Discussion Nvidia Blackwell in Q1-2025

[Heartbreaker]

Just try scrolling a website or dragging a window around. It's a huge difference. But I do think a mouse brings it out more. I have a 60hz display on my laptop and see the difference all the time.

A lot of games 20+ years ago were designed to only work properly at 60fps, and sometimes even 30fps. Force Unleashed was like that and it felt horrible.

I have an old 60Hz LCD next to my gaming monitor as a second monitor, and I can notice those differences 120 vs 60 moving around windows, but my gaming monitor also looks slightly better even 60Hz vs 60Hz on the old monitor. Moving a window around seems to be more noticeable than gaming with a controller.

But I don't notice any difference going from 120Hz to 165Hz anywhere.

Each step is diminishing returns.

The difference between 30 and 60 is hits like a hammer to the face.

The difference between 60 and 120 is more subtle, and even when detectable, seems fairly minor.

120 to 165 I really can't detect anywhere.

This all assumes even frame pacing. Even 120 with spikes feels horrible.

Is there a way to setup an in between refresh on a monitor?

I only have 60, 120, 144, 165. I'd like to try 90 Hz. It's a VRR monitor so it should be able to run at 90 Hz.

 

[SolidQ]

Some info about 5070/ti/5080

https://videocardz.com/newz/alleged-geforce-rtx-5080-3dmark-leak-15-faster-than-rtx-4080-super

 

[coercitiv]

Is there a way to setup an in between refresh on a monitor?

NVIDIA Support

nvidia.custhelp.com

 

[Thunder 57]

Some info about 5070/ti/5080

https://videocardz.com/newz/alleged-geforce-rtx-5080-3dmark-leak-15-faster-than-rtx-4080-super

The 5080 is getting crucified in the comments over on Videocardz. Can't say I blame them.

 

[techjunkie123]

Nuh uh. Don't twist Honorable Mr. Jensen's words. It's 4090 performance for $549. Because he said so.

Heck why call it a 5070??

Presenting the new 4090 12GB.

Available for only 550$.

 

[Josh128]

Are we possibly on the cusp of the ridiculous GPU price bubble bursting? I mean, 5080 might not be great, but 15% faster than a 4080 Super for $999 is not a terrible deal when looking at the ridiculous prices in the post COVID era. We all want AMD to save the day with 9070/XT, but the reality is its a not so small monolithic die produced on the ridiculously sought after TSMC 4nm process, which means they are definitely limited in what they can offer price wise. If you consider that Huang potentially has much more stroke as a customer due to its AI GPU explosion, he may even have negotiated better rates than AMD has. Or on the flip side, he might have paid more to reserve more capacity.

Either way, the bubble might be getting ready to pop.

 

[adroc_thurston]

all want AMD to save the day with 9070/XT, but the reality is its a not so small monolithic die produced on the ridiculously sought after TSMC 4nm process

It's sub 350mm^2, uses commodity mem and N4 isn't ridiculously sought after at all.
Not a single bleeding edge node has been wafer limited for 2 years.

 

[Heartbreaker]

Are we possibly on the cusp of the ridiculous GPU price bubble bursting? I mean, 5080 might not be great, but 15% faster than a 4080 Super for $999 is not a terrible deal when looking at the ridiculous prices in the post COVID era. We all want AMD to save the day with 9070/XT, but the reality is its a not so small monolithic die produced on the ridiculously sought after TSMC 4nm process, which means they are definitely limited in what they can offer price wise. If you consider that Huang potentially has much more stroke as a customer due to its AI GPU explosion, he may even have negotiated better rates than AMD has. Or on the flip side, he might have paid more to reserve more capacity.

Either way, the bubble might be getting ready to pop.

I'm not seeing the bubble part. That's usually some kind of huge demand.

 

[Josh128]

It's sub 350mm^2, uses commodity mem and N4 isn't ridiculously sought after at all.
Not a single bleeding edge node has been wafer limited for 2 years.

The AI "revolution" is less than 2 years old and showing no signs of slowing down yet.

Every single 3 trillion dollar market cap NV GPU, commercial and consumer, most of what AMD produces, and much of what Intel is currently producing uses 4nm/5nm fabs. AI companies, MS, Meta, Amazon, Tesla, and others buying NV commercial stuff by literally hundreds of thousands of pieces says otherwise to your claim that it isnt sought after, lol. What proof do you have for your claim that "N4 isnt ridiculously sought after at all"? TSMC publicly announced they were raising prices due to huge demand months ago, and recently posted record net income, exceeding same quarter last year by over 50%. Net income= high ASPs.

And RDNA4 if it turns out to be sub 350mm2 (doubt), that is exactly still the definition of "not so small".

