Kingpin Skipping GeForce RTX 50 Generation, Cites "Impossible" Logistics

[T0@st]

Vince "Kingpin" Lucido's latest "Let's Talk" video blog has revealed difficult circumstances—affecting the development of a new-generation "KP Card." The extreme overclocker has announced a planned move back to North America (prioritizing his family) around mid-2025—he has lived in Taiwan for fifteen years—along with a sad project cancellation. Kingpin stated that: "there's no 'my card' this generation. It can't come for Gen 5. It's impossible. There are just too many issues over here, and it would create a lot of problems if I jumped on board so quickly." Mid-last year, Kingpin partnered with PNY—this collaboration reached a "wishlist" phase following tests performed in 3D Mark, according to today's VLOG.

Timing is cited as a major factor in his decision-making—he did not want to sign a Taiwanese vendor's contract and commit significant man-hours during a trying life period. He reckons that he has options further down the line—potentially with US-based companies. His next Kingpin card design (skipping a generation) could be something "unbelievable" and he hinted about it being a surprise. On this subject he stated: "I can't really say much about that. Well, because it has to do with my previous contractual stuff. I love graphics cards. I love making graphics cards, and for sure, I don't think that's going to die. But I think we'll have to wait-still have to wait for another generation. And who knows? Maybe it won't be Team Green; maybe it'll be another color."




Skip to the 2-minute 11-second mark for "New Card Talk."

Kingpin's description states: "in this video, I give a final update on the possibility of a KINGPIN Graphics Card for this new generation of 5 series NVIDIA cards. I also talk about new CA regional availability of my KPx High Performance Thermal Grease and the future of this channel."

View at TechPowerUp Main Site | Source

 

[Daven]

I can really relate to this guy. I also don't want an RTX 50 series GPU. What are the odds?

 

[Onasi]

And absolutely nobody who isn’t into XOC would care. I mean, it’s a shame since his collabs with EVGA were nice cards, but realistically they don’t have much place in the market anyway.

 

[lexluthermiester]

I can really relate to this guy. I also don't want an RTX 50 series GPU. What are the odds?

The main problem most people have is the power/heat aspect. The projected performance is not going to disappoint.

 

[agent_x007]

Main issue :
(X)OC is dead, unless you can make NV to release higher TDP vBIOS for specific card.
If you lack NV "say" - they may simply not allow it (or wall it with "some additional requirements needs to be done..." for you to get it).
^Speculation warning.

 

[x4it3n]

The main problem most people have is the power/heat aspect. The projected performance is not going to disappoint.

In RAW performance the 5090 is going to be 30-40% faster than the 4090, but definitely not even close to 2x lol. You'll need AI for that.
Also all models below will be closer to 20-30% faster vs their RTX 40s counterparts when looking at the official benchmarks from Nvidia.

 

[lexluthermiester]

In RAW performance the 5090 is going to be 30-40% faster than the 4090

Ok, yeah, and those are very compelling numbers.

but definitely not even close to 2x lol

No one who has any credibility has said 2x! It's a meritless statement without some sort of miracle in IC development.

 

[x4it3n]

Ok, yeah, and those are very compelling numbers.

No one who has any credibility has said 2x! It's a meritless statement without some sort of miracle in IC development.

The problem is that people were expecting the same performance gap as 3090 Ti to 4090 but that increase only happened because Nvidia switched from Samsung 8nm to TSMC 4N !
Blackwell is a new architecture but we can see that CUDA Cores IPC is not getting much better anymore with every new generation, so the only way to get more performance is to increase the number of CUDA Cores & Core Frequencies too, but both increase the Wattage, Voltage and Heat too... RTX 40s and RTX 50s being on a pretty similar node we couldn't expect much! If anything GDDR7 and the 512-bit bus are the big changes here, and ~33% more CUDA Cores.
But the 4090 had ~52% more CUDA Cores and ~35% Higher Core Clocks than the 3090 Ti ! And also a lot more L2 Cache (72MB vs 6MB) to make up for the lack of Memory Bandwidth.

 

[lexluthermiester]

The problem is that people were expecting the same performance gap as 3090 Ti to 4090 but that increase only happened because Nvidia switched from Samsung 8nm to TSMC 4N !

That wasn't twice the performance, not even close. So whoever was expecting such is fool to themselves.

But the 4090 had ~52% more CUDA Cores and ~35% Higher Core Clocks than the 3090 Ti !

Observe;

From the 3090ti to the 4090 was at most a 48% jump. The RTX 5000 series may not be as big a jump, but it'll be within a reasonably close margin. It will NOT be 2x the performance.

 

[mkppo]

He wants to maybe do another colour? Go on then release a badass UDNA card

AMD better make it badass in the first place though. Many maybes

 

[x4it3n]

That wasn't twice the performance, not even close. So whoever was expecting such is fool to themselves.

Observe
View attachment 380910
From the 3090ti to the 4090 was at most a 48% jump. The RTX 5000 series may not be as big a jump, but it'll be within a reasonably close margin. It will NOT be 2x the performance.

