Nvidia can't afford TSMC anymore

I'm guessing adoredfaggot's "rumors" about yields being on par with 14nm are actually just made up bullshit, again, and TSMC's 7nm yields are in the dumpster where they were expected to be. And Samsung's EUV 7nm is probably doing much better, especially since Nvidia has it all to itself since the big players like Apple, Qualcomm, Mediatek, etc. all locked into TSMC for the foreseeable future.

Wouldn't be surprised. Samsung skipped 7nm DUV which left them vulnerable and without partners. They need someone big other than their own Exynos team to use their process.

Even with all the memes Nvidia is maintaining a high gross margin above 50% so there's nothing wrong with their cost structure.

samshit's 7nm EUV is going to be delayed because they can't buy supplies from Japan anymore and TSMC will blow the fuck out of them again
novideo btfo

Now they have another excuse to charge more for their products

>they have another excuse
they never needed the excuses

If Samsung will offer a better price, Jensen will go for them. Higher profit margin = bigger bonus = more leather jackets.

TSMC 7nm is kinda shit
It's focused almost entirely to low frequency power efficiency.
It's like 20% more power efficient than 14/16nm with shit clock speeds.

I bet Samsung's 7nm EUV will allow Nvidia to hit 2500mhz

haha yeah right 14nm samshit was used for vega

aren't GPUs usually sub 2ghz? and isn't relatively good power-performance one of the reasons maxwell and pascal did so well

Samshit processes are even more mobile oriented than TSMC

yeah it's shit but samshit doesn't have anything equivalent atm

Probably because samsung costs 1/2 the price for like 95% of the performance of 7nm TSMC.

Also, Nvidia knows Turing is still ahead by like 50% when it comes to GPU architecture over AMD....

Their 12nm (16nm++) is literally on par with AMD's 7nm. If AMD ever provides competition, Nvidia will simply launch 7nm cards.

The hope is that AMD's Navi cards will be good enough to force Nvidia to start trying again, but we'll have to see how they actually perform first.

>Samsung will make Intel GPUs
>now Nvidia as well

fpbp
TSMC doesn't want to make 400+mm2 chips on 7nm.
Others ask them to make tiny little chips which yield well, no additional costs so they gave most of wafer cap to AMD and others.
5700XT is x2.5 smaller than 2070S, it's a good arch, sad they fucked up marketing.

They will launch 7nm next year regardless.
lmfao they did a super refresh and literally fucked AMD out of any shitty gain they were building with their fake news.

lol cope.

AMD is in no way on the same scale as the other two. They don't even ship as many CPU+GPU combined as Nvidia does GPUs. AMD had a 15% lower gross margin than Nvidia with half the revenue for the past few quarters.

>Others ask them to make tiny little chips which yield well, no additional costs so they gave most of wafer cap to AMD and others.
>5700XT is x2.5 smaller than 2070S, it's a good arch, sad they fucked up marketing.
When are they going to do chiplet GPUs?

Leatherjacketman always jews for the max, so Samsung was probably more cost-efficient, especially with everybody and their mother already ordering 7nm from TSMC. He has zero reason to rush things since AMD GPUs are still DoA as far as the PC market is concerned.

Never. Nvidia already researched it and proposed chiplets for compute GPUs but there's too much latency for realtime

fuck offfffffffff you fucking retard

>Nvidia already researched it and proposed chiplets for compute GPUs but there's too much latency for realtime
According to AMD their proposed MCM GPU would have functioned like a crossfire/SLI system with less latency, so it's more like devs are too lazy to optimize for multi-GPU without a full shift from both AMD and Nvidia to MCM/chiplets.

Oh for the love of god you people make up anything to justify your disappointment.
AMD doubled their core density and needed to increase (relative to 14nm density) power to give any clock speed increases.
TSMC advertised from the beginning their 7nm gives +35% transistor performance at ISO power or -50% power at ISO performance.
TSMC 7nm wasn't designed for low clocks you stupid retarded fat nigger piece of shit fuck off back to /v/ or reddit or fagbook or tiktok 9r junior high or wherever the fuck you came from
did I forget to mention you're a brainless stupid nigger

Not really, in a few years Nvidia will have big issues to address.

There are low power and high power libraries for 7nm, dumbass.

Modern games use lots of temporal effects which make multi-GPU use literally impossible unless the two GPUs work effectively as one (= very low latencies, rendering the same frame at the same time).