Thread replies: 12
Thread images: 2
Thread images: 2
File: amd-vega.jpg (145KB, 1280x720px) Image search:
[Google]
145KB, 1280x720px
>The only confirmed metric we know is an undisclosed Vega GPU running Doom between 60 and 80FPS on Ultra 4K
>This makes it sit (as of now) between the GTX 1080 and the GTX 1080 Ti (http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1080-ti,4972-3.html)
>They just released the RX580 which sits at $200 - $250 and they will likely not touch that price point
>Starting prices will probably be around the 1070 level, so $400
>Nvidia themselves will probably have at least some info about Vega performance
>Volta (at least for consumers) is not due until 2018 so Nvidia would be hard pressed if they didn't have a counter
The most probable case is that Nvidias "answer" to Vega is already released, and it's the 1080 Ti.
Vega performance is right between the 1080 and the 1080 Ti, but the only confirmed case was Vulkan, so we don't know how optimized it is for DX11 yet.
This allows Nvidia to remain the raw performance king, as AMD will probably undercut the 1080 Ti roughly $100 with their top tier model.
Which, in combination with Nvidias marketing department (who don't want to antagonize 1080 Ti buyers) means that there will be no high-end price war. Rather the 1080 Ti and 1080 will stay at their price, and AMD will fit their top tier model right in between. HBM2 is still not cheap after all.
There will probably be a second model at about 1070 price point which could sit between the 1070 and the 1080 performance wise.
Nothing indicates AMD is sandbagging hard and will deliver anything unexpected, but it's going to be a solid release with AMD catching up to >=1070 cards, with the 1080Ti remaining the "raw power" king but the top-tier Vega model (however it will be called) a high-end price-to-performance king.
Thoughts?
>>
>>59998433
Do video cards really need to be "optimized" for running a specific library like DX11? I thought the game is what needs to be optimized
>>
>>59998433
I think you're heavily underestimating this thing, 4096 shaders at 1536+mhz is no joke. That and the news of how they focused on fixing the scaling and pipeline bubbles that Fiji had which killed the performance of the card. What I'm expecting is a balls hair under the 1080 Ti in DX11 and pounding the tuna at 4K and DX12.
>>
When will it get released? I'm planning to buy 2x 1080 in may but might buy 2x Vega if the performance is good.
>>
What's the fucking point of this kind of speculation? 90% of the time it's all wrong anyway. Remember how the rx480 was supposed to be faster than the 1070? Or how the 1080ti was supposed to have 16gb HBM2?
>>
>>59999642
Who said that outside of wccf poo in loos?
There is already solid evidence that at least one SKU of Vega places pretty snugly between the 1080 and the 1080 Ti. Whether this is the most high end SKU or not no one knows, but it is probable.
>>
>>60000065
Care to provide a link? I've been holding off on buying a 1080ti for the off chance of vega being actually competitive with it.
>>
i hope the finally show Vega at Computex
>>
>>60000234
https://www.youtube.com/watch?v=Y8tDaPLHxiE
At 0:50
>>
>>60000234
first heard of vega from google tech news , please look stuff up mane
>>
File: RX580-guru-3d-1.png (732KB, 1364x6707px) Image search:
[Google]
732KB, 1364x6707px
>>59998433
>Vega performance is right between the 1080 and the 1080 Ti, but the only confirmed case was Vulkan, so we don't know how optimized it is for DX11 yet.
That was on Fiji drivers and with the Vega card underclocked. lol.
In ideal cases, some DX12 and Vulkan games and/or games optimized for it like Prey, it'll probably beat the Titan Xp.
In other cases, it'll fall short of the 1080Ti, at least at 1080p, I'm pretty sure. But it may still beat it at 4K even in those worst case examples as GCN has always scaled much better into higher resolutions than any Nvidia cards.
What we know definitively is that if you simply doubled the size of the RX580, that alone would be competitive with the 1080, in many cases Titan X performance. (well, more like 2.5x then clocked lower so it's not a housefire)
But instead of simply being a 2x sized RX580, Vega is a die that's over 3 times larger. Twice the CUs. 2x the triangles drawn per clock per ROP. It's a big architecture change. And HBM2 that's not only going to be good bandwidth, but lower latency, and a greatly improved memory/cache controller.
I'm really expecting it to completely BTFO like how the 7970 BTFO the GTX 580. That's a reasonable expectation.
>>59998777
Games are almost all shittily programmed so the driver has to do extra optimizations.
In a lot of cases with Nvidia, those "optimizations" just plain mean lowering the quality down lower than what the user set.
>>
>>60000976
Nvidia must be shitting bricks rn
Thread posts: 12
Thread images: 2
Thread images: 2