Support NeoGAF

[Tqaulity]

That's false though. The 7700 XT isn't less than 50% of the performance of the 7900 XT. It's around 60-65%. That's quite a bit above what you're suggesting.

Yep you're right. I was typing in a rush. I edited my original post.

 

[Leonidas]

Yep you're right. I was typing in a rush. I edited my original post.

Make Sony's purported 45% uplift make sense to me.

That's why I, and others beleive it will perform like 7700 XT in raster.

 

[Tqaulity]

Make Sony's purported 45% uplift make sense to me then.

That piddly (for a mid-gen upgrade) 45% uplift is why people (like me) say it will perform like a 7700XT (in raster).

To be clear, I'm not saying it won't only be 45% uplift. That's my point...first of all I (we) can't say anything concreate on these unconfirmed specs. But to your question, I can't make it make sense anymore than anyone else has been able to. There are several things about those leaked specs simply don't add up which is why I don't really trust them. One of them is the 45% number which doesn't match the other details in the leak. If it is only 45%, then the 60CU, 256bit bus, 576GB/s bandwidth, cache size increases and 16GB of RAM can't be right. If those are in fact the GPU config, then the 45% figure is either complete wrong or at least misinterpreted and taken out of context (I suspect the latter).

 

[asdasdasdbb]

To be clear, I'm not saying it won't only be 45% uplift. That's my point...first of all I (we) can't say anything concreate on these unconfirmed specs. But to your question, I can't make it make sense anymore than anyone else has been able to. There are several things about those leaked specs simply don't add up which is why I don't really trust them. One of them is the 45% number which doesn't match the other details in the leak. If it is only 45%, then the 60CU, 256bit bus, 576GB/s bandwidth, cache size increases and 16GB of RAM can't be right. If those are in fact the GPU config, then the 45% figure is either complete wrong or at least misinterpreted and taken out of context (I suspect the latter).

Because while theoretically the Pro can run at similar clocks as the original, it can't sustain it.

 

[CuteFaceJay]

Not sure the value is there to upgrade from a 3080. We'll see though!

prolly not for a while tbh. the high end 30 cards still rip.

 

[SantaC]

7900XT performance should be 299 imo

 

[SolidQ]

7900XT performance should be 299 imo

4090 only for 100$

 

[TheRaidenPT]

I'm curious about the next evolution of RDNA, just sucks that they are leaving the high-end (....again....)

The 7900XTX was a really good product as long as you accept the Ray Tracing limitations, personally for me I set RT to Low/Medium. In terms of pricing if you could get it under 980€, I got mine for 820 in a open box deal. It's a hell of a card especially paired with the 7800X3D.

*Disclaimer: PS5 Pro specs are still only based on rumors until Sony makes an official announcement

Exactly. Everyone on here is still talking about the Pro being ~7700XT based on the (unconfirmed) leaked specs and it makes no sense. Everything about the leaked specs point to a card that is more powerful than a 7700XT (I've already broken this down on the PS5 Pro thread). The GPU config is most closely matched to a Navi48 and even with the 7% reduction in shaders and assuming a 10-20% reduction in clocks, that should still put it's performance between a 7800XT and 7900GRE (assuming the NAVI 48 is >= a 7900XT).

Keep in mind that a 7900XT is ~25-30% faster than a 7800XT and a 7800XT is ~20-25% faster than a 7700XT. In other words, a 7900XT is over 50% faster than 7700XT. Now in what world would Sony take a GPU and gimp it to the point of reducing it's performance by that much for yields and power limitations?! That has never happened in any console generation. 20%? Maybe. >30% not likely!

True, plus with extra game optimization and PSSR I'm sure we gonna see some more "magic tricks"

 

[Hudo]

The problem is software killer apps and dev ecosystem for such software and tools.

This.

AMD and Intel aren't that far behind when it comes to hardware (well, Intel still have to build up their offering...). The main competitive edge that Nvidia have over others is their software stack. It's also how they manage to keep people (like myself) vendor locked. CUDA and the ecosystem around it just doesn't have a viable alternative. OpenCL is not even close, Metal is mainly being pushed by Apple and mostly only useful if you want to utilize their M chips. ROCm is also not even close. And Intel's OneAPI, at least so far, is just there so Intel can say that they're doing something.

Maybe the newly formed alliance between Qualcomm, Google, Intel and others to develop an open-source CUDA competitor will break Nvidia's stranglehold. Who knows.

 

[Kenpachii]

So basically the 300 series all over again

 

[TheRaidenPT]

This.

AMD and Intel aren't that far behind when it comes to hardware (well, Intel still have to build up their offering...). The main competitive edge that Nvidia have over others is their software stack. It's also how they manage to keep people (like myself) vendor locked. CUDA and the ecosystem around it just doesn't have a viable alternative. OpenCL is not even close, Metal is mainly being pushed by Apple and mostly only useful if you want to utilize their M chips. ROCm is also not even close. And Intel's OneAPI, at least so far, is just there so Intel can say that they're doing something.

