maybe in a decade we will get there but honestly we aren't close yet. Just like Ray Tracing. It was thrown down our throats and it still end of the day isn't that relevant. Give me a good game. Ray tracing or not.

There’s no need to be reductive. DLSS, FrameGen, RT - whatever, they are all tools in the box, just like any other technique. The issue is developers use them as a crutch instead of enhancing an already well running and decently looking game. DLSS is a “we don’t need to optimize” button. RT is a “we don’t need to bake light and shadows, who cares it runs worse” button. It’s all a shitshow. I am yet to see a title besides CP2077 with its RT lighting that would look convincingly better than the old raster techniques to justify the performance loss. Hell, the new idTech 8 in Doom TDA with its RT requirement looks no better than Eternal while running at a third of the framerate. It’s just developers being monkeys giving shit to fling at the wall. Nvidia’s only fault in this regard is giving them access to said shot in the first place, I guess.

 

[LittleBro]

Sooner or later, Nvidia is going to move HQ to Taiwan. Why? Because of $$$.

 

[Icon Charlie]

It looks like Nvidia has gotten a few in the Youtube community a bit pissed.

Heres one from Paul Hardware.

Another is Daniel Owen.

Cherry picking reviewers... Yea...

 

[Rightness_1]

And I thought Bitcoin was a waste of electricity... A.I. said "hold my beer..."

 

[Dr. Dro]

It is a bit personal when a company is lying while making the whole industry put along with the brakes on and not moving the needle. DLSS is garbage. Fake frames are fake frames. Fake frames create latency. It is all garbage. Maybe in a decade we will get there but honestly we aren't close yet. Just like Ray Tracing. It was thrown down our throats and it still end of the day isn't that relevant. Give me a good game. Ray tracing or not.

In the mean time Gamers Nexus has shown that Nvidia is going to play vindictive games if you don't play by their rules. Which could violate US laws when it comes to how people are to disclose if they have been given a product and if it influenced their review etc. Which is why reviewers will state I was given this product for review but all opinions are my own. Which isn't the case with how Nvidia wants it to be.

It isn't a review if Nvidia has any input at all. If they want input they need to make all the reviews sponsored videos.

So far now.... Another acronym.

Narcissistic
Vindictive
Irrational
Disrespectful
Impudent
Assholes

Haha, I said it was juvenile, not necessarily wrong.

I've recently been under a bit of a spotlight for calling a programming practice that AMD employs "sewage", it turns out that many, many things don't exactly smell of roses here at the green camp, either. It'd be pretty foolish to think any of the corpos are our friend, but as @Onasi brought up, they are dangerously close to considering their GeForce business a liability. We're being treated accordingly, and the press must feel like a particularly nasty itch now that they started to lose ground due to high prices and recent software woes. Well, that's Nv getting some well deserved crabs for fooling around

Unfortunately, throwing $50,000 cash prizes at feature marketing campaigns can be quite effective... lots will forgive them on that alone. What a cheap fix!

 

[Caring1]

"NVIDIA announces Issac Groot humanoid robot foundation model"

But

 

[mechtech]

For those who have a hard time wrapping their head around 220MW, that's about 295,024.8 HP or in other terms getting close to half of a 500MW nuke plant or enough to roughly power 110,000 to 220,000 homes.

Crazy stuff right there.

 

[JustBenching]

So this gives them the right and license to be childish jerks? NVidia does something unacceptable, they simply refuse and tell them what they're doing is not ethical. The press doesn't need to throw the tantrums about it. It's called professional civility, decorum and maturity. The press has an obligation to exercise those things.

Well, it's called entitlement. These channels think that nvidia has the obligation to put food on their table. The issue here isn't that they can't review the product, they can, they just have to go buy it off the shelf like the rest of us. But no, they want special treatment so they can release day 1 review, else theyll take a hit to their income. As if that is nvidias problem.

EG1. Before the usual lurkers start with the "nvidia defense", I think that putting FG "performance" on graphs is at BEST misleading, it should probably be illegal, but that's a different issue.

Looking for consumer-oriented ethics in billionaire global corporations is like looking for a needle in a haystack, except that you actually have a chance to come across the needle instead of just vacuous corporate statements. Thankfully we have people like Rossmann, the greatest technician that's ever lived, GN, etc. to stand against them and, at the very least, put the brakes on it.

TBF it's not even the megacorps. Even you me or any small business, if spending marketing money (free samples etc.) doesn't get you back the money you spent for that marketing, then you'd stop spending it, right? You are under no obligation to continue sending free samples to people that don't actually help you increase your sales, that would be a stupid business decision.

 

[lexluthermiester]

Well, it's called entitlement.

