AMD Radeon HD 7970 3 GB

[Trackr]

Am I the only one who's disappointed here?

It's 8-15% more powerful than the GTX 580.

Seriously?

Only 2000 ALUs.. on 28nm, when they got 1536 on 40nm.

What gives?

 

[the54thvoid]

Nice summary from Anand.

http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/25

The fact of the matter is that since 2008 we’ve become spoiled by AMD’s aggressive pricing. More than anything else the low prices of the Radeon HD 4870 and Radeon HD 5870 made those products superstars thanks to their performance for the price and their undercutting of NVIDIA’s competing cards. The Radeon HD 5870 was definitely fast, but at $379 it was a steal, which is part of the reason prices for it never stabilized at that low a level.

At the same time the 7970 is not the 5870. The 5870 relative to both NVIDIA and AMD’s previous generation video cards was faster on a percentage basis. It was more clearly a next-generation card, and DX11 only helped to seal the deal. Meanwhile if you look at straight averages the 7970 is only around 15-25% faster than the GTX 580 in our tests, with its advantage being highly game dependent. It always wins at 2560 and 1920, but there are some cases where it’s not much of a win. The 7970’s domination of the 6970 is more absolute, but then again the 6970 is a good $200 cheaper at this point in time.

This is regards to its gaming prowess. Anand is very optimistic about GCN for compute and thinks with proper coding it will be a barnstormer.

 

[Benetanegia]

So a 30% increase from HD6970, like I said it would happen based on specs.

I told you so guys. Now where's the people that called me names for pointing out the obvious?

 

[HellasVagabond]

i use an out of the box install of the drivers without any change to settings

Well that's why i asked. A few months back AMD admitted to having used lower IQ in the default settings than they used to in the past so i was wondering if they reverted that to the original state or they left it as is with the 7970 series.

 

[dj-electric]

1200Mhz core clock with air cooling isnt a dream?
look on the OC of HH

(No-one's lookin) *fap* *fap* *fap* *fap* *fap* *fap* *fap*

 

[cdawall]

I don't think the card offers enough of a performance increase. Yes, it beats the 580, as it doesn't trounce it, but it should do - after all the 580 has been out for a year now. I would have expected around 50% better performance than the 580 minimum in all benchmarks, but we don't get that.

Note that PCI-E 3.0 won't make much of a difference to performance, as W1zz stated in an earlier post.

Overclocking is good though and the dual BIOS is a very worthy feature for peace of mind.

Great review as ever, W1zz.

Mature AMD drivers normally boost a few more percents across the board. 50% I don't think has ever happened in a single generation jump most two generation jumps don't offer that.

 

[arnoo1]

Now crossfire that damn thing and see some real numbers xd

 

[Kaotik]

Any chance of double checking the numbers?

I can see how 590 would jump over 6990 with different games and drivers, same obviously with 570vs6970, but 5850 going over 6870, both are even VLIW5 chips, simply shouldn't happen.
It seems that whole HD6-lineup performance is lower than it should be.

 

[Benetanegia]

Mature AMD drivers normally boost a few more percents across the board. 50% I don't think has ever happened in a single generation jump most two generation jumps don't offer that.

What is this selective memory or what? Most generation jumps have been well over 50% increase and often times close to 100% increase, especially when a new process is used.

 

[W1zzard]

but 5850 going over 6870, both are even VLIW5 chips, simply shouldn't happen.

hd 5850 is high-end, hd 6870 is mid-range. remember, amd boosted their model numbers for hd 6000 series.

 

[Trackr]

What is this selective memory or what? Most generation jumps have been well over 50% increase and often times close to 100% increase, especially when a new process is used.

Wow, Qubit agrees with this?

You're suggesting the HD 4870 had the performance of an HD 5870 when the latter was released?

How unfathomably absurd.

 

[Benetanegia]

Wow, Qubit agrees with this?

You're suggesting the HD 4870 had the performance of an HD 5870 when the latter was released?

How unfathomably absurd.

Whaaaaaaat?! You need to get some reading skills ASAP.

I said that most generation jumps have a 50% to 100% INCREASE in performance. The HD5870 was ~80% faster than the HD4870 so it falls in that range.

 

[cdawall]

What is this selective memory or what? Most generation jumps have been well over 50% increase and often times close to 100% increase, especially when a new process is used.

Review your facts again


6870->6970 18%
5870->6870 -7%
4890->5870 26%
3870->4870 46%
2900XT->3870 -2%
Overall performance summaries

As I don't see anything that exceeded 50% let alone came close to 100%.

 

[Benetanegia]

Review your facts again


6870->6970 18%
5870->6870 -7%
4890->5870 26%
3870->4870 46%
2900XT->3870 -2%
Overall performance summaries

As I don't see anything that exceeded 50% let alone came close to 100%.

Most of those are not generation jumps.

HD3800 was not a new gen. HD6870 is not a new gen and it's not even the same market segment as the HD5870. HD6970 is the same gen and forms part of the same lineup as the HD6870. Pff I'm sure you can do better.

Plus you need some math skills. For example is the HD3870 is 64% as fast as the HD4870 that means that 100/64*100= 156%. So the HD4870 was 56% faster. On the exact same 55nm process BTW.

EDIT: With the HD4890 vs HD5870, you are mostly correct, except for the failed math, that is. I totally forgot about the HD4890.

 

[qubit]

hd 5850 is high-end, hd 6870 is mid-range. remember, amd boosted their model numbers for hd 6000 series.

"Boosted", I like it. One could say "did an nvidia", too.

