Regarding the RX 480 AOTS benchmark at Computex

[Indra Adha]

16 hours ago, AmbarChakrabarti said:

 

$400 GPU setup beats a $600 GPU! Holy Balls!

$400? is it 4gb version (rx 480)?

[Big-G]

9 minutes ago, Indra Adha said:

$400? is it 4gb version (rx 480)?

He wishes... the AMD slide was mentioning sub $500... that would be the 8GB items...

[Indra Adha]

16 hours ago, AmbarChakrabarti said:

 

$400 GPU setup beats a $600 GPU! Holy Balls!

$400? is it 4gb version (rx 480)?

 

2 minutes ago, GidonsClaw said:

He wishes... the AMD slide was mentioning sub $500... that would be the 8GB items...

if i play at 1080p,the 8gb version worth?

[DXMember]

25 minutes ago, GidonsClaw said:

The thing is, it's well known... the only thing Nvidia guarantees is the base clock, boost clock is an estimation, but GPU Boost will always try and get the best clock range it can get within limits. So technically it is not cheating... you are getting the best you can get...

 

AMD on the other hand advertises a boost clock, but generally the card will throttle down from that to match the environment. So essentially most AMD owners always get less than they thought...

first of all we were talking about the snow

second of all the GPU Boost is pretty much a benchmark crusher as under heavy gaming loads the boost clock at least on the founders cards pretty much settles down to base after 15~20mins so it does kind of falsely advertise the card when people base their opinions on the public benchmarks - especially when being on boost or on base clock is the difference between being at 60fps or below

 

I would also like to have some clarification on the Hallocs clarification of the AotS benchmark - he noted only one FPS figure for the Single batch but he gave GPU utilization for all three Single, Medium and Heavy batch - I would expect individual FPS numbers for those batches as well from the AotS benchmark... what's up with that?

[Big-G]

2 minutes ago, Indra Adha said:

$400? is it 4gb version (rx 480)?

 

if i play at 1080p,the 8gb version worth?

Na, $200 4GB should be fine

[BOT Zee]

16 hours ago, Fulgrim said:

Not the ones used at computex, but here's some benchmarks

RX 480 Crossfire

http://www.ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/ac88258f-4541-408e-8234-f9e96febe303

GTX 1080

http://www.ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/a957db0f-59b3-4394-84cc-2ba0170ab699

 

Those benchmarks literally prove the bullshit of AMD looool the 1080 does WAY BETTER! 95FPS VS 72FPS IN THE AVERAGE LOL SO WHY IS THAT? WHEN THEY BOTH ARE USING THE SAME CPU? THOSE BENCHMARKS are not in AMD's favor.... if a 400$ solution has equivalent fps/$ as a 600$ gpu then no one cares and AMD lost once again....

[BOT Zee]

2 minutes ago, DXMember said:

first of all we were talking about the snow

second of all the GPU Boost is pretty much a benchmark crusher as under heavy gaming loads the boost clock at least on the founders cards pretty much settles down to base after 15~20mins so it does kind of falsely advertise the card when people base their opinions on the public benchmarks - especially when being on boost or on base clock is the difference between being at 60fps or below

 

I would also like to have some clarification on the Hallocs clarification of the AotS benchmark - he noted only one FPS figure for the Single batch but he gave GPU utilization for all three Single, Medium and Heavy batch - I would expect individual FPS numbers for those batches as well from the AotS benchmark... what's up with that?

No the boost clock does not infact settle down after 15-20mins ahahaha it does if your fan is set to 30% mind you this is talking about the founders edition which youd be an idiot to buy anyways, if you set the fan higher like 60-70% on founders itll maintain a 2ghz oc indefinately. Okay so say you dont want fans that high... yeah buy an aftermarket gpu which is cooler and quieter and can DEFINITELY MAINTAIN THE OC.

[DXMember]

8 minutes ago, BOT Zee said:

No the boost clock does not infact settle down after 15-20mins ahahaha it does if your fan is set to 30% mind you this is talking about the founders edition which youd be an idiot to buy anyways, if you set the fan higher like 60-70% on founders itll maintain a 2ghz oc indefinately. Okay so say you dont want fans that high... yeah buy an aftermarket gpu which is cooler and quieter and can DEFINITELY MAINTAIN THE OC.

so you do agree that the FE benchmarks and reviews recommending it are a hoax after all?

