Overclock.net › Forums › Industry News › Hardware News › [Hexus] Review: Sapphire Radeon R9 Fury Nitro
New Posts  All Forums:Forum Nav:

[Hexus] Review: Sapphire Radeon R9 Fury Nitro - Page 6

post #51 of 179
FYI, I have to annotate to those who suppose frame rates in excess of 144 FPS mean anything - they don't really.
The panel in question which is overclocked to 165 Hz is NOT calibrated to function any faster than its 144 Hz mode. In essence, you are rastering faster - that is correct - however the pixels are not switching any faster therefore 165 Hz fails at being instrumental for faster monitor response.
The Machine
(14 items)
 
Nexus 7 2013
(11 items)
 
 
CPUMotherboardGraphicsRAM
A10 6800K Asus F2A85-V MSI 6870 Hawx, VTX3D 5770, AMD HD6950(RIP), Sap... G.skill Ripjaws PC12800 6-8-6-24 
Hard DriveOptical DriveOSMonitor
Seagate 7200.5 1TB NEC 3540 Dvd-Rom Windows 7 x32 Ultimate Samsung P2350 23" 1080p 
PowerCaseMouseAudio
Seasonic s12-600w CoolerMaster Centurion 5 Logitech G600 Auzen X-Fi Raider 
CPUMotherboardGraphicsRAM
Quad Krait 300 at 1.5Ghz Qualcomm APQ8064-1AA SOC Adreno 320 at 400mhz 2GB DDR3L-1600 
Hard DriveOSMonitorKeyboard
32GB Internal NAND Android 5.0 7" 1920X1200 103% sRGB & 572 cd/m2 LTPS IPS Microsoft Wedge Mobile Keyboard 
PowerAudio
3950mAh/15.01mAh Battery Stereo Speakers 
  hide details  
Reply
The Machine
(14 items)
 
Nexus 7 2013
(11 items)
 
 
CPUMotherboardGraphicsRAM
A10 6800K Asus F2A85-V MSI 6870 Hawx, VTX3D 5770, AMD HD6950(RIP), Sap... G.skill Ripjaws PC12800 6-8-6-24 
Hard DriveOptical DriveOSMonitor
Seagate 7200.5 1TB NEC 3540 Dvd-Rom Windows 7 x32 Ultimate Samsung P2350 23" 1080p 
PowerCaseMouseAudio
Seasonic s12-600w CoolerMaster Centurion 5 Logitech G600 Auzen X-Fi Raider 
CPUMotherboardGraphicsRAM
Quad Krait 300 at 1.5Ghz Qualcomm APQ8064-1AA SOC Adreno 320 at 400mhz 2GB DDR3L-1600 
Hard DriveOSMonitorKeyboard
32GB Internal NAND Android 5.0 7" 1920X1200 103% sRGB & 572 cd/m2 LTPS IPS Microsoft Wedge Mobile Keyboard 
PowerAudio
3950mAh/15.01mAh Battery Stereo Speakers 
  hide details  
Reply
post #52 of 179
Quote:
Originally Posted by Slink3Slyde View Post

However we are currently talking about how relevant their benchmarks are towards the Fury. Most people paying 500 dollars or Euros for their GPU would be planning on running as close to ultra settings as possible in all games, I would imagine.

Yes, I'm totally sure they will run Rise of the Tomb Raider at sub 30 FPS average like their latest review is pointing out. Because everyone knows that a stutterfest/slideshow @ highest settings is better than something running fairly nice with a lower IQ.
Quote:
Originally Posted by Slink3Slyde View Post

I'd fully support them re adjusting their settings to high with no AA for example on mid range card comparisons, as it is though it is an easier way of keeping all testing consistent so that they can compare all cards on the same playing field. Otherwise we wouldn't be able to tell exactly how much better the top end cards were all things being equal.

Because they're not equal. You can't ask the same to a 2 GB card than a 6 GB one for example.
Quote:
Originally Posted by Slink3Slyde View Post

I dont think we can appreciate exactly how many hours go into running these tests, running them twice with different settings would be a serious challenge, and I wouldnt want to be the one to do it.

Seriously, it's their bloody job.