 

[cdawall]

Most of those are not generation jumps.

HD3800 was not a new gen. HD6870 is not a new gen and it's not even the same market segment as the HD5870. HD6970 is the same gen and forms part of the same lineup as the HD6870. Pff I'm sure you can do better.

The 5870->6970 still isn't a 50% bud. The 5870->6870 still represents a core change and a lateral performance change between generations. The 6970 later took the top spot without pulling off 100% performance increase.

Plus you need some math skills. For example is the HD3870 is 64% as fast as the HD4870 that means that 100/64*100= 156%. So the HD4870 was 56% faster. On the exact same 55nm process BTW.

I just went off percent difference.

EDIT: With the HD4890 vs HD5870, you are mostly correct, except for the failed math, that is. I totally forgot about the HD4890.

It's not really failed math its just not the way you were phrasing it. If the 5870 is 100% and the 4890 is 74% the 5870 still has 26% higher performance.

 

[badtaylorx]

point being this card is going to be a great one....considering amd blew their wad 3 weeks early and we still get performance this good means nvidia is in for a world of hurt untill 6xx

 

[Benetanegia]

The 5870->6970 still isn't a 50% bud. The 5870->6870 still represents a core change and a lateral performance change between generations. The 6970 later took the top spot without pulling off 100% performance increase.

HD5000 to HD6000 is NOT a new gen, unless you really fall for AMD's (and Nvidia's) tactics. It was also made on the same 40nm process.

With a new process, and all the additional transistors that you can use, a lot more is expected.

It's not really failed math its just not the way you were phrasing it. If the 5870 is 100% and the 4890 is 74% the 5870 still has 26% higher performance.

No, that's failed math, plain and simple. Let's go with round numbers. 50% and 100%. According to you 100 is 50% more than 50, and that's false. 100 is 2x, or double of 50. 100 is also a 100% increase or 200% as much as 50.

50 + 50% == 50 + 1/2 * 50 == 75 != 100

 

[cdawall]

HD5000 to HD6000 is NOT a new gen, unless you really fall for AMD's (and Nvidia's) tactics. It was also made on the same 40nm process.

With a new process, and all the additional transistors that you can use, a lot more is expected.

Was a 65nm althon x2 the same generation as a 65nm phenom?

No, that's failed math, plain and simple. Let's go with round numbers. 50% and 100%. According to you 100 is 50% more than 50, and that's false. 100 is 2x, or double of 50. 100 is also a 100% increase or 200% as much as 50.

50 + 50% == 50 + 1/2 * 50 == 75 != 100

You could call it lazy after work math not to mention the magarita and couple of beers . My point that no generation was a 100% or 2x increase still stands. Also I was using the card at 100% as the starting point meaning card A at 100% and card B at 75% card A has a 25% performance lead while also being 50% faster.

 

[GSquadron]

Only the price ruins this card. You get power for noise so it is normal i think.
I think it was much better than gtx 580, but still it needs to evolve as a 28nm gpu
so would not get dissapointed for being the best single gpu

 

[dj-electric]

People are used to being spoiled with AMD products, always expecting X5 performance for 150$ like its some kinda of freaking magic they can perform

 

[Benetanegia]

Was a 65nm althon x2 the same generation as a 65nm phenom?

Different situation, but mainly yes. many Athlon X2's are the same generation as Phenom. Athlon II is same generation as Phenom II, etc.

You could call it lazy after work math not to mention the magarita and couple of beers . My point that no generation was a 100% or 2x increase still stands. Also I was using the card at 100% as the starting point meaning card A at 100% and card B at 75% card A has a 25% performance lead while also being 50% faster.

If you use the faster card as the starting point, you can only say that card B is x% slower than card A. No matter how you word it, it's not correct to say it's x% faster unless the slower card was used as the baseline.

And not trying to bring you down completely, but 100% would be 33% faster than 75%, not 50%. I suggest a cup of coffee and/or a little bit of sleep.

 

[Kaotik]

hd 5850 is high-end, hd 6870 is mid-range. remember, amd boosted their model numbers for hd 6000 series.

Yes I'm fully aware of this, however I'm also aware that 6870 has been always faster than 5850 (while 6850 has been slower)

Previously the difference has been something around 6-7% in favor of 6870 on average at 1920x1200, and now suddenly 5850 is supposed to be nearly 4% faster at same res?

That's just not right, if 6800's were VLIW4's it could possibly be explained by some driver optimizations just working for VLIW5's and not VLIW4's, but since they're both VLIW5's it's just not right.

 

[cdawall]

Different situation, but mainly yes. many Athlon X2's are the same generation as Phenom. Athlon II is same generation as Phenom II, etc.

On that note an Athlon X2 is a K8 and a Phenom I would be a K10. Those are two completely different cores. Same thing as saying a Pentium 4 65nm is the same as a Core 2 Duo 65nm chip. Fab means shit as far as generations go if the core is new its a different generation. Not to mention there are plenty of cards of the same generation that received die shrinks 9800GTX vs GTX+ for example same exact core with a die shrink.

If you use the faster card as the starting point, you can only say that card B is x% slower than card A. No matter how you word it, it's not correct to say it's x% faster unless the slower card was used as the baseline.

I was comparing % difference which is Card A-Card B and my post reflects that.

And not trying to bring you down completely, but 100% would be 33% faster than 75%, not 50%. I suggest a cup of coffee and/or a little bit of sleep.

Fuck it I am having another beer.

 

[Andrei23]

Came here to read the review first, great job w1zz