Vega GPU benchmarks emerge, revealing several GPU specifications

[AngryGoldfish]

It needs to be at least as fast as a 1080. The Fury X is a good rival to the 1070 in DX12 today, because of Async. So it needs to be faster, otherwise what's the point?

I've never really thought about that until now, but yes, if it comes along with 1070 performance then it's pretty much going to be a Fury X with more VRAM.

A huge side step.

So with any luck yeah, it will offer 1080 performance for less and I will be able to jump onto that

It's 1080 performance I am looking for, I'm just not paying £500 for it.

This is what I like to say to people who claim Vega will only be slightly faster than a 1070. Considering a Fury X is already potentially up there with a 1070, being only marginally behind in certain benchmarks, surely two years later they'd be able to dramatically increase that performance at a significantly lower cost. Goodbye Radeon if they can't.

 

[Shingara]

well i think vega is gonna be a beast, these benchmarks peg it close to the 1080, and nvidia are bouncing around the idea that volta is the answer to vega, so who knows, could be junk, could be a tech demo or it could be godlike.

 

[WYP]

Vega's GPU performance will depend on how well Vega's design changes will pay off, with the new "Next Generation Compute Unit" design being of particular interest.

With an expected core count of 4096 (same as the Fury X) and the promise of "higher clock speeds and IPC", the GPU should yield some good results. The Fury X has clock speeds of 1000MHz, with Polaris (RX 480) coming in with reference boost clocks of 1266MHz was already a decent increase in terms of clocks, Vega already promised higher clocks.

Let's estimate that Vega's boost clock speeds are at least 1300MHz (estimated number taken from the arse dimension), giving it a 30% increase in performance without any other design changes compared to the R9 Fury X. Then combine this with the design changes in the geometry pipeline and the use of a new "Draw Stream Binning Rasterizer" (tile based rasterizer) to cull necessary geometry/work on a scene and Vega should gain some decent performance per clock.

The question is how much these design changes will yield when it comes to increased game performance, especially when it comes to using features like Primitive shaders and rapid packed math (both of which require developer support/intervention to function).

Most of Vega's design changes are to get more work out of Vega per clock and limit the amount of unnecessary shading work in any scene, reducing the work that needs to be done and decreasing the number of clock cycles required to do it.

These are some very smart design decisions from AMD, so it would be great to see it pay off with the RX Vega.

Right now the R9 Fury X performs in GTX 980Ti/GTX 1070 territory, though some games it can fall way behind that. I really hope that AMD does well with the RX Vega, as we really need some competition in the GPU space.

 

[AlienALX]

This is what I like to say to people who claim Vega will only be slightly faster than a 1070. Considering a Fury X is already potentially up there with a 1070, being only marginally behind in certain benchmarks, surely two years later they'd be able to dramatically increase that performance at a significantly lower cost. Goodbye Radeon if they can't.

I don't think big Vega will be 1070 perf tbh man. I just think the small one will and will probably come along first. Who knows though? here's hoping I am wrong

 

[russler]

if vega doesnt destroy 1080s then what was the point? everyone seems to forget that the 1080 released almost a year ago, another thing to consider is the price point, after the recent price drops nvidia carried out is a theoretical £400 gunna cut it. love amd but they take way to long to release anything and the benefits are only noticed months/years down the line.

 

[Shingara]

Well i dont think its so much the grunt that its got behind it thats the key, i think its stuff like WYP was pointing at, its how it does the bulk of the work. Its a case of work smarter then work harder. And when we look at stuff like that it gets into the gameworks territory where we see huge impacts even on the last gen nvidia cards.

 

[AlienALX]

if vega doesnt destroy 1080s then what was the point? everyone seems to forget that the 1080 released almost a year ago, another thing to consider is the price point, after the recent price drops nvidia carried out is a theoretical £400 gunna cut it. love amd but they take way to long to release anything and the benefits are only noticed months/years down the line.

Nvidia have cut the price of the 1080 to $499, which so far has equated to £499. Mainly due to VAT etc. Any way, that's still not cheap, using the same logic that a 1080ti is not cheap at £699 just because the Titan X was £1200 or whatever.

So yeah, if it performs like a 1070 then whilst it will be of no use to me the key will be the price. The 1070 is still mid £300s to £400 if AMD can come in at say £250 they would sell loads.

Just like the 480. It hasn't beaten anything, but it's sold well because of the keen price. A 1070 performing AMD card would be the perfect upgrade for those using a 480.

I am hoping for more though. Me personally, likes.

 

[AngryGoldfish]

I don't think big Vega will be 1070 perf tbh man. I just think the small one will and will probably come along first. Who knows though? here's hoping I am wrong

I imagine Vega 10 will come before Vega 11. We know Vega 11 is not replacing Polaris, which I always thought sounded off when people suggested that Vega 11 was going to replace Polaris, and we haven't heard a peep regarding it so I imagine Vega 10 will drop first.

 

[Doomslayer]

IMO AMD doesn't have an answer to the GTX 1080Ti or Titan XP but they do have an answer to the GTX1080 and below.

Plus anyway i wouldn't be buying any AMD GPU's IMDO they run way to hot.