[DXMember]

13 minutes ago, BOT Zee said:

Those benchmarks literally prove the bullshit of AMD looool the 1080 does WAY BETTER! 95FPS VS 72FPS IN THE AVERAGE LOL SO WHY IS THAT? WHEN THEY BOTH ARE USING THE SAME CPU? THOSE BENCHMARKS are not in AMD's favor.... if a 400$ solution has equivalent fps/$ as a 600$ gpu then no one cares and AMD lost once again....

the CPU framerate doesn't matter as it's not displayed on the screen - that metric is there to help understand whether the system is bottlenecked by the CPU or not, that's all

[BOT Zee]

3 minutes ago, DXMember said:

the CPU framerate doesn't matter as it's not displayed on the screen - that metric is there to help understand whether the system is bottlenecked by the CPU or not, that's all

Yeah but like i said WHY DOES IT VARY BETWEEN THE TWO SYSTEMS IF THE ONLY DIFFERENCE IS THE CARDS??? This doesnt look good for AMD idc if that number is cpu only i see that it is but they are both using the exact same CPU SO why is it different? Invalid test? Maybe? 

[patrickjp93]

Sorry but believing an AMD representative in this matter is worthless.

[DXMember]

1 minute ago, patrickjp93 said:

Sorry but believing an AMD representative in this matter is worthless.

I believe we got these:

RX 480 Crossfire

http://www.ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/ac88258f-4541-408e-8234-f9e96febe303

GTX 1080

http://www.ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/a957db0f-59b3-4394-84cc-2ba0170ab699

[DXMember]

2 minutes ago, BOT Zee said:

Yeah but like i said WHY DOES IT VARY BETWEEN THE TWO SYSTEMS IF THE ONLY DIFFERENCE IS THE CARDS??? This doesnt look good for AMD idc if that number is cpu only i see that it is but they are both using the exact same CPU SO why is it different? Invalid test? Maybe? 

well first of all they apparently have different code paths in the Nitros engine

[BOT Zee]

1 minute ago, DXMember said:

well first of all they apparently have different code paths in the Nitros engine

AKA AMD's "code path" is inferior....

[DXMember]

1 minute ago, BOT Zee said:

AKA AMD's "code path" is inferior....

you can't say that - you're not in a position to judge objectively

[patrickjp93]

40 minutes ago, DXMember said:

I believe we got these:

RX 480 Crossfire

http://www.ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/ac88258f-4541-408e-8234-f9e96febe303

GTX 1080

http://www.ashesofthesingularity.com/metaverse#/personas/b0db0294-8cab-4399-8815-f956a670b68f/match-details/a957db0f-59b3-4394-84cc-2ba0170ab699

Tells us nothing about who is rendering the terrain shader correctly. I want in-sequence debug logs.

[DXMember]

30 minutes ago, patrickjp93 said:

Tells us nothing about who is rendering the terrain shader correctly. I want in-sequence debug logs.

heh, closes you'll get is a side by side video replay of the benchmark

 

however what it does show is that the game engine settings were set to equal and from the video demo we did in fact see that nVidia run appeared to have sharper shadows and some terrain LoD pops in one of the scenes where it zoomed out, however it was hard to judge if the terrain looked different because of softer shadows on AMD or because of actually more snow

[patrickjp93]

42 minutes ago, DXMember said:

heh, closes you'll get is a side by side video replay of the benchmark

 

however what it does show is that the game engine settings were set to equal and from the video demo we did in fact see that nVidia run appeared to have sharper shadows and some terrain LoD pops in one of the scenes where it zoomed out, however it was hard to judge if the terrain looked different because of softer shadows on AMD or because of actually more snow

The settings part is doctor-able. Scrubbing video isn't tough to do anymore, especially with text.

[Prysin]

34 minutes ago, patrickjp93 said:

The settings part is doctor-able. Scrubbing video isn't tough to do anymore, especially with text.

lol, you so want the 1080 to win don't you?

there is nothing out of the extraodinairy with the 2x 480 >= 1080.

