First gaming benchmarks up

pcgeekesq

[H]ard|Gawd
Joined
Apr 23, 2012
Messages
1,399
Oh, it's just grainy photos from that article in that Polish (I think) magazine. Old news.

Anyone know if they are even a credible source? Maybe they just made up some numbers to sell their rag ...
 

harmattan

Supreme [H]ardness
Joined
Feb 11, 2008
Messages
4,688
Gibbo over at OCUK just stated (this morning) 480 is categorically slower than GTX 980 except perhaps in DX12, and it's close to 970.
 

Reality

[H]ard|Gawd
Joined
Feb 16, 2003
Messages
1,912
Gibbo over at OCUK just stated (this morning) 480 is categorically slower than GTX 980 except perhaps in DX12, and it's close to 970.

ouch

not necessarily that the card is bad but once again, everyone had their hopes wayyyyyyyyyyy too high for an AMD card

AMD generally lets people down, whereas Nvidia cards lately have been as good or better than expected
 

harmattan

Supreme [H]ardness
Joined
Feb 11, 2008
Messages
4,688
Yea, right? What most of us level-headed kids have thought all along.

Anyhoo, here's the first real review.


Gets beat in most benches and games by a 390, toasted by an overclocked 970. Overclocking is crap: 1.5% top out on this sample with major heat issues.
 

Revdarian

2[H]4U
Joined
Aug 16, 2010
Messages
2,613
Wow, what a mixed bag... i am seriously wondering what determines numbers like for example the Fallout 4 ones, i mean, it usually is slightly behind or on par with the 390, but on FO4 it goes way past it by a great deal. The performance is perfectly nice for 200$ tbh, of course that if you already had a 390-970+ this is a sidegrade at best, but it was widely known since a bit ago.
 

harmattan

Supreme [H]ardness
Joined
Feb 11, 2008
Messages
4,688
Haven't been able to pour over those results (a lot of games there), but just guessing 390 is doing a better job in games where traditional AA was turned on.

Yep, looks like a sidegrade at most for anyone with a 970/390/390x/290x (and maybe a downgrade even if the overclocking results are true for most chips). Also, looks like Kyle may have been right on the heat issues.
 

5150Joker

Supreme [H]ardness
Joined
Aug 1, 2005
Messages
4,569
Wow, what a mixed bag... i am seriously wondering what determines numbers like for example the Fallout 4 ones, i mean, it usually is slightly behind or on par with the 390, but on FO4 it goes way past it by a great deal. The performance is perfectly nice for 200$ tbh, of course that if you already had a 390-970+ this is a sidegrade at best, but it was widely known since a bit ago.

I expected it to at least match the gtx 980 but it just got pummeled in most games. Even the 390 beat it in most cases. And the temperatures of that thing, especially when a tiny OC is applied is beyond abysmal. I'm curious to see what nvidia does with 1060s release schedule but based on these results I'd say they don't have much to worry about.
 

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,594
Wow, what a mixed bag... i am seriously wondering what determines numbers like for example the Fallout 4 ones, i mean, it usually is slightly behind or on par with the 390, but on FO4 it goes way past it by a great deal. The performance is perfectly nice for 200$ tbh, of course that if you already had a 390-970+ this is a sidegrade at best, but it was widely known since a bit ago.

There isn't such thing as "generic" performance given the complex systems involved.

For Fallout 4 I'd assume it is because of Polaris 10's "enhanced geometry engine" lends itself very well to the Fallout 4 given the work load resulting in much more effective geometry throughput (eg. through better culling). This is also why Nvidia, who has been stronger in this area, also leads. It is also likely that color compression is comparatively rather effective with Fallout 4 meaning that the raw memory bandwidth advantage of Hawaii is less at play. You can contrast this with say AoTS which from what I've heard favors raw bandwidth over color compression.

This is really just rather fuzzy guess work though. In order to actually explore this would be much more complex.
 

Revdarian

2[H]4U
Joined
Aug 16, 2010
Messages
2,613
With all the new effects and techniques gaming has become absurdly complicated >.< >.< >.<
 

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,594
With all the new effects and techniques gaming has become absurdly complicated >.< >.< >.<

It's actually always been this way. That is why one singular benchmark cannot be representative and never has been.

Even in a single game you can have performance variance due to the given scene being asked to be rendered.
 

Revdarian

2[H]4U
Joined
Aug 16, 2010
Messages
2,613
It's actually always been this way. That is why one singular benchmark cannot be representative and never has been.

Even in a single game you can have performance variance due to the given scene being asked to be rendered.

I know but this is a derivative Chip, in theory performance should have been a little more inline, but like you pointed out they did an improvement on their geometry engine so maybe that is what we are seeing. Ohh well, for the price *it is nice*, specially with warranties and the like.
 

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
51,662
How much of a price premium are we looking at with those extremely good coolers?
Dunno yet, and I don't know if AIBs know yet either. Going to take a lot of culling to pull good OCing CPUs for custom SKUs.
 

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,594
I know but this is a derivative Chip, in theory performance should have been a little more inline, but like you pointed out they did an improvement on their geometry engine so maybe that is what we are seeing. Ohh well, for the price *it is nice*, specially with warranties and the like.

I assume the discussion here is against Hawaii, but consider the following -

Is Polaris 10 an iterative step over Hawaii? Or is it Pitcairn?

