AMD Radeon HD 7970 3 GB

[theJesus]

Quick everybody, find something else wrong with the graphs; we mustn't let W1zzard have any time to enjoy the holidays!

 

[W1zzard]

Quick everybody, find something else wrong with the graphs; we mustn't let W1zzard have any time to enjoy the holidays!

"holidays" = extra time for work

 

[GSquadron]

I think working with graphics is quite a holiday ^_^

 

[jrs3000]

"holidays" = extra time for work

Hey wizz can you give us a gift and do a quick test or two of the card with a bulldozer processor. I haven't seen any and I'm sure people are curious how it would stack up compared to intel.

 

[cadaveca]

"holidays" = extra time for work

No kids, eh? Holidays for me = 4 kids running around the house buzzed on sugar from holiday goodies.

 

[mediasorcerer]

problem found why hd 6870 sucks so much since the rebench:

http://img.techpowerup.org/111224/Capture244.jpg

result of the damage -> card running at pcie x2 (that's 2 lanes vs. 16 it should run at)

off to find another hd 6870

Could you try resoldering? has it bridged between c106/c107? seems a shame anyway.

 

[Jonap_1st]

a perfect remedy for curing the failure of bulldozer..

.. i want one,

 

[buggalugs]

In the 7970 review on the heaven benchmark page, the 6970 gets 32 FPS(@1920x1200).

http://www.techpowerup.com/reviews/AMD/HD_7970/25.html

On an old 6970 review for the HIS model the 6970 gets 45 fps in heaven:

http://www.techpowerup.com/reviews/HIS/Radeon_HD_6970/26.html

The test systems look to be exactly the same, same res, same everything, how can a 6970 lose 25% performance over a few months??

 

[cowie]

A card loses performance after a few months by the drivers needing to actually apply aa/af or tessellation.
On the other hand they do gai some fps over time by being better optimized(Thats the one you hear most

Anyway i cant believe wizz dont use 03 05 or 06 anymore!!wtf?

 

[cadaveca]

03, 05, and 06 are all CPU-limited. In other words, vastly more powerful cards don't score as much as they should due to CPU limitations. Those tests are good for overclock testing only, IMHO.

 

[cowie]

lol
I know its low res when default but i just am so used to seeing his reviews include those benches.
At the same cpu speeds i could get the jist of a score and ponder what a card could do by certain benchmark areas.
I'm not that big a gamer so i dont care if one card can run a game 10fps faster or stuff like that.

 

[cadaveca]

Ah. well, if benching, I think the reports of 1335 Mhz stock cards bodes well for future benchmark records.

 

[cowie]

You aint kidding,and the way they clock on dice/ln2 they will be top of the boards till either a refreash or nv cards are out(maybe).

 

[WhoDecidedThat]

nice.. hope to see real performance if this card test on pciE 3.0 mobo..

Not Possible. SB-E has 8 PCIe 3.0 lanes whose bandwidth is equal to 16 PCIe 2.0 lanes since PCIe 3.0 doubles the bandwidth of PCIe 2.0 per lane. So there shouldn't be much performance improvements.

 

[WhoDecidedThat]

50% is a lot to expect, I don't know if any generation gap has brought those kind of figures.

GTX 480 was about that much faster than the GTX 285.

 

[WhoDecidedThat]

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?

It was designed for 2304 SPs at 1 GHz. They backed it down.

 

[Casecutter]

Wow! Home Run with two on base, bottom of the 7th! Not a base loaded, but definitely has the crowd on their feet!

Sure the price... while high I don't know if it's truly all falls on AMD's shoulders. TSMC is a player in the new price structure, and I wouldn't make that call until Nvidia has their "Über Enthusiast" part is in the market to pass judgment. Heck, Nvidia being 6 months (or better) on that release might play in their favor, that’s a millennium for TSMC/Nvidia to tweak their process(s) and provide cost ratio on deliverables. Though Nvidia big die strategy is still as always a factor (strike) against them, it again is AMD taking lead, being the "guinea pig" and on parts for their star player first. That was a baall'ies move IMO.

Then if you’re thinking of this for 1920x don't think of coming from popcorn stand to the dugout! Unless it's for BF3, Skyrim or STALKER COP; those are titles really pushing the hardware and that's how it works. (you have to pay to play to be a Top Tier Player)

With awesome power/performance it's a long wait for Kepler, and what's now I’m re-terming as "Super Über segment", and dropping "Über Enthusiast" to 580/570/6970/6970 or what will be the 7870 when it shows..

Is it time for the average Joe to grab one? Only if you’re a Joe with disposable income.

 

[Sir Alex Ice]

The testing method used seems to be old and outdated on 2 accounts:
1. the resolution used 1900x1200 should be replaced with the much more popular 1920x1080
2. Unigine's Heaven has reached version 2.5 and should be used to test all recent cards.