Maybe the newly formed alliance between Qualcomm, Google, Intel and others to develop an open-source CUDA competitor will break Nvidia's stranglehold. Who knows.

Question is, how will AMD & Intel at least accomplish to dethrone 30-40% of the market share that NVIDIA has?

 

[Zathalus]

$499 for 7900XT performance + better RT is pretty good. Looking at what Nvidia is doing the 5070 will likely still be 12GB, cost $599 and still be 10% slower.

 

[SHA]

I'm not optimistic about 10th gen consoles graphics, the future doesn't look bright with graphics, is it really gonna be an issue with next gen consoles?

 

[yurinka]

AMD RDNA 4 "Radeon RX 8000" GPUs Reportedly Debut In 2025: Navi 48 At CES & Navi 44 In Q2

AMD is reportedly going to launch its next-gen RDNA 4 "Radeon RX 8000 GPUs, primarily the Navi 48 & Navi 44 chips, in the first half of 2025.

wccftech.com

Tl:;dr:
- Navi 48 (top SKU) likely announced at CES in January, launching Q1, around 7900XT performance but lower power consumption for $500 or less

- Navi 44 (entry level) coming Q2

- high end/enthusiast tier is canceled, won’t have a high end GPU til RDNA5

Is it possible that the PS5 Pro would feature a tweaked version of that navi 48 (removing some stuff, adding other stuff)?

 

[DonkeyPunchJr]

Is it possible that the PS5 Pro would feature a tweaked version of that navi 48 (removing some stuff, adding other stuff)?

The rumors I’ve heard are that the PS5 Pro specs were finalized a long time ago, so probably not

 

[Kataploom]

I'm ok with that, I have expenses until first or second half of next year so my upgrade plans got delayed anyway, so I rather get a more modern and cheaper GPU with same performance of than the one I was planning to buy anyway

 

[Hudo]

Question is, how will AMD & Intel at least accomplish to dethrone 30-40% of the market share that NVIDIA has?

The only real way to do that would be to price their stuff aggressively and to keep up with CUDA. Monumental tasks but if they can get a lot of the industry behind one open standard, it might be possible.

 

[Gaiff]

The only real way to do that would be to price their stuff aggressively and to keep up with CUDA. Monumental tasks but if they can get a lot of the industry behind one open standard, it might be possible.

Wouldn't work. For one, NVIDIA can respond in kind. For two, this has never really bore fruits every time it was attempted. AMD has always been seen as the budget brand and even when they priced their GPUs competitively, it didn't really help them. What they need is to release products that are competitive with NVIDIA in the same tiers, not inferior but cheaper cards.

AMD didn't start stealing Intel's lunch until they released better CPUs again with Ryzen.

 

[Hudo]

Wouldn't work. For one, NVIDIA can respond in kind. For two, this has never really bore fruits every time it was attempted. AMD has always been seen as the budget brand and even when they priced their GPUs competitively, it didn't really help them. What they need is to release products that are competitive with NVIDIA in the same tiers, not inferior but cheaper cards.

AMD didn't start stealing Intel's lunch until they released better CPUs again with Ryzen.

Huh? That's what I meant, tho. With "pricing their stuff aggressively". Obviously that includes being on par hardware-wise, which is a realistic goal. The harder part is the software stack because this is where AMD are lacking quite severely. And if AMD can join the alliance set up by Intel, Google and Qualcomm, it might be possible to compete with Nvidia there as well.
And if Nvidia respond, all the better! A 4090 shouldn't be more than $1200. The current pricing is just pure insanity.

Also: I vividly remember that, long ago, AMD/ATI had cards that were not only on par but also arguably better than Nvidia's offerings at the time. I did love my Radeon 9800 Pro. Or my X1950 XT.

 

[Crayon]

Forget the tech. Just make it a good price AT LAUNCH instead of a year later. They could come out with some crazy awesome card but then jack up the price and then it's no better than rdna 2 or 3. People say you save money with amd and you can, but not by buying them at full price. You have to wait a year or year and a half for them to slash the prices before they are a really good deal.

 

[YeulEmeralda]

Sounds like another L for AMD. The only reason most people hope AMD lowers their prices is to force Nvidia to lower theirs, and then they’ll buy Nvidia anyway.

 

[SolidQ]

CUDA and the ecosystem around it just doesn't have a viable alternative. OpenCL is not even close, Metal is mainly being pushed by Apple and mostly only useful if you want to utilize their M chips. ROCm is also not even close.

SCALE documentation

docs.scale-lang.com

 

[Allandor]

CUDA and the ecosystem around it just doesn't have a viable alternative. OpenCL is not even close, Metal is mainly being pushed by Apple and mostly only useful if you want to utilize their M chips. ROCm is also not even close. And Intel's OneAPI, at least so far, is just there so Intel can say that they're doing something.

It is astonishing what a company could do if it blocks other standards they should also actively develop with other companies (for years) when you can do their own thing and everybody uses it without questioning.
And now, nvidia is more or less leaving the gaming market (they can sell the same hardware for much, much higher prices in other sectors). Maybe now there is a chance that opencl, DirectML etc get the attention those standards deserve.