 

[Win2012R2]

Are we possibly on the cusp of the ridiculous GPU price bubble bursting?

We are potentially on a cusp of whole AI bubble bursting, thanks to DeepSeek

 

[jpiniero]

The AI "revolution" is less than 2 years old and showing no signs of slowing down yet.

Most of that demand is solely because Wall Street loses it's mind over anything AI. They get bored, AI chip demand will dry up in an instant.

 

[PJVol]

I'd love to see some testing of how many people could detect higher frame rate difference when gaming only using a controller

I switched to 175hz OLED a couple days ago, from a 75hz VA (both 1440p), and I can say that if on the desktop the difference is clearly noticeable, in fast paced games, e.g. Doom Eternal, it's not so obvious.

 

[maddie]

The AI "revolution" is less than 2 years old and showing no signs of slowing down yet.

Every single 3 trillion dollar market cap NV GPU, commercial and consumer, most of what AMD produces, and much of what Intel is currently producing uses 4nm/5nm fabs. AI companies, MS, Meta, Amazon, Tesla, and others buying NV commercial stuff by literally hundreds of thousands of pieces says otherwise to your claim that it isnt sought after, lol. What proof do you have for your claim that "N4 isnt ridiculously sought after at all"? TSMC publicly announced they were raising prices due to huge demand months ago.

And RDNA4 if it turns out to be sub 350mm2 (doubt), that is exactly still the definition of "not so small".

AFAIK, packaging capacity is the bottleneck.

 

[maddie]

We are potentially on a cusp of whole AI bubble bursting, thanks to DeepSeek

I'm surprised at the lack of attention that's getting. Maybe the shock is still affecting involved parties.

A lot of CapEx plans probably blew up.

 

[Josh128]

4nm/5nm is probably the highest demand production line TSMC has ever produced.

Taiwan Semiconductor Manufacturing Company Limited TSM, better known as TSMC, saw its shares scale northward after posting record fourth-quarter net income, which topped analysts’ projections. The foundry behemoths’ fourth-quarter net profit soared 57% to $374.68 billion from a year earlier, driven by a surge in demand for artificial intelligence (AI) chipsets. - Jan 23, 2025 Zacks

TSMC to Raise Wafer Prices by 10% in 2025, Customers Seemingly Agree

Taiwanese semiconductor giant TSMC is reportedly planning to increase its wafer prices by up to 10% in 2025, according to a Morgan Stanley note cited by investor Eric Jhonsa. The move comes as demand for cutting-edge processors in smartphones, PCs, AI accelerators, and HPC continues to surge...

www.techpowerup.com

TSMC's Arizona Fab 21 mass produces 4nm chips at a higher price than Taiwan

But customers are willing to pay extra, says TSMC.

www.tomshardware.com

 

[adroc_thurston]

4nm/5nm is probably the highest demand production line TSMC has ever produced.

That doesn't mean it's a wafer-limited node at all.

 

[DAPUNISHER]

I mean, 5080 might not be great, but 15% faster than a 4080 Super for $999 is not a terrible deal when looking at the ridiculous prices in the post COVID era.

My gut brain says that's hopium. I'll be impressed if there is a single AIB model selling for that price at launch. The reference will probably be low stock and sellout immediately.

 

[Win2012R2]

I'm surprised at the lack of attention that's getting

It's getting it right now BIG TIME.

 

[Josh128]

That doesn't mean it's a wafer-limited node at all.

II didnt say that it was, I said AMD would be limited in what they can offer price wise. But being that you brought it up, its very likely that it is wafer limited. If customers are willing to pay more for 4nm from the Arizona fab, its not because there is an abundance of unused capacity in Taiwan.

 

[Josh128]

It's getting it right now BIG TIME.

https://twitter.com/x/status/1883149286088819029

https://twitter.com/x/status/1882525126039916819

AMD trying to get in on the action:

https://x.com/AMD/status/1882851449991737473

https://x.com/HotAisle/status/1883216290409771422

 

[maddie]

It's getting it right now BIG TIME.

Sorry, I meant from the big public players.

 

[poke01]

As long as Meta suffers that’s all I want.

 

[Heartbreaker]

https://twitter.com/x/status/1883149286088819029

https://twitter.com/x/status/1882525126039916819

AMD trying to get in on the action:

https://x.com/AMD/status/1882851449991737473

https://x.com/HotAisle/status/1883216290409771422

Do we have something more substantive than some hype tweets?

 

[maddie]

Do we have something more substantive than some hype tweets?

Check out the site yourself.

deepseek-ai/DeepSeek-R1 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

 

[branch_suggestion]

GB202 Dieshot

https://twitter.com/x/status/1883153126011892140

Interesting floorplan tweaks.
Not just required to fit the reticle, it seems a bit more drastic.