I don't know how you can write off having a totally messed up methodology for lower end cards. At the end of the day their methodology is still wrong.
post #53 of 179
Quote:
Originally Posted by Slink3Slyde View Post

I think I agreed with you already on the mid range and lower cards thing already, that makes sense to me. I can believe how the different cards might react more or less positively to certain features being disabled, its something I'm sure we've all done when trying to run a new game on a mid range or slightly outdated card.

However, they do disable hairworks in TW 3 for example, run GTA 5 with MSAA off, they dont just run everything maxed with no regard, if there's a setting that kills everything they appear to recognize and compensate for that in some way.

Nowhere is going to be perfect. This is the internet theres always going to be a way to pick holes, but I still feel that their results are on a level playing field for all cards, and as such are a good gauge of relative performance.

 

I'm moving back into hardware benchmarking, for the site I write for, now that I have a permanent testing platform on the way, and this is something I'm going to include in my results. If I find that there's something unfairly affecting the performance of one of the cards, I'll mention it and provide test results that show what the performance is like when things are fixed. As it is, I'm already going to have to test GTA V again with my Radeon R7 265 to make sure the MSAA bug is well and truly gone. 

 

And this also means that I won't be testing something idiotic like a GTX 960 at 4K. It doesn't make sense and it wastes everyone's time. I won't be testing as many cards as W1zzard does at a time, so I don't have to accept benchmark results that unfairly mark down cheaper cards because I won't be doing dozens of tests for each one. This is why Anandtech's iGPU testing is done at 720p where applicable.

 

Quote:
Originally Posted by Imouto View Post

After getting to know that PCPer doesn't know how to interpret their own data and people at these forums are fine with that everything is possible.

 

Is this a known fact?


Edited by CataclysmZA - 1/31/16 at 11:13am
post #54 of 179
Quote:
Originally Posted by CataclysmZA View Post

Is this a known fact?

http://www.overclock.net/t/1564303/various-amd-r9-fury-reviews/320#post_24155285
post #55 of 179

 

So...

 

Fury X Crossfire 39 / 20 = 1.95 ~ 95% scaling

Fury X  3x Crossfire 57 / 39 = 1.46 ~ 46% scaling

 

GTX 980 Ti SLI 35 / 19 = 1.84 ~ 84% scaling

GTX 980 Ti 3 x SLI 51 / 35 = 1.4571 ~ 45% scaling

 

PCPer does round down, as Ryan as admitted on the podcast before, but the math works out. Scaling behaves as expected when you add on a third or even fourth card, it's always an improvement that equals about half of the gain you got from the previous card. 

 

You're not calculating the percentage change between the results, you're trying to match those results up with how you think the scaling should work. You're the one not interpreting the data here correctly.


Edited by CataclysmZA - 1/31/16 at 11:17am
post #56 of 179
Quote:
Originally Posted by CataclysmZA View Post

So...

Fury X Crossfire 39 / 20 = 1.95 ~ 95% scaling
Fury X Crossfire 57 / 39 = 1.46 ~ 46% scaling

GTX 980 Ti SLI 35 / 19 = 1.84 ~ 84% scaling
GTX 980 Ti SLI 51 / 35 = 1.4571 ~ 45% scaling

PCPer does round down, as Ryan as admitted on the podcast before, but the math works out. Scaling behaves as expected when you add on a third or even fourth card, it's always an improvement that equals about half of the gain you got from the previous card. 

You're not calculating the percentage change between the results, you're trying to match those results up with how you think the scaling should work. You're the one not interpreting the data here correctly.

Nah its stupid. When you add a 3rd card you dont expect the double performance of 2 cards.
Workstation
(4 items)
 
  
CPUMotherboardGraphicsMonitor
Xeon E5-2690 Supermicro 2011 Nvidia GP100/ Vega FE Dell ultrasharp 4k 
  hide details  
Reply
Workstation
(4 items)
 
  
CPUMotherboardGraphicsMonitor
Xeon E5-2690 Supermicro 2011 Nvidia GP100/ Vega FE Dell ultrasharp 4k 
  hide details  
Reply
post #57 of 179
Quote:
Originally Posted by sugarhell View Post


Nah its stupid. When you add a 3rd card you dont expect the double performance of 2 cards.

 

Well, yeah, because that would be impossible. Scoring the scaling benefits of multi-GPU arrays against a single card seems a bit odd to me. You're dividing the available work between the GPUs, the CPU has to work harder to keep them fed, latency in the display chain gets injected at multiple points now, and at some point adding on another card makes little sense for the performance boost. 

