Potential RTX 5090 and RTX 5080 Pricing in China Leaks

[evernessince]

I'll be quite frank...

New games aren't that good to invest into it either way... TW4 maybe? But it is years away. We all know GTA6 will need a 6000 series to run good as it is always with rockstar and I expect it to flop in certain ways, there are certain signs of that.

Yeah, spending $2,600 on just the GPU to game is kind of nuts IMO. If you work on it maybe but for games your rig will be approaching $6,000.

 

[Ferrum Master]

Yeah, spending $2,600 on just the GPU to game is kind of nuts IMO. If you work on it maybe but for games your rig will be approaching $6,000.

Nvidia kinda hallucinates... like the AI does... if the toy will start to be more expensive than a car... and this already costs like two used Volvos... people will turn into other toys... time for winter beater for certain places never ceased to exist, you know what I mean. You cannot pull the price endlessly out of your arse.

And don't give me all this CUDA crap, like anyone of us does it like next thing besides posting here, it is simply not worth it for what it is, let's be real.

 

[Bobaganoosh]

Should we not be comparing the 5080 price to the 4080 Super? The 4080 Super fully replaced the original 4080 and was price-adjusted because they overpriced the 4080 to begin with. Being they completely EOL'd the 4080, it doesn't seem fair to compare to that. Also, specifically referring to MSRP, not scalper pricing.

 

[ymdhis]

Here I am gaming on an iGPU and laughing so hard.

 

[Steevo]

Excellent, let the scalping begin

 

[Lew Zealand]

Here I am gaming on an iGPU and laughing so hard.

LOL I've been gaming on an iGPU for a few days and it's been a blast. Horsepower is nice but matching great games to your system works at every price point.

 

[jesdals]

So China scalper price will be 35000 RMB....

 

[Dr. Dro]

Should we not be comparing the 5080 price to the 4080 Super? The 4080 Super fully replaced the original 4080 and was price-adjusted because they overpriced the 4080 to begin with. Being they completely EOL'd the 4080, it doesn't seem fair to compare to that. Also, specifically referring to MSRP, not scalper pricing.

No! Where is my fake outrage and my pro AMD rhetoric! I demand $599 flagships and the ATi glory days!

LOL I've been gaming on an iGPU for a few days and it's been a blast. Horsepower is nice but matching great games to your system works at every price point.

Stop making sense, you! Where is your sense of shame, young man!

 

[Onasi]

LOL I've been gaming on an iGPU for a few days and it's been a blast. Horsepower is nice but matching great games to your system works at every price point.

Very based, honestly. And quite in vein of actual PC enthusiasts of old. What the fuck happened? We used to buy cheaper hardware and tinker with it for more performance, we tweaked the shit out of games for more frames at comparable image quality, hardcore FPS enthusiasts used console commands to run Q3 with flat textures to frag more efficiently on their old 800x600 CRTs. We had tons of genres that weren’t HW intensive, yet QUINTESSENTIALY PC. And some that WERE intensive AND PC centric. Consoles used to be memed on. Now it’s all about “I bought this 9999$ GPU and a scalped 800$ CPU, turned up PT to Ultra at 16K in the latest AAA console slop and shit doesn’t work good. Fucking NVidia, PC gaming is dead”. I tell ya, we didn’t gatekeep hard enough.

 

[Klemc]

Thinkink about this growing prices, their drivers won't magically become better (that they need) so all this money is waste on the consumer side, totally !!!!!!!

 

[ymdhis]

Very based, honestly. And quite in vein of actual PC enthusiasts of old. What the fuck happened? We used to buy cheaper hardware and tinker with it for more performance, we tweaked the shit out of games for more frames at comparable image quality, hardcore FPS enthusiasts used console commands to run Q3 with flat textures to frag more efficiently on their old 800x600 CRTs. We had tons of genres that weren’t HW intensive, yet QUINTESSENTIALY PC. And some that WERE intensive AND PC centric. Consoles used to be memed on. Now it’s all about “I bought this 9999$ GPU and a scalped 800$ CPU, turned up PT to Ultra at 16K in the latest AAA console slop and shit doesn’t work good. Fucking NVidia, PC gaming is dead”. I tell ya, we didn’t gatekeep hard enough.

What happened is that GPUs grew to a trillion dollar market, and this brings with it much stronger marketing, telling you that you need to buy the latest greatest card. Why do you think TPU makes a new article about every rumor or stray photograph of the 5090 too?

 

[Prima.Vera]

Me, reading those kind of news, looking at my RTX 3080:

 

[ir_cow]

Titan Z had 8.8 times the FP64 performance of the 980 Ti and whopped the next gen 1080 Ti in that regard as well. Not a good comparison, the 5090 will not have a 1:3 FP64 ratio.

Sure it had much better FP64, but they sold it as a gaming card. Also you could get Radeon card for FP64 if you really needed it.

 

[evernessince]

Sure it had much better FP64, but they sold it as a gaming card. Also you could get Radeon card for FP64 if you really needed it.

You might want to check GTC 2014 videos, the only time gaming is mentioned is near the start:

The rest of the time is spent talking about it's uses outside of gaming and they have an entire 2nd part on it just for rendering (pixar, ect).

 

[R-T-B]

Until the Feds step in but there is a new Admin coming.

No one honestly thinks the new admin is going to lower tariffs though.

 

[Verpal]

It's not about tariffs, but compute power exceeding what is allowed to be sold in China.