True!

The issue here isn't that they can't review the product, they can, they just have to go buy it off the shelf like the rest of us.

Exactly! That's what the press did BITD. They waited for the item to hit shelves, bought their own and then reviewed it.

Companies should NEVER have any input or say in how a product is reviewed.

 

[Hecate91]

Well, it's called entitlement.

It isn't entitlement, please learn how the review process works.
A review takes weeks of testing, a reviewer can't simply buy a card on launch day and have a thorough set of benchmark charts ready on the same day.

The issue here isn't that they can't review the product, they can, they just have to go buy it off the shelf like the rest of us.

It isn't about having to buy it themselves either, but when a company refuses to send out a product for review it doesn't look good for the company.

TBF it's not even the megacorps. Even you me or any small business, if spending marketing money (free samples etc.) doesn't get you back the money you spent for that marketing, then you'd stop spending it, right? You are under no obligation to continue sending free samples to people that don't actually help you increase your sales, that would be a stupid business decision.

Or maybe the company can just make a product good enough to be worthy of sending out review samples, I would expect that from a multi-trillion dollar company.
When a company is so afraid of reviewers possibly criticizing a product, then maybe the product sucks, making sh*tty products not even worth a review is a stupid business decision.

 

[JustBenching]

It isn't entitlement, please learn how the review process works.
A review takes weeks of testing, a reviewer can't simply buy a card on launch day and have a review ready on the same day.

They don't have to release the review the same day.

Or maybe the company can just make a product good enough to be worthy of sending out review samples, I would expect that from a multi-trillion dollar company.
When a company is so afraid of those mean reviewers possibly criticizing a product, then maybe the product sucks, making sh*tty products not even worth a review is a stupid business decision.

Sure, the product sucks, whatever, doesn't even matter. No company should or is forced to send out review samples to people they don't want to work with. It's absurd to even suggest such a thing.

 

[Dr. Dro]

Oh. Ah. Oopsie. You have to wonder if they were judged too harshly here.

Hopefully this means 9060 XT 8G at $299.

It isn't entitlement, please learn how the review process works.
A review takes weeks of testing, a reviewer can't simply buy a card on launch day and have a thorough set of benchmark charts ready on the same day.

It isn't about having to buy it themselves either, but when a company refuses to send out a product for review it doesn't look good for the company.

Or maybe the company can just make a product good enough to be worthy of sending out review samples, I would expect that from a multi-trillion dollar company.
When a company is so afraid of reviewers possibly criticizing a product, then maybe the product sucks, making sh*tty products not even worth a review is a stupid business decision.

The truth and nothing but the truth. You know, all things considered, we actually see eye to eye on far more things than previously thought, if we can put that "but you like the opposite brand" feud aside.

 

[Rabit]

Hopefully this means 9060 XT 8G at $299.

RX 9060 XT have PCIe 5.0×16 ~ 64GB/s this can make 8GB edition faster than Nvidia RTX 5060 TI 8GB PCIe 5.0×8 ~ 32GB/s

 

[Dr. Dro]

RX 9060 XT have PCIe 5.0×16 ~ 64GB/s this can make 8GB edition faster than Nvidia RTX 5060 TI 8GB PCIe 5.0×8 ~ 32GB/s

Fortunately, W1zzard tested this scenario, and the conclusion is that using the graphics card at x8 4.0 is not going to cause any significant performance degradation, a very minor one (~2-3%) occurs at x8 3.0, so dropping two generations, at effectively just about 25% of the bandwidth that a 5.0 slot would offer with the same lane count.

NVIDIA GeForce RTX 5060 Ti PCI-Express x8 Scaling

The new NVIDIA GeForce RTX 5060 Ti uses the modern PCI-Express 5.0 bus interface, but supports only an 8-lane configuration. This is just half the bandwidth of the higher-end models, which run at x16. Are you in trouble with an older system, which only supports running at PCIe 4.0?

www.techpowerup.com

It has little bearing even going all out with the 5090:

NVIDIA GeForce RTX 5090 PCI-Express Scaling

The new NVIDIA GeForce RTX 5090 is the first high-end graphics card that makes use of a PCI-Express 5.0 bus interface. Are you in trouble when trying to run it on PCIe 4.0? What about x8, like when an SSD is using up some bandwidth? We've also tested various PCI-Express 3.0, 2.0 and 1.1 configs...

www.techpowerup.com

I understand you may be meaning WDDM shared memory spill (when 8 GB is exceeded), but regardless of the lane count available experience will take a hit and this is a scenario that should be avoided at all costs.