 

Facts we know.

1080 is around 5-15% faster then a Titan X in normal games.

R9 295x2 is around 5-15% faster then a Titan X in normal games (assuming proper CF profile)

RX 480 is estimated around R9 390X/GTX 980 perf.

R9 295x2 is roughly equal to the projected performance of the RX 480

 

2+2 = 4

 

I dont need any more evidence, because simple math tells us that the performance seen is within reasonable probability 

[GlassBomb]

56 minutes ago, Prysin said:

lol, you so want the 1080 to win don't you?

there is nothing out of the extraodinairy with the 2x 480 >= 1080.

 

Facts we know.

1080 is around 5-15% faster then a Titan X in normal games.

R9 295x2 is around 5-15% faster then a Titan X in normal games (assuming proper CF profile)

RX 480 is estimated around R9 390X/GTX 980 perf.

R9 295x2 is roughly equal to the projected performance of the RX 480

 

2+2 = 4

 

I dont need any more evidence, because simple math tells us that the performance seen is within reasonable probability 

It's all just speculations coming from each and every corner thus far. As all that we have to go for is those slides and whatever.

 

I am personally guessing that the RX 480 will be slightly faster than a GTX 970, or a R9 3-/290. Simply for the fact that it's priced within reason of those cards, and because they are touting it as a "VR-Ready" card. And the prior mentioned cards it eh minimum requirements that Vive has.

 

And if you compare them to the numbers that you have thrown out, then it sort of fits.

 

Edit: But waiting for solid numbers from actual people is the better thing to do. So far the only new card that's available to Average Joe is the GTX 1080, and nothing else.

[3DOSH]

(+5 tflops) + (+5 tflops)= +10 tflops 

+10 tflops > 9 tflops= Gtx 1080

 

(+10flops) - (Crossfire scaling) ~ +9 tflops 

Fuck me, that makes no sense!

[Citadelen]

2 minutes ago, 3DOSH said:

-snip-

The Founders Edition can't reach 9Tflops, and the RX 480 pushes more than five. So take that into account. Also, explicit multi-GPU was used, which is vastly superior to CrossFireX.

[Trixanity]

4 hours ago, BOT Zee said:

Yeah but like i said WHY DOES IT VARY BETWEEN THE TWO SYSTEMS IF THE ONLY DIFFERENCE IS THE CARDS??? This doesnt look good for AMD idc if that number is cpu only i see that it is but they are both using the exact same CPU SO why is it different? Invalid test? Maybe? 

Most likely culprit is that AMD is running multi-GPU meaning that the CPU needs to feed two GPUs hence the CPU framerate drops. That's also why the multi-GPU scaling improves with the "heavy" batches. It simply puts more strain on the GPUs which eases the CPU bottleneck.

4 hours ago, patrickjp93 said:

Sorry but believing an AMD representative in this matter is worthless.

And believing an Intel representative isn't?

4 hours ago, DXMember said:

first of all we were talking about the snow

second of all the GPU Boost is pretty much a benchmark crusher as under heavy gaming loads the boost clock at least on the founders cards pretty much settles down to base after 15~20mins so it does kind of falsely advertise the card when people base their opinions on the public benchmarks - especially when being on boost or on base clock is the difference between being at 60fps or below

 

I would also like to have some clarification on the Hallocs clarification of the AotS benchmark - he noted only one FPS figure for the Single batch but he gave GPU utilization for all three Single, Medium and Heavy batch - I would expect individual FPS numbers for those batches as well from the AotS benchmark... what's up with that?

You could probably ask him but there are FPS numbers in the wild eg. the 1440p Crazy preset online benchmarks and if necessary, try to extrapolate from there.

[xAcid9]

3 minutes ago, Trixanity said:

That's also why the multi-GPU scaling improves with the "heavy" batches. It simply puts more strain on the GPUs which eases the CPU bottleneck.

It's the other way around, heavy batches mean more drawcalls, more CPU intensive.

[Pardeep01]

Sorry if this is a stupid question but may I know what is AOTS?