Considering TPP limit of sanction is 4800 for FP8, if NVIDIA is to maintain a semblance of core count, they will have to cut down tensor core drastically, and maybe drop the official TGP quite a bit too.

That being said, current DLSS doesn't seems to completely saturate tensor core, so maybe NVIDIA can get away with that, and performance will be fine for gaming?

 

[R-T-B]

Is that without the 60% alleged tariff?

They aren't in effect yet so it would have to be.

 

[Garrus]

LOL I've been gaming on an iGPU for a few days and it's been a blast. Horsepower is nice but matching great games to your system works at every price point.

I hope we see something like socketed Strix Halo in the future. Would be nice to finally get 256 bit memory bus width on the desktop. Maybe even if they have to sell it as a MB+CPU+RAM kit.

Me, with reading those kind of news, looking at my RTX 3080:
View attachment 377519

I used an RTX 3080 for 4 years straight and only upgraded to the 4070 Ti to get 12GB of ram to stop the 10GB issue. (bunch of games need more than 10GB now especially at 4k even with upscaling enabled)

 

[igormp]

$2,600 is pretty much around worst case predictions. It's a straight $1,000 price increase. It may actually stay in stock assuming it isn't soaked up for AI. That price is high enough to where I can see even many ultra-enthusiasts question it. Heck I'm questioning it and I do use my 4090 for both work and AI.

I had considered it for a short time, but getting two of those is not worth it at all for me, given that one wouldn't make much sense vram-wise.
For the supposed price of a single 5090 I can get over a thousand hours with a GH200 in the cloud, which is way faster and has way more vram.

And don't give me all this CUDA crap, like anyone of us does it like next thing besides posting here, it is simply not worth it for what it is, let's be real.

Hi, I only use GPUs for CUDA and shitposting in forums, and don't game at all
But still, not worth it for me. I still can see it being "cheap" for many others, tho.

I hope we see something like socketed Strix Halo

Sadly I doubt this will ever happen.

 

[ir_cow]

You might want to check GTC 2014 videos, the only time gaming is mentioned is near the start:

The rest of the time is spent talking about it's uses outside of gaming and they have an entire 2nd part on it just for rendering (pixar, ect).

Yeah the first 50 seconds is him saying it's a gaming card and the rest is "oops" we didn't know people would use the FP64 part. Its still two 780s in SLI. cheaper to get 2x Titans for $2,000 at the time with more performance. People bought this for gaming and not compute. It just didn't make sense to overpay for two Titans on a single PCB.

 

[Paganstomp]

I switched back to an old hobby after 45 years. Now...

 

[agent_x007]

Its still two 780s in SLI.

It's two Titan Blacks in single PCB card (1:3 FP64 enabled). Clock adjusted to make TDP... reasonable (for 2014 ).

 

[evernessince]

Yeah the first 50 seconds is him saying it's a gaming card and the rest is "oops" we didn't know people would use the FP64 part. Its still two 780s in SLI. cheaper to get 2x Titans for $2,000 at the time with more performance. People bought this for gaming and not compute. It just didn't make sense to overpay for two Titans on a single PCB.

Take a look at the next part, it's entirely dedicated to Titan Z non-gaming related uses.

 

[ModEl4]

The original plan regarding Ada was $600/$900/$1200/$1600 for RTX4070/4080 12GB/4080 16GB /4090 and this kinda shows Nvidia's way of thinking and intentions. Because the raster performance jump for Blackwell will be underwhelming and based on what I expect regarding performance for Blackwell lineup I thought that the pricing will be $600/$900/$1100/$1700 for RTX 5070/5070Ti/5080/5090 but now that I think about it again it should be probably $600/$900/$1200/$1800.
RTX 5070 should stay at the same price since the raster performance increase should be 10% or something around this level and also there is competition from Navi 48, RTX 5070Ti should be finally at $900 (original intention of Nvidia was to name RTX 4070Ti , RTX 4080 12GB and price it at $900, this time 5070Ti will have the same memory size as it 5080 and also it will be based on the same chip instead of a lower tier like it was in 4080/4070Ti case, so I think Nvidia will try again to force the $900 price tag).
RTX 5080 should have $300 difference from RTX 5070Ti, I don't think the gap will be smaller than the gap between RTX 5070 ($600) / RTX 5070Ti ($900) also RTX5080 is full GB203 so Nvidia will try to artificially lower the demand for RTX5080 since the yields potentially won't be enough for the demand of a $1100 RTX 5080 and most importantly there is also the mobile RTX5090 which should be full GB203 also and finally regarding RTX 5090 since demand for RTX 4090 was not affected after the launch of 4080 Super at $1000 it should maintain the $600 gap from RTX 5080, so it should be $1800 (also 2X the price of RTX 5070Ti, like it was in RTX4090 $1600/4070Ti $800 case).
Regarding high-end the cost for RTX 5090 will be higher for Nvidia in relation with RTX 4090, the process is similar with mild difference, AD102 4N (5nm based) 609 mm² vs GB202 rumored at 744mm² in a process based on 4nm, the memory bus also is 512bit vs 384bit with 32GB vs 24GB memory and the ICs are GDDR7 based instead of GDDR6X, the interface is gen5 instead of gen4, the power consumption is higher and these changes affect also other factors potentially like PCB layers and other components so the $200 increase seems logical to me to happen in this case, regarding the others prices are a major letdown also (except maybe RTX 5070 which is just meh)

 

[Nostras]

Me, with reading those kind of news, looking at my RTX 3080:
View attachment 377519

10GB is squarely in the danger zone considering current VRAM usage trajectory.