If we compare against Pitcairn then I believe Polaris 10 has improvements across the board.

But if we compare against Hawaii then even looking just at the higher level specs (such as ROP count, memory bandwidth, etc.) we obviously see trade offs. And these differences are where workload differences are going to expose performance differences.
 

CSI_PC

2[H]4U
Joined
Apr 3, 2016
Messages
2,193
Yea, right? What most of us level-headed kids have thought all along.

Anyhoo, here's the first real review.


Gets beat in most benches and games by a 390, toasted by an overclocked 970. Overclocking is crap: 1.5% top out on this sample with major heat issues.

Yeah interesting spread of results depending upon game, seems to do better with the more recent releases.
But interesting it is slightly slower to the 390 in games such as AoTS, Star Wars:Battlefront, where one would expect it to be a bit stronger than 390.
A big positive is some aspects to games where it was really weak historically such as Fallout 4 due to the lighting/AO and in that game it is 16% faster than 390, although this does not reflect with all older games tested where it can be slower than 390.
A solid card IMO for the price, just a shame it is not consistently good across a broad range of games though.
Cheers
 

CSI_PC

2[H]4U
Joined
Apr 3, 2016
Messages
2,193
I assume the discussion here is against Hawaii, but consider the following -

Is Polaris 10 an iterative step over Hawaii? Or is it Pitcairn?

If we compare against Pitcairn then I believe Polaris 10 has improvements across the board.

But if we compare against Hawaii then even looking just at the higher level specs (such as ROP count, memory bandwidth, etc.) we obviously see trade offs. And these differences are where workload differences are going to expose performance differences.
More applicable to 380x Tonga and some aspects possibly from Fiji (this had slight performance increase in terms of geometry/Tessellation throughput, so probably more development of this for Polaris)
Cheers
 
Joined
Jun 11, 2016
Messages
56
Based on AMD track record, I would expect a 20% boost over the next couple of years due to driver updates. For the other cards, which are EOL, I think most performance has already been squeezed out.

My take is AMD 480 for 1080p, nvidia 1070 for 1440p and 1080 for 4k.
 

trandoanhung1991

[H]ard|Gawd
Joined
Aug 26, 2011
Messages
1,113
Can you ask them if they are binning the chips themselves ?
If these chips were not lottery then AMD was sleeping when they decided on the RX 480 at 1266mhz as default megahertz ...

It's not like they can push the clocks much more on that cooler. 1.5% overclock leading to a 8oC jump, are you kidding me.
 

Dayaks

[H]F Junkie
Joined
Feb 22, 2012
Messages
8,288
It's not like they can push the clocks much more on that cooler. 1.5% overclock leading to a 8oC jump, are you kidding me.

It matters more how fast the fan is spinning at 80*C.

For example my 1080 won't spin up the fan significantly until I get around 80*C but it's silent and has headroom.

Getting close to the safe design temp != large heat output or poor cooler design
 

TheLAWNoob

Limp Gawd
Joined
Jan 10, 2016
Messages
330
It says in the video that 8GB is $239, not $229. I'm expecting "extremely good custom coolers" to push $300.

So much for $199. Looks like it will be mostly 8GB at launch. I didn't even get that hyped up and I'm already bummed.

Waiting for 4GB RX 470.
 

madgun

[H]ard|Gawd
Joined
May 11, 2006
Messages
1,782
Hoping that 1060 GTX will supersede 980s performance combined with decent acoustics. Need to upgrade my GPU for the HTPC setup.
 

harmattan

Supreme [H]ardness
Joined
Feb 11, 2008
Messages
4,688
Nvidia will do that and for less then $200 because they love to sell you premium products at very low price.

Yep, I wouldn't hold my breath (sarcasm understood). I'm guessing we'll see just over gtx 980 performance at $280 for 1060, but who knows. If priced too high, it might make rx 480 more attractive.
 

limitedaccess

Supreme [H]ardness
Joined
May 10, 2010
Messages
7,594
More applicable to 380x Tonga and some aspects possibly from Fiji (this had slight performance increase in terms of geometry/Tessellation throughput, so probably more development of this for Polaris)
Cheers

To clarify I was referring more towards placement (and associated design constraints) than the design/technology itself, in terms of the latter it would make sense to iterate from your most recent (in this case Tonga/Fiji).

Similarly for example Kepler->Mawell->Pascal, but its GK104->GM204->GP104. Nvidia just makes this easier as they consistently addressed the entire stack from top to bottom while AMD did not.
 

trandoanhung1991

[H]ard|Gawd
Joined
Aug 26, 2011
Messages
1,113
Who are you trying to troll here ?
Custom AIB boards those have better cooling ...

Yeah, but then Perf/$ goes out the window with AIB boards.

You have to choose one or the other. And from the looks of it, since rumours state that AIBs are binning themselves, I'm expecting even more expensive custom parts with high overclocks and huge wattage.
 

buttons

2[H]4U
Joined
Oct 12, 2011
Messages
2,052
I am pretty disappointed with the results. I bought a pair of 290x for $200 each -- granted used with no warranty and power hogs... I am curious though what the actual power consumption is under load. I saw a spec sheet claiming 110 watts. my 290s are pulling about 275-300... equal performance for 1/3 the power draw sounds pretty good actually.
 
Top