News AMD's upcoming integrated graphics matches seven years old GTX 1060 in Geekbench 6 — Ryzen 5 8600G iGPU benchmarks leak

[dmitche31958]

If true, they will be perfect for budget E-sports rigs.

Perhaps they should start a league of imbedded GPU gamers only. lol.

 

[PiersPlowman]

Its why they didn't make any GPU below 7600, they have been saying it will replace a lot of low end cards.

Integrated graphics have a long and illustrious history since the early days of home computing. The Commodore Amiga and its variants set the high watermark in the early days, followed by the Atari ST (though that was basically just a framebuffer and a CPU before Atari got around to tossing in a blitter, hardware scrolling, genlock, and one or two other goodies). However, I feel that such solutions belong in that bygone era; okay, so having a fallback for VGA is great, but what we are witnessing is the "consolization" or "Macifying" of the Windows desktop PC. We are being locked down more and more unless we are prepared to pay for the "unlocked" version - the dGPU.

This is ongoing anti-consumer behaviour from AMVidia and I hope, nay, pray, Intel offers us something compelling in the sub $300 category.

 

[Colif]

Expecting Intel to save you is a pipe dream. They're no better than the other two

 

[PiersPlowman]

Expecting Intel to save you is a pipe dream. They're no better than the other two

Read my post again. Hope indicates a desire or wish (probably fruitless in this case); it does not indicate an expectation. I do agree with your sentiment, however.

 

[hannibal]

Intel battlemage will definitely at least douple Intel prices… unless battle mage is crap.
Intel Stock owners want to have fast profit margins… and Current gen does not provide that.
So these apus that offer minimal level gaming is theonly hope to budget gamers. Old used GPU is the next update ower these is the next step above to those who has little bit more money.

 

[H4UnT3R]

After having seen how much DLSS affects FPS on an RTX2050 4GB.
If 8000G series supports FSR3 and frame generation, then I can see it being an absolute slayer of everything low end.

If is my Spanish correct, it can match 3050 with FSR (sarcasm, I don't understand <Mod Edit> in Spanish).
https://www.youtube.com/live/b3Uel4LbvCs?si=VDxLsWcAhEZNdIr5

 

[palladin9479]

The dream (no need for a dGPU in laptop) will be realised when AMD makes a Ryzen 9 sized chip with 1 CCD compute tile and 1 CCD graphics tile. 40 CU and 2560 shader cores, competing with PS5 GPU in terms of core count.

Won't happen. With iGPU's it's all about the shared memory subsystem. Most every CPU made in the past two decades use's two 64-bit memory channels for 128-bit aggregate interface. System memory is designed for low latency / fast command response times as scalar code frequently needs the results of the operations to decide which path to take next. GPU's processing is almost entirely vector code, which isn't nearly as sensitive to command times and prefers just having massive bandwidth it can use to run hundreds if not thousands of calculations simultaneously. That is why GDDR has terrible latency but higher bandwidth. HBM was supposed to be the solution as it's both low latency and high bandwidth but turned out to be very expensive.

 

[Lucky_SLS]

^ true, but that is what apple is trying to do but with its M series chips. And in a few years, cuz of this approach, i think we will see Macs as a formidable gaming PCs and ecosystem...

 

[thestryker]

I think Qualcomm is going to be the one who has a chance to push us beyond 128bit bus for everything desktop/laptop. It sounds like they've been able to do great things with 128bit but if they plan on continuing to integrate high performance graphics (as opposed to relying on discrete) they're absolutely going to need a higher bus width available for their next chip.

 

[palladin9479]

^ true, but that is what apple is trying to do but with its M series chips. And in a few years, cuz of this approach, i think we will see Macs as a formidable gaming PCs and ecosystem...