 

I can see Imotou's point, but it's a confusing way to look at the data. Perfect scaling is probably never going to happen for the majority of games and software tested.


Edited by CataclysmZA - 1/31/16 at 11:32am
post #58 of 179
Quote:
Originally Posted by Imouto View Post

Yes, I'm totally sure they will run Rise of the Tomb Raider at sub 30 FPS average like their latest review is pointing out. Because everyone knows that a stutterfest/slideshow @ highest settings is better than something running fairly nice with a lower IQ.
Because they're not equal. You can't ask the same to a 2 GB card than a 6 GB one for example.
Seriously, it's their bloody job.

I don't know how you can write off having a totally messed up methodology for lower end cards. At the end of the day their methodology is still wrong.

I didnt write it off, I agreed with you and I stated thats it's irrelevant when we are talking about how the Fury matches up with the 980ti. Which apparently as far as price performance goes is quite well in relation to a reference 980ti, particlularly in 4k.

It is indeed their job, journalists often have tight deadlines. I'm amazed that they manage to do as much as they do sometimes, of course if they decided to do more it would be great. As it is they keep things consistent, and it is a fair gauge of performance in that the testing is consistent. Places like Hard OCP will provide the sort of IQ setting comparisons you seek, its best not to just take one places results as gospel in pretty much everything in my experience.
Quote:
Originally Posted by CataclysmZA View Post

I'm moving back into hardware benchmarking, for the site I write for, now that I have a permanent testing platform on the way, and this is something I'm going to include in my results. If I find that there's something unfairly affecting the performance of one of the cards, I'll mention it and provide test results that show what the performance is like when things are fixed. As it is, I'm already going to have to test GTA V again with my Radeon R7 265 to make sure the MSAA bug is well and truly gone. 

And this also means that I won't be testing something idiotic like a GTX 960 at 4K. It doesn't make sense and it wastes everyone's time. I won't be testing as many cards as W1zzard does at a time, so I don't have to accept benchmark results that unfairly mark down cheaper cards because I won't be doing dozens of tests for each one. This is why Anandtech's iGPU testing is done at 720p where applicable.

I'll look out for your reviews, as you say when you're testing a lot less cards its much easier to look into things more closely. thumb.gif
post #59 of 179
Quote:
Originally Posted by CataclysmZA View Post

PCPer does round down, as Ryan as admitted on the podcast before, but the math works out. Scaling behaves as expected when you add on a third or even fourth card, it's always an improvement that equals about half of the gain you got from the previous card.

It is not rounding, it's plain bad math. He is taking as base an aftermath, the outcome of adding two factors.

Let's check his results, ok?

GTX 980 Ti = 19 Single -> 35 SLi = 84% scaling (from a laughable 1.84~ to 84% conversion).

He just divided 35 by 19 so he's considering 35 (ie, the aftermath of a previous operation) as base for his calculations. Now look what he is presenting instead of scaling:

(19*1) + (19*0.84~) = 19 + 16 = 35

He's doing it backwards and considering the SLI/CF results as 100% and then picking whatever he wants because he doesn't know what he's doing.
Quote:
Originally Posted by CataclysmZA View Post

You're not calculating the percentage change between the results, you're trying to match those results up with how you think the scaling should work. You're the one not interpreting the data here correctly.

There's no other way to interpret these data. You can't pick whatever you want and the easier and only one with any academic worth in this case is the base. So with each set of data you're looking at a different target for perfect scaling. First 100%, then 50%, then 33%, then ... Excuse me but that's bollocks.

Presenting data like Ryan did for any kind of study would get him instantly fired in any field. If you look at any study with a similar set of data I guarantee you that you won't find any resembling what Ryan did. Wrong to boot.

Imagine that you're working with batteries, photovoltaic panels or whatever instead of graphic cards. It doesn't make any sense to present the data like he did.

Bad math and not knowing what he's doing I say.
Edited by Imouto - 1/31/16 at 12:31pm
post #60 of 179
so a single 980ti has 19fps, then adding another is +16 fps gain. Then another card adds another 16 fps but this time scaling is interpreted as 45%. Funny.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Hexus] Review: Sapphire Radeon R9 Fury Nitro