The 3090 Ti has a full GA102 die whereas the 4090 is a cut-down of GB202. The 4090 is about 2x a 3090 (non Ti) at 4K with RT/PT but about 70-80% in Raster which is already a lot more than 30-40% of Blackwell.
We haven't had a full AD102 die (4090 Ti) but it could have done 2x in Raster imo. Also we probably won't have a full GB202 either since Nvidia are all about selling full die for AI chips only...

 

[lexluthermiester]

The 3090 Ti has a full GA102 die whereas the 4090 is a cut-down of GB202. The 4090 is about 2x a 3090 (non Ti) at 4K with RT/PT but about 70-80% in Raster which is already a lot more than 30-40% of Blackwell.

Did you actually look at the screen shot? Take a closer look. It's not twice the performance of a 3090. Why are you arguing this point?

 

[x4it3n]

Did you actually look at the screen shot? Take a closer look. It's not twice the performance of a 3090. Why are you arguing this point?

I never said it was 2x and even the 5090 will only be 2x with MFG (A.I.)
Also if you read my post again the 4090 is a cut-down of AD102, a full AD102 could have been 15% faster and much closer to 2x, we just never had one.

 

[Assimilator]

When was the last time he was actually relevant?

 

[lexluthermiester]

I didn't know he had stopped being relevant.

 

[N/A]

Did you actually look at the screen shot? Take a closer look. It's not twice the performance of a 3090. Why are you arguing this point?

The new GPU testing methodology shows 1.52 (1.67 with RT) and enabling the remaining 10% of the die is another 5% max.

 

[lexluthermiester]

The new GPU testing methodology shows 1.52 (1.67 with RT) and enabling the remaining 10% of the die is another 5% max.

Ok. Still, that's nowhere near twice.

 

[trsttte]

The main problem most people have is the power/heat aspect. The projected performance is not going to disappoint.

I think the main problem is the price increase, with the current expected numbers you're getting no perf/dollar increase, just paying more for more performance.

 

[lexluthermiester]

I think the main problem is the price increase

Except that it's not going to increase. It's either going to stay the same or drop. NVidia already confirmed that.

 

[mkppo]

Except that it's not going to increase. It's either going to stay the same or drop. NVidia already confirmed that.

The price is going to increase and nvidia confirmed that, maybe you're confusing 5090 with the 5080 which isn't getting one. So $1600 > $2000 would be a 25% increase for what i'm guessing will be 30% more performance. Meh at best, but it'll be the fastest GPU and nvidia can charge whatever so it is what it is.

Meanwhile the 4090 increased by $100 over 3090 and had ~55% more performance.

 

[x4it3n]

Ok. Still, that's nowhere near twice.

I never said the 4090 was twice as powerful as the 3090 Ti. What I said was that it could be 2x faster at 4K RT/PT vs 3090 (non Ti). Read my messages once again.

The price is going to increase and nvidia confirmed that, maybe you're confusing 5090 with the 5080 which isn't getting one. So $1600 > $2000 would be a 25% increase for what i'm guessing will be 30% more performance. Meh at best, but it'll be the fastest GPU and nvidia can charge whatever so it is what it is.

Meanwhile the 4090 increased by $100 over 3090 and had ~55% more performance.

Agree, the RTX 50s are probably not going to go down unless Intel and AMD really have something that destroys Nvidia GPUs on a Performance/Price ratio.
RTX 40s are becoming harder to get (mostly 4090s) so I don't see the 5090, 5080 prices going down anytime soon. 5070 and 5070 Ti could maybe drop if the 9070 XT is between the 2 in performance but the price is the same as the 5070 or even cheaper! Imo if the 9070 XT was like a 4080 SUPER for $500 or less it would definitely sell really well.

 

[MCJAxolotl7]

He wants to maybe do another colour? Go on then release a badass UDNA card

AMD better make it badass in the first place though. Many maybes.

I would like that a lot. If we get a card that will be equivalent to or slightly below the 6090, good.

 

[tpuuser256]

The problem is that people were expecting the same performance gap as 3090 Ti to 4090 but that increase only happened because Nvidia switched from Samsung 8nm to TSMC 4N !
Blackwell is a new architecture but we can see that CUDA Cores IPC is not getting much better anymore with every new generation, so the only way to get more performance is to increase the number of CUDA Cores & Core Frequencies too, but both increase the Wattage, Voltage and Heat too... RTX 40s and RTX 50s being on a pretty similar node we couldn't expect much! If anything GDDR7 and the 512-bit bus are the big changes here, and ~33% more CUDA Cores.
But the 4090 had ~52% more CUDA Cores and ~35% Higher Core Clocks than the 3090 Ti ! And also a lot more L2 Cache (72MB vs 6MB) to make up for the lack of Memory Bandwidth.

The 5090 seems to have great overclock potential but I have yet to see official support for more than 675 watts of power (75 from pci)

 

[N/A]

The 5090 seems to have great overclock potential but I have yet to see official support for more than 675 watts of power (75 from pci)

Dual 12HPWR as a single source, the slot is no longer used for power delivery. Nvidia is saving this for the Titan.

I would like that a lot. If we get a card that will be equivalent to or slightly below the 6090, good.

More like slightly below 6070 with a 50% bigger die area. They're cooked.

 

[ZoneDymo]

Could go for AMD or Intel this time around just saying, also weirdly negative comments about this legendary overclocker...