Macs are just ARM SOC's on TSMC's newest fabrication node. Apply pays a ridiculous sum of money to TSMC to prepurchase capacity before any of the other big players. They are using LPDDR memory that's been soldiered to the board. Anyone else could accomplish the same task by just having four 64-bit memory bus's to four separate DIMM slots. There used to be some high end CPU sockets that had quad channel memory support, ended up being very expensive which caused mainstream not to pick it up.

 

[Lucky_SLS]

^Very true. Maybe not a 40cu, 2500 core gpu that is starved for bandwidth but still something more powerful than the 780M.

The sucess of Steam deck with 2 channel memory and 90gbps bandwidth made me think of this. Handheld gaming is definitely not loosing steam and we will see more advancements in this space!

 

[thestryker]

^Very true. Maybe not a 40cu, 2500 core gpu that is starved for bandwidth but still something more powerful than the 780M.

The 780M is already bandwidth starved as evidenced by the performance difference going from 6400 to 7500 in the Ally. There's only so much that can be done without increasing memory bandwidth. Consider that the RX 6400 has equivalent bandwidth to DDR5 8000. The 780M is the same hardware configuration as the 6400, but RDNA 3 instead of RNDA 2 and has higher boost clocks.

 

[Brian31525]

Still running a 1060 6gb on a new rig. 30 plus Fps in forza horizon 5 on medium settings. I had it on Sandy b4 i built my 13700kf and even down a lane it ran most games. I've always gone for the 60s series cards from a 560 ti on up.

 

[H4UnT3R]

^ true, but that is what apple is trying to do but with its M series chips. And in a few years, cuz of this approach, i think we will see Macs as a formidable gaming PCs and ecosystem...

Nope. For basic tasks, ok, but M3 is comparable to RTX3080... without option of any upgrade like switching GPU

 

[H4UnT3R]

Macs are just ARM SOC's on TSMC's newest fabrication node. Apply pays a ridiculous sum of money to TSMC to prepurchase capacity before any of the other big players. They are using LPDDR memory that's been soldiered to the board. Anyone else could accomplish the same task by just having four 64-bit memory bus's to four separate DIMM slots. There used to be some high end CPU sockets that had quad channel memory support, ended up being very expensive which caused mainstream not to pick it up.

Well you can't. There is some memory even on chip, they said somewhere smth about speeds around 900GB/s, which is comparable to nvlink, but still - it's limited and for applications like this, most people will prefer switchable GPU.

 

[TerryLaze]

Why is nobody ever reading the actual article???
These are the things tested by the benchmarks....
anybody please go ahead and tell us how any of that is relevant to gaming even in the remotest of relations. That's also how it's better than the 1060 but also worse than the 1030, because it has nothing to do with gaming but only with compute.
It's a strong compute GPU, it might be good for gaming too but that would be irrelevant to these benches.

OpenCL Performance​

OpenCL Score	24842
Background Blur	15492 64.1 images/sec
Face Detection	9338 30.5 images/sec
Horizon Detection	28934 900.4 Mpixels/sec
Edge Detection	29648 1.10 Gpixels/sec
Gaussian Blur	20407 889.2 Mpixels/sec
Feature Matching	8293 326.9 Mpixels/sec
Stereo Matching	97636 92.8 Gpixels/sec
Particle Physics	70745 3113.6 FPS

Vulkan Performance​

Vulkan Score	30770
Background Blur	15890 65.8 images/sec
Face Detection	9844 32.1 images/sec
Horizon Detection	28691 892.8 Mpixels/sec
Edge Detection	32901 1.22 Gpixels/sec
Gaussian Blur	48167 2.10 Gpixels/sec
Feature Matching	9326 367.7 Mpixels/sec
Stereo Matching	105548 100.3 Gpixels/sec
Particle Physics	114791 5052.1 FPS

 

[Lucky_SLS]

Nope. For basic tasks, ok, but M3 is comparable to RTX3080... without option of any upgrade like switching GPU

I was talking more about big large gpu like the 40 core m2 max with 400gbps memory bandwidth...

As for being equivalent to the 3080, I won't comment.