 

[crazyeyesreaper]

for one 1920x1200 and 1920x1080 are so damn similar you wont see a difference worth bothering with so its a moot point

Heaven Bench at 2.5 or 2.0 dosent really matter all that much 2.5 is just a longer bench run but in the end it dosent matter because all gpus are tested in 2.0 meaning the performance head to head is still indicative of DX11 performance.
1024x768 4x3 = 786,432 pixels
530k difference
1280x1024 5x4 = 1,310,720 pixels
450k difference
1680x1050 16x10 = 1,764,000 pixels
300k difference
1920x1080 16x9 = 2,073,600 pixels
300k difference
1920x1200 16x10 = 2,304,000 pixels
1.7m difference
2560x1600 4,096,000 pixels

the reason 1920x1200 is popular for testing is because it has more distance away in terms of pixel count from 1680x1050

300k pixel difference from 1680x1050 to 1920x1080 the difference is small enough that the change in FPS would be a bit less

using 1920x1200 offers up 600k pixel difference which gives a better overall indication of performance for those at 1080p because they can expect that not all systems are the same so the bit of an extra drop for 1200p gives people breathing room, not to mention it tends to even off the pixel increase aka 500 450 600 is a bit more preferable to 500 450 then drop to 300.

not to mention 1080p might be more POPULAR but youll be hard pressed to find many on this forum that prefer 1080p to 1200p most would take 1920x1200 if they can get it.

thats how I see it anyway.

 

[Sir Alex Ice]

Full HD is the current norm for gaming systems, either PCs or console based. Not using is living in the past.
Remaining with the version 2.0 when 2.5 is out and has been out for a while is also living in the past.
Finally, if there is not difference between them there should be no problem in using the new resolutions and versions.
I can understand keeping the test data for comparison, but when a new generation of VGA comes out the testing method should really be updated.

 

[crazyeyesreaper]

well then go read other reviews that offer only 3D Mark 11 Vantage and 3-4 games then

fact is w1zzards isnt like most other reviewers he dosent keep stagnant old data around

every so often he rebenchs every single gpu you see on those charts benchmarks are kept as they are for comparison reasons as throwing away valuable data is worse then using an earlier benchmark version

get over it.

find me another review site that tests the same number of GPUs review tests and tests it on as many games as W1zzard does.

not to mention he tests at resolution ranges from 1024x768 up to 2560 x1600

most review sites test at 1680x1050 and up thats it. W1zzards work load is nearly 3-5 times more then most other GPU reviewers.

you want him to get with the times okay in that case lets drop half the gpus from the chart and some of the games.

if your going to nit pick a single damn benchmark and a single resolution you need to shape up

you do realize that your in favor of 16:9 1920 x 1080

well guess what 1680x1050 is 16:10 and its the most common PC resolution period at nearly 3x the number of PC users vs 1920x1080 res

to be blunt 16:10 aspect ratio still has greater then 30% market share for desktop resolutions

1680x1050 1920x1200 and 2560 x 1600 are all thats right 16:10 = 30% market demographic according to steam

1920x1080 might have 17% but 16:9 displays only make up 20% of the market

untill 16:9 takes over as the dominate aspect ratio i see no point and switching out 1 resolution from 1920x1200 to 1920x1080 when it serves no purpose.

also your telling a reviewer to go out and buy more hardware to suit your needs. when its not neccsary

why not replace all resolutions remove the 4:3 the 5:4 and switch everything to 16:9 oh wait thats right expect for 1920x1080 almost all other 16:9 desktop resolutions have crap market share.

According to Steam
Total Aspect ratio of general users

4:3 = 7%
5:4 = 11%
16:9 = 21%
16:10 = 31%
misc = 4%
theres another 23% roughly misc resolutions thats not reported. aka 1920x1200 2560 x 1600 and other lower resolutions

http://store.steampowered.com/hwsurvey

of those resolution w1zzards supports the better PC market share for overall penetration in aspect ratio aka 16:10

but he also represents the most used 4:3 and 5:4 resolutions as well

no to switch to 1920x1080 from 1920x1200 is maybe 2% margin of error to the point it would be a worthless expense for w1zzard. to switch.

Altho im sure if you were so kind as to donate a 1920x1080 screen he might help you. but im willing to guess hes using the 2560x1600 res monitor to test all resolutions there for sticking to 16:10 works best over all since thats the aspect ratio the monitor was design with in the first place.

 

[ViperXTR]

haha im sorry if im not rich enough to get 1080p display combo'd with a "capable" gpu.
I always tend to go in the mid range segment GPUs since i only use non full HD display, up to now im still using a 1280 x 1024 5:4 display (!!) on my weak GTX 460SE overclocked, at least it makes me play my game with less hassle vs running on a 1080p, so thats why Wiz's benchmark results are pretty helpful in my case at least

 

[crazyeyesreaper]

its why youll notice i posted the 1 line there about 4:3 and 5:4 1024x768 and 1280x1024 both represent the largest group of uses in those 2 aspect ratios, for entry lvl gaming

1680x1050 1920x1200 and 2560x1600 of course are all 16:10 and represent the proper resolution for the monitor hes testing on.

monitor size dosent matter, just point out obvious facts for why w1zzard does what he does. most people think its SO easy to just change shit. they have no idea how much work is involved.

 

[buggalugs]

A card loses performance after a few months by the drivers needing to actually apply aa/af or tessellation.

We actually looked into that very question here, to see if driver performance improved over time and the answer was it stays about the same overall, with maybe a 2 or 3 % improvement over several months.

You can get the odd game where drivers improve performance by 25% or something but overall its more like 3%. That what the reviews show.

Theres no way a 6970 can lose 25% performance in a benchmark(heaven), in a couple of months.

I still think there is something wrong with the AMD numbers in the 7970 review, they are not consistent with Wizzards past reviews, using exactly the same hardware.

Wizzard, can you explain why the