5080 16gb is literally the 4080 12gb (aka 4070ti) all over again

[N/A]

yawn...specs don't matter, the actual performance and price are what matter.

For all we know 5090 could be 40% faster than 5080, even with 2x the specs.

Ideally it should be 95%, but the CPU bottleneck is not excluded even with the best X3D and 14th generation processors.

 

[Waldorf]

@Krit
the problem isnt with ppl buying expensive stuff, its ppl spending more than they should, buying a certain model nbr, instead of what they can afford with their wallet.

if i can get an xx70ti/80 for what i want to spend, im fine, and im willing to spend a little more for a better (less noisy/cooler) card, but im not wasting my money just so i can have a certain model in my rig, unless im getting 6 figure lottery win..

 

[agent_x007]

It could be worse... a quarter of GPU for 1/3 of price :
[G80] 8800 GTX 128SP 384bit/768MB = 600$ (Nov. 2006)
[G84] 8600 GTS 32SP 128bit/256MB = 200$ (April 2007)

 

[nomdeplume]

xx70 actually xx60 , xx60 actually xx50 at least since the 3000 series, I am not surprised.

XD

Closer to 8 = 6 ¼ being a rough forumula that holds true when you plug in 5-7.
Time to draw up a blacklist of members who bought 4xxx

 

[Macro Device]

a quarter of GPU

Can't say it's really a quarter. Significantly higher clocks (675 VS 576 MHz) and 32 GBps VS 86.4 GBps VRAM bandwidth. This is 29% GPU speed and 37% VRAM speed for 33% price. Still terrible though.

 

[RUSerious]

I was called a deranged AMD fan despite only having a GT 1030 at the moment. Society is never reasonable I guess.

The rabble on the internet are often unreasonable. There are far more reasonable ppl in real life.


As to the rest, we will just have to see how the actual cards perform and where they are priced. Rumors have proven to be off in the past (and way off for the example of the Ryzen 9000 series CPUs).

 

[leonavis]

As to the rest, we will just have to see how the actual cards perform and where they are priced. Rumors have proven to be off in the past (and way off for the example of the Ryzen 9000 series CPUs).

They do deliver significant gains with Linux, weirdly enough. https://www.phoronix.com/review/amd-ryzen-9950x-9900x/15

Productivity with 9900X and 9950X being 20 and 18% faster than their 7900X/7950X counterparts. Similar effects could be seen with the 9700X for gaming (German).

Apparantly a driver issue for Windows?

Yeah I know entirely off topic. To the topic at hand... Nvidia will name their line up how they wish. They wanted to call the 4070 4080 already, apparantly now they follow through. Probably economically smart to make the 5090 very, very pricey and the 5080 way less performant.

 

[agent_x007]

Can't say it's really a quarter. Significantly higher clocks (675 VS 576 MHz) and 32 GBps VS 86.4 GBps VRAM bandwidth. This is 29% GPU speed and 37% VRAM speed for 33% price. Still terrible though.

You look from buyers/performance perspective.
From NV point of view, after design is done they only pay for die manufacturing.
So, G80 vs. G84 would look like this in NV books : 681M @ 484mm^2 vs. 210M @ 127mm^2.
Which means on 12" wafer NV can fit ~4x more G86s, than G80s (depending on actual dimensions, more square = better). More dies = more GPUs can be sold, math is simple : 4x 200$ > 1x 600$.
I doubt cost of G80 die was actually 4x higher than G84* for card manufacturers in 2007 (but this is pure speculation).
Sure binning takes place, but rejects from GTS just get thrown into 8600 GT in this case (still 160$ card in 2007).

*Note : G84 is 80nm part, while G80 is 90nm.
Manufacturer cost per die WILL be different vs. same technology I assumed above.

 

[Macro Device]

You look from buyers perspective

To look from their standpoint one needs to know how much this all costs. How much for one wafer of this, how much for a wafer of that, logistics, rejects, yadda, yadda. I have zero clue. But I do know that 8600 GTS is faster than 25% of 8800 GTX.

And I'm fully aware of why they do it like that. Still, doesn't mean I don't have a right to be bitter about it.

 

[RUSerious]

They do deliver significant gains with Linux, weirdly enough. https://www.phoronix.com/review/amd-ryzen-9950x-9900x/15

Productivity with 9900X and 9950X being 20 and 18% faster than their 7900X/7950X counterparts. Similar effects could be seen with the 9700X for gaming (German).

Apparantly a driver issue for Windows?

Linux delivers. Shocking - lol! Zen5's ccds are designed for server first - probably at least 10x more Epyc's sold are running a version of Linux vs Windows (hyperscalers/data centers).

Yeah I know entirely off topic. To the topic at hand... Nvidia will name their line up how they wish. They wanted to call the 4070 4080 already, apparantly now they follow through. Probably economically smart to make the 5090 very, very pricey and the 5080 way less performant.

They will. There was enough backlash via YouTuber's and in social media against a '4070' class card being called a 5800. The same thing is happening again - but I have a bad feeling they NV will care less this time. They are rolling in $$$s over the AI craze.

 

[solarmystic]

It could be worse... a quarter of GPU for 1/3 of price :
[G80] 8800 GTX 128SP 384bit/768MB = 600$ (Nov. 2006)
[G84] 8600 GTS 32SP 128bit/256MB = 200$ (April 2007)

Funny thing about the 8000 series, while it started off being overpriced af for top end performance (in 2006), NVIDIA later on (barely a year later) also delivered one of the best bang for buck cards ever with the 8800 GT at a shocking 249 USD. While it is true you could never find the card at 249 for a long time (it often went for 300 USD minimum), it gave you 85-90% of the 8800 GTX's performance for half the price!

Imagine NVIDIA pulling off a move like that in 2024

 

[Krit]

Imagine NVIDIA pulling off a move like that in 2024

That would be end for AMD and intel final KO blow! But it will not happen because of current buyers iq's. They will buy everything what nvidia will offer even low end 500$+ garbage nice!