At best, this could make the 9060 XT behave a little better on a 3.0 system, but RDNA 4 is simultaneously in the worst position for this as AMD dropped support for legacy systems and requires an UEFI compliant BIOS for the RX 9000 series, so people looking to put these on some earlier Ivy Bridge PCs might not have luck. Should be great for the Intel X99 platform, though, and that's my intended use case. I plan on buying one sometime to install on my server to use for video transcoding workloads and if possible, some level of virtualized graphics under Linux. The GPU I have on it right now is not exactly capable of doing any of that.

 

[Lianna]

Fortunately, W1zzard tested this scenario, and the conclusion is that using the graphics card at x8 4.0 is not going to cause any significant performance degradation, a very minor one (~2-3%) occurs at x8 3.0, so dropping two generations, at effectively just about 25% of the bandwidth that a 5.0 slot would offer with the same lane count.

NVIDIA GeForce RTX 5060 Ti PCI-Express x8 Scaling

The new NVIDIA GeForce RTX 5060 Ti uses the modern PCI-Express 5.0 bus interface, but supports only an 8-lane configuration. This is just half the bandwidth of the higher-end models, which run at x16. Are you in trouble with an older system, which only supports running at PCIe 4.0?

www.techpowerup.com

It has little bearing even going all out with the 5090:

NVIDIA GeForce RTX 5090 PCI-Express Scaling

The new NVIDIA GeForce RTX 5090 is the first high-end graphics card that makes use of a PCI-Express 5.0 bus interface. Are you in trouble when trying to run it on PCIe 4.0? What about x8, like when an SSD is using up some bandwidth? We've also tested various PCI-Express 3.0, 2.0 and 1.1 configs...

www.techpowerup.com

I understand you may be meaning WDDM shared memory spill (when 8 GB is exceeded), but regardless of the lane count available experience will take a hit and this is a scenario that should be avoided at all costs.
View attachment 400534

No, W1zzard did not test THIS scenario (yet). 5060 Ti was tested only with 16 GB version, 5090 obviously with 32 GB.
This is not the first time you ignore parts of questions that specifically ask for 8 GB cards and/or selected games.
There are tests showing 8 GB cards exhibit much higher degradation in a lot of real life situations (where 16 GB cards have no problem running at good fps).
Degradation may not be present in all titles, so maybe not a big hit to "n-game average", heavily depending on game selection.
In some cases it may not show (much) in average fps, but 1% lows (stutter) may make game unplayable or at least unpleasant.
In some cases it shows in Nvidia recommended settings, as FG/RT settings push VRAM requirements.
Here it may be compunded by 5060 Ti 16GB to 8 GB degradation being (for whatever driver reason?) visibly worse than on 4060 Ti.

So no, repeating "a very minor [performance degradation]" does not make it true.
I'm still waiting for W1zzard's promised 5060 Ti 8 GB PCIe scaling review.

 

[Dr. Dro]

No, W1zzard did not test THIS scenario (yet). 5060 Ti was tested only with 16 GB version, 5090 obviously with 32 GB.
This is not the first time you ignore parts of questions that specifically ask for 8 GB cards and/or selected games.
There are tests showing 8 GB cards exhibit much higher degradation in a lot of real life situations (where 16 GB cards have no problem running at good fps).
Degradation may not be present in all titles, so maybe not a big hit to "n-game average", heavily depending on game selection.
In some cases it may not show (much) in average fps, but 1% lows (stutter) may make game unplayable or at least unpleasant.
In some cases it shows in Nvidia recommended settings, as FG/RT settings push VRAM requirements.
Here it may be compunded by 5060 Ti 16GB to 8 GB degradation being (for whatever driver reason?) visibly worse than on 4060 Ti.

So no, repeating "a very minor [performance degradation]" does not make it true.
I'm still waiting for W1zzard's promised 5060 Ti 8 GB PCIe scaling review.

Yeah, I figured it out after I wrote the post and added that in later, but the truth is that it could have the RTX 5090's monstrous throughput (cache, high data rates, large bus, etc.) - venture past the local framebuffer and your experience is 100% gonna take a hit. 8, 16, 32 GB - if you have to swap the data set, you are wasting precious performance cycles there.

 

[Rightness_1]

Oh. Ah. Oopsie. You have to wonder if they were judged too harshly here.

View attachment 400533

Hopefully this means 9060 XT 8G at $299.



The truth and nothing but the truth. You know, all things considered, we actually see eye to eye on far more things than previously thought, if we can put that "but you like the opposite brand" feud aside.

The biggest issue is the VRAM. 8GB is not enough VRAM in 2025. Watch their review.

 

[lexluthermiester]

The biggest issue is the VRAM. 8GB is not enough VRAM in 2025. Watch their review.

Please don't start that silliness.