GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
Is the 9060xt 16gb a worthwhile upgrade from the b580? I found a 9060 for about $100 more than my b580. Should I get it or is it just a sidegrade?
when you look at the specs a b580 slaps the 9060xt on the face for a 4GB vram tradeoff.
bruh.
i don't know what kind of crack intel was smoking when they made the b580 but i hope they continue, the gaming market is in dire need of good fucking gpu's that don't bend over to AI bullshittery, maybe if blue team starts getting these gamers nvidia and amd might move their fat fucking arses or just ditch gaming altogether and go to AI.
 
when you look at the specs a b580 slaps the 9060xt on the face for a 4GB vram tradeoff.
bruh.
i don't know what kind of crack intel was smoking when they made the b580 but i hope they continue, the gaming market is in dire need of good fucking gpu's that don't bend over to AI bullshittery, maybe if blue team starts getting these gamers nvidia and amd might move their fat fucking arses or just ditch gaming altogether and go to AI.
The 8 GB 9060 XT is 28% faster (1080p), 23% faster (1440p) than the B580 in TPU's review: https://www.techpowerup.com/review/powercolor-radeon-rx-9060-xt-reaper-8-gb/33.html

The 16 GB version is about 29.6% faster than the B580 at 1440p. +35.3% in Tom's Hardware's review.

B580 is $280 on Newegg. So even at $380 (+36%), the 9060 XT 16 GB isn't that bad, and price/perf is better if you managed to grab it at $350. If the 8 GB model falls below MSRP to around where the B580 is and you can cope with 8 GB VRAM, it's going to be faster.

Don't fall for the Intel meme.
 
  • Like
Reactions: Combustion Engine
Currently still rocking my 3080 10gb FE card. I am a graphics whore and looking to get a 5080 FE but gawddamn these fuckers have no stock at any of my local Best buys. I would've thought that post pandemic and post scalper shit supply wouldn't be so fucked still.
 
I would've thought that post pandemic and post scalper shit supply wouldn't be so fucked still.
Well all the chip fabrication time went to AI shit. Intel, AMD, and NVidia all use TSMC right now so there's only so much supply to go around.

I am a graphics whore and looking to get a 5080 FE
Yeah it'll be at least 3 more months before there starts to be decent stock levels. Though I'd personally just keep waiting and get a 5080 Super in your situation.
 
I don't see the low and entry level GPU market surviving, With AMD perfecting the APU. Make a PS6 equilvent APU for PC use and a Big APU and there you go.
Unless amd can massively increase memory bandwidth for the apus, they will always have decent limits on how much performance they can push out. compared to an equivalent dedicated gpu This is why the 780m is in theory faster than a gtx 1060 but in practice is much slower. The strix halo GPU can be much faster but it has less bandwidth than the slowest gddr6, so even if you were to put 200w watts on it there would be diminishing returns. Unless they use gddr for desktop memory or increase the memory bus for the cpus on desktop. the low end market will always have a place.
 
Yeah it'll be at least 3 more months before there starts to be decent stock levels. Though I'd personally just keep waiting and get a 5080 Super in your situation.
yeah that might be the correct play assuming that there even is a super for this gen. At this point i'm just getting to the point where it seems like it might be a better idea to wait for the 6000 series.
 
  • Optimistic
Reactions: StacticShock
Don't fall for the Intel meme.
tell that to my wallet, lisa, it's far cheaper for me to buy a b580 than a 9060 XT and seeing how this shit loses it's value over time, safest best ever.

the low end market will always have a place.
low end is mainly meant to fool suckers or to be sold in third world, if apu's make their way to the prebuilt computers space then they will eat up the entire reason of why low end gpu's should exist because arsing yourself with a prebuilt low end pc with a low end gpu versus a high memory apu pc makes you see that the price difference is staggering but seeing how lisa su is retarded, it will take a while until they put someone who isn't a nigger in there that has familial connections to their competitor.
 
  • Feels
Reactions: Brain Problems
Unless amd can massively increase memory bandwidth for the apus, they will always have decent limits on how much performance they can push out. compared to an equivalent dedicated gpu This is why the 780m is in theory faster than a gtx 1060 but in practice is much slower. The strix halo GPU can be much faster but it has less bandwidth than the slowest gddr6, so even if you were to put 200w watts on it there would be diminishing returns. Unless they use gddr for desktop memory or increase the memory bus for the cpus on desktop. the low end market will always have a place.
The biggest market for APUs is laptops and miniPCs, and those are the platforms most likely to switch entirely to soldered memory in the near future. Soldered memory massively increases bandwidth, especially if AMD were to make a high bandwidth channel specifically for the graphics chip. Like you could have 16GB of LPDDR5 to share between CPU and GPU, and 3GB GDDR7 dedicated for the APU.
 
