• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD VEGA confirmed for 2017 H1

Status
Not open for further replies.
Nvidia wont have the efficiency they have now once they actually start putting hardware on their cards that use the hardware features of DX12 and Vulkan etc.

AMD are paying this penalty right now, Nvidia have brute forced their way so far and it shows in DX12 / Vulkan performance where AMD Cards hardware is getting fully utilised and closes the gaps on Nvidia.

AMD cards still power these bits of hardware in Dx11 where it goes mostly unused, hence the bigger power draws.

This is all coming for Nvidia if they start adding Hardware to their cards to fully utilise newer API features, of course they could still use a Brute force and software approach, but i believe hardware approach will always eventually win out in this type of scenario.
 
Efficiency has always been AMDs biggest problem and still is. However, as far as their own products are concerned they are starting to do better. If Vega is 10% faster than a 1080 then that will mean that it is 15% slower than a Titan XP.....at the right price it will sell by the bucketload even if it takes more power to do so. Again depends on the price/performance and we will not know that for a while yet. I dont think it will be 50% more power to get that 10% performance over the 1080...maybe 20-30%.

Obviously with only a 30% market share AMD will not get enough profit on desktops alone, but they do also have the console contracts and depending on Vega's price/performance that marketshare may (probably will) increase again. AMD continue to increase efficiency in their own products even though it isnt as good as Nvidia's and that is still a good thing. :)

Its not always the case. the 290X uses about the same amount of power for about the same performance as the 780TI.

The Fury-Nano uses about the same amount of power as the GTX 980 for a bit more performance.

And when your talking about performance per watt you can't simply ignore the levels of performance AMD get from high feature level DX12 and Vulkan, the RX 480 uses nearly as much power as the 1070 but when the API feature level suits it its also nearly as fast.

AMD's problem isn't so much a lack of performance per watt, their problem is building full fat architectures. those architectural features cost power, those are architectural features Nvidia don't have in their GPU's.
Power to performance in old API's is not the be all and end all.

Banging on about power this and power that instead of performance is the very reason Nvidia strip their GPU's down to the bare bones, its the reason my GPU doesn't do DX12 beyond DX11.3 feature level while a much older 290X stomps all over it.
 
Last edited:
Nvidia wont have the efficiency they have now once they actually start putting hardware on their cards that use the hardware features of DX12 and Vulkan etc.

AMD are paying this penalty right now, Nvidia have brute forced their way so far and it shows in DX12 / Vulkan performance where AMD Cards hardware is getting fully utilised and closes the gaps on Nvidia.

AMD cards still power these bits of hardware in Dx11 where it goes mostly unused, hence the bigger power draws.

This is all coming for Nvidia if they start adding Hardware to their cards to fully utilise newer API features, of course they could still use a Brute force and software approach, but i believe hardware approach will always eventually win out in this type of scenario.

NV are relying on massive clocks to keep ahead of AMD, 1060 v 480.
 
Efficiency has always been AMDs biggest problem and still is. However, as far as their own products are concerned they are starting to do better. If Vega is 10% faster than a 1080 then that will mean that it is 15% slower than a Titan XP.....at the right price it will sell by the bucketload even if it takes more power to do so. Again depends on the price/performance and we will not know that for a while yet. I dont think it will be 50% more power to get that 10% performance over the 1080...maybe 20-30%.

Obviously with only a 30% market share AMD will not get enough profit on desktops alone, but they do also have the console contracts and depending on Vega's price/performance that marketshare may (probably will) increase again. AMD continue to increase efficiency in their own products even though it isnt as good as Nvidia's and that is still a good thing. :)

Laptop profit>>>>>Console peanut profit.
 
I might be wrong, probably, but Rx480 is 1060 competitor. 1060 uses 50w less than 480 so 1080 competitor, vega, why couldnt that be equal but use just 50w more, so 220w. I would be fine with that.
 
I had fury X and i felt it lacks in high end product features compare to GTX 980 Ti with better features at the same price.

AMD cards disparately need efficiency for high end GPUs due to that it will able to get good overclocking and it will be much cooler. The reason today AMD has no high end GPU market is because AMD PR and AMD management thinks only Raw power is the key feature and requirement for a successful high end product. I really hope they learned something from Fury X launch or else VEGA will follow same path as fury X.
 

Read what i said above, Nvidia have the edge *Right Now* as they design their hardware for the here and now, Pascal is still a DX11 architecture, AMD gambled and built for the future.