 

[Davva2004]

My interpretation is that AMD are trying to implement a way of unifying system memory between the CPU and GPU using their infinity fabric tech, rather than having the rigid boundary of DRAM for CPU and VRAM for GPU. Which makes a lot of sense, considering their work in the console market.

 

[Shingara]

IMO AMD doesn't have an answer to the GTX 1080Ti or Titan XP but they do have an answer to the GTX1080 and below.

Plus anyway i wouldn't be buying any AMD GPU's IMDO they run way to hot.

Well that just gives us a reason to build sweet sweet water cooling systems, and its a bit off saying it like nvidias run cool, its not like they run ice cold themselves.

 

[Doomslayer]

Well that just gives us a reason to build sweet sweet water cooling systems, and its a bit off saying it like nvidias run cool, its not like they run ice cold themselves.

For me Nvidia DOES run cooler, Last time i own a AMD GPU was a R9 390 that thing was running 20C hotter than my Nvidia GPU.

That's why i have said i don't like AMD GPU's purely of the fact they run hotter, I will never own another AMD GPU ever again.

I have never been happy with a AMD GPU's compared to Nvidia GPU's.

 

[Shingara]

For me Nvidia DOES run cooler, Last time i own a AMD GPU was a R9 390 that thing was running 20C hotter than my Nvidia GPU.

That's why i have said i don't like AMD GPU's purely of the fact they run hotter, I will never own another AMD GPU ever again.

I have never been happy with a AMD GPU's compared to Nvidia GPU's.

Well one mans gold is another mans lump of coal, but for me i plan getting 2 vegas, at best i get 2 cards running together, at worst the main card can nab the hbm from the other one.

 

[AngryGoldfish]

Well one mans gold is another mans lump of coal, but for me i plan getting 2 vegas, at best i get 2 cards running together, at worst the main card can nab the hbm from the other one.

I wonder whether it would be worth waiting for AMD's flagship, the Vega 10x2 or whatever it's going to be called. It's highly likely they'll make a dual GPU card. The only issue will be the price. For some reason AMD charge an arm and a leg for their dual GPU cards.

 

[Shingara]

I wonder whether it would be worth waiting for AMD's flagship, the Vega 10x2 or whatever it's going to be called. It's highly likely they'll make a dual GPU card. The only issue will be the price. For some reason AMD charge an arm and a leg for their dual GPU cards.

That is a good question, but im sort of pidgeon holed into the fact my ryzen build is waiting for a vega not the vega.

Edit, plus we arnt held to the old constraints which were vram, before now if you had 2 cards and only 1 card was being used, the vram on the second card was a paperweight, with vega it doesnt matter if the vram is on one card, 2 cards or even on an m.2.

 

[AngryGoldfish]

That is a good question, but im sort of pidgeon holed into the fact my ryzen build is waiting for a vega not the vega.

Edit, plus we arnt held to the old constraints which were vram, before now if you had 2 cards and only 1 card was being used, the vram on the second card was a paperweight, with vega it doesnt matter if the vram is on one card, 2 cards or even on an m.2.

Hopefully that theory becomes a reality once Vega is released and tested. 8GB is still enough for most 4k gaming, but having potentially 512TB to utilise can't go amiss—though I do hope they didn't allocate too much attention on the die to HBC over other hardware features.

 

[Shingara]

Hopefully that theory becomes a reality once Vega is released and tested. 8GB is still enough for most 4k gaming, but having potentially 512TB to utilise can't go amiss—though I do hope they didn't allocate too much attention on the die to HBC over other hardware features.

Well, this is something that goes amiss alot of the time i think but when i game im not only gaming, i have multiple screens, im always doing stuff on other screens, that eats up vram.

And the 512TB ye, god knows what we need 512TB's for but im sure that there are plenty of people on here who will try there damned hardest to try and max out 512TB on something

And going forwards if we are in a position where game developers just assume we are running over 10GB of vram for ultra settings then im sure they wont be shy in throwing all those polys about.

 

[Avet]

Vega doesn't need to beat 1080Ti. It doesn't need to be the fastest card on the planet. It needs to perform better than 1060/1070/1080 for less money. And when i say better it means to be competitive with power consumption, and temperatures with same or better performance than mentioned cards for less money. Because as we can see on Steam data thingy 1080p gaming is still by far the most used resolution in gaming. 4K and VR are just overhyped. Very few people play on 4K compared to the whole gaming industy.

It basically needs to do what Ryzen 7 has done to X99. It has beaten them at the targeted consumer segment (workstation PC). It performs around the same +/- 10%, it doesn't matter really, for HALF THE PRICE! No one would pay 2x more money for same or 10% increase in performance. Yes there are people who would, and who can. But would you? Still that is the very, very small percent of the targeted market.

AMD just needs to sort out availability, BIOS, RAM, and the rest of the mess. But they will take a very big peace of intel's cookie.

If Vega does the same to nVidia... We have glory days ahead of us.

 

[Shingara]

Dont take that steam data as golden, that stuff is really doing my nut in, the thing is optional, only 4.7% of people in it have cpus at or over 3.7ghzs. The majority of people on there are running at 2.6 ghz, let that sink in.

 

[Avet]

Either way, what I have said stands. If Vega offers optimized cards that perform on the pair with competition for less money it is a big win. Previous Radeon cards were furnaces that just gobbled power, and they were not good value for the money, because for a little more cash you got much better product from the green team.