I don't see the low and entry level GPU market surviving, With AMD perfecting the APU. Make a PS6 equilvent APU for PC use and a Big APU and there you go.
Unless amd can massively increase memory bandwidth for the apus, they will always have decent limits on how much performance they can push out. compared to an equivalent dedicated gpu This is why the 780m is in theory faster than a gtx 1060 but in practice is much slower. The strix halo GPU can be much faster but it has less bandwidth than the slowest gddr6, so even if you were to put 200w watts on it there would be diminishing returns. Unless they use gddr for desktop memory or increase the memory bus for the cpus on desktop. the low end market will always have a place.
It's likely that AM6 will remain 128-bit, so socketed desktop APUs and the mainstream APUs they are based on will continue to eat shit, although it will be better shit than ever before (DDR6 and CUDIMMs). But they could improve on it by finally starting to include Infinity Cache (L3 cache for GPU) in mainstream APUs, which raises the "effective bandwidth" and helps mitigate the bandwidth issue. This will be something to watch out for with Medusa Point and future APUs. Strix Halo has 32 MiB, but even 16 MiB (like Navi 24) could help.

Zen 6 desktop (Olympic Ridge) could boost the iGPU after two gens of lowly Radeon 610M graphics. We have a tentative prediction of up to 8 CUs of RDNA4, which could put it near the 8600G. 4 CUs of RDNA3 or better would match or beat the 8500G. But it might not make it fully intact to the low-end 8/10-core of the lineup.

Strix Halo is AMD's first attempt at a consumer "mega APU", and they can make big changes to it going forward as long as the physical size doesn't grow too crazy. Medusa Halo might support 384-bit LPDDR6 which could be a big upgrade. LPDDR6 was just officially announced, and it uses 24-bit channels (12-bit sub channels), which is probably where the 384-bit leak is coming from. What was 256-bit became 384-bit with the same number of memory chips, probably. We'll see what the experts have to say about it.

Strix Halo's pricing and relative GPU performance makes it irrelevant for the low-end. Maybe that can change in the future. For now it's an LLM powerhouse.

Mini PCs and mobile-on-desktop chips soldered onto motherboards, which can include a "Halo" but not for budget users anytime soon, are the way forward for the low-end. There are Rembrandt/Phoenix mini PCs with RAM and storage for around $300 to $500 on sale, undercutting almost any new desktop build and offering convenience and a small form factor. They don't offer the price/perf of throwing a $300 GPU with the cheapest parts you can find, since they have more CPU and less GPU performance than needed for gaming.
 
Unless amd can massively increase memory bandwidth for the apus, they will always have decent limits on how much performance they can push out. compared to an equivalent dedicated gpu
Yeah, but in 2025, that performance limit is something like 60 fps @ 1080p and a mix of medium & low settings. FSR & XeSS ease the pressure on bandwidth even further. Sure, my iGPU can't handle 200 fps or 4K, but who cares?
 
  • Agree
Reactions: Rololowlo

"Zen 6 is targeting 7 GHz, if I'm being honest, it's actually targeting above 7 GHz. And for those who think that sounds crazy, remember, they're not going from N4P to N3, they're going from N4P to N3P to N2P to N2X. If you look at the latest TSMC foundry cadence, they consider N2X to basically be a node better for some customers than N2P. With that in mind, that's almost three node jumps...

They're definitely achieving 6.4 GHz or higher, and the goal is above 7 GHz... 7 GHz is actually kind of in the middle of where they're trying to achieve... know that when I say 7 GHz, it's in the middle of the top goal, and what they're sure they will achieve already."

He says N2X would be for Olympic Ridge desktop CPUs and maybe some high-clock Epyc chips. Other parts such as APUs would use N2P and even N3P.

I wouldn't call N2X a "node jump" from N2P, since it should be design compatible and the same size, but it's focused on delivering "+10% performance" from higher clocks, voltage, and current. That could mean +600 MHz over N2P.
 
They're definitely achieving 6.4 GHz or higher, and the goal is above 7 GHz... 7 GHz is actually kind of in the middle of where they're trying to achieve... know that when I say 7 GHz, it's in the middle of the top goal, and what they're sure they will achieve already."
It's Y2K all over again, baby! Netburst 2: Electric Boogaloo
 
Yeah, but in 2025, that performance limit is something like 60 fps @ 1080p and a mix of medium & low settings. FSR & XeSS ease the pressure on bandwidth even further. Sure, my iGPU can't handle 200 fps or 4K, but who cares?
The 1% lows on the apus are the problem. I've tried using the igpu on my laptop (780m) and on the turbo setting (50-90w) the stuttering is terrible.

"Zen 6 is targeting 7 GHz, if I'm being honest, it's actually targeting above 7 GHz. And for those who think that sounds crazy, remember, they're not going from N4P to N3, they're going from N4P to N3P to N2P to N2X. If you look at the latest TSMC foundry cadence, they consider N2X to basically be a node better for some customers than N2P. With that in mind, that's almost three node jumps...
If this is true I will swap my 9700x for it.
 
Back