Once Nvidia move to a proper DX12 / Vulkan architecture on their GPU's they will have to add hardware to keep the performance up, this will require more power and generate more heat.

AMD gambled on stuff like DX12 and Vulkan becoming mainstream earlier than it has, they tried to force the movement forward with Mantle and to a degree have achieved this.

Once more and more software arrives to take use of this DX12 type of architecture, unless Nvidia add similar hardware they will fall behind, drivers and software will only take you so far, the API's are designed to talk directly to the hardware, and if its not there to talk to, then how do you leverage the performance?

Like i have been saying, AMD have been paying the cost on power consumption for a long time as their GCN hardware has never been fully utilized in DX11 yet the hardware requires powering for the cards to function, this generates heat etc.

You only have to look at the 1060 vs 480, or the other AMD cards, in DX11 many of them fall behind Maxwell and Pascal etc, but in DX12 and Vulkan a lot of them punch much further above their weight, its because the software utilizes this dormant hardware.

The real question is this, will Nvidia get out DX12 hardware before it becomes mainstream? because if they do not, they will be the ones playing catchup as the AMD hardware pulls significantly ahead.
 
Read what i said above, Nvidia have the edge *Right Now* as they design their hardware for the here and now, Pascal is still a DX11 architecture, AMD gambled and built for the future.

Once Nvidia move to a proper DX12 / Vulkan architecture on their GPU's they will have to add hardware to keep the performance up, this will require more power and generate more heat.

AMD gambled on stuff like DX12 and Vulkan becoming mainstream earlier than it has, they tried to force the movement forward with Mantle and to a degree have achieved this.

Once more and more software arrives to take use of this DX12 type of architecture, unless Nvidia add similar hardware they will fall behind, drivers and software will only take you so far, the API's are designed to talk directly to the hardware, and if its not there to talk to, then how do you leverage the performance?

Like i have been saying, AMD have been paying the cost on power consumption for a long time as their GCN hardware has never been fully utilized in DX11 yet the hardware requires powering for the cards to function, this generates heat etc.

You only have to look at the 1060 vs 480, or the other AMD cards, in DX11 many of them fall behind Maxwell and Pascal etc, but in DX12 and Vulkan a lot of them punch much further above their weight, its because the software utilizes this dormant hardware.

The real question is this, will Nvidia get out DX12 hardware before it becomes mainstream? because if they do not, they will be the ones playing catchup as the AMD hardware pulls significantly ahead.

Nvidia had edge since maxwell (2014). Now on 14nm AMD only manage get on par with Nvidia while Nvidia was on 28nm. It only shows that how much years and how far AMD is behind.
 
i'd never build another SLi nvidia until their arch is full DX12 supported via hardware, where as RX 480 and hopefully vega, would do it in a heart beat. seems by the looks of it i could easily leave it in the shed for a year and with DX12 fully being used and hopefully MS getting their Uni app sorted with Gears of War 4 ( make or break for the gaming app i think), I could get it, dust it off and still have amazing performance like it was brand new

and this coming from an Nvidia owner since the GeForce 256
 

So you ignore the 1070 and 1080 founders? The latter throttled a lot, and even NVIDIA took flack for showcasing it only running at ~60 degrees OC'd to 2000Mhz; when infact they had the fan running at 100%.

The 1080 runs at 83 degrees and throttles the boost clock speeds. The Titan X does the same.

I've owned several top NVIDIA reference cards, they all run hot and are loud; even more so when you up the fan speeds to compensate for the heat and to stabilize the boost clocks.
 
Last edited:
So you ignore the 1070 and 1080 founders? The latter throttled a lot, and even NVIDIA took flack for showcasing it only running at ~60 degrees OC'd to 2000Mhz; when infact they had the fan running at 100%.

The 1080 runs at 83 degrees and throttles the boost clock speeds. The Titan X does the same.

I've owned several top NVIDIA reference cards, they all run hot and are loud; even more so when you up the fan speeds to compensate for the heat and to stabilize the boost clocks.

yes I ignored them since I was originally replying to a 1060 vs 480 post 370 :confused:
 
yes I ignored them since I was replying to a 1060 vs 480 post :confused:

Not at all; he gave those two cards as an example, as NVIDIA are relying on massive clocks on all their cards.

IPC wise Pascal and Maxwell are extremely similar; in fact a well overclocked 980Ti or TitanX Maxwell easily match the much higher clocked Pascal cards.
 
Status
Not open for further replies.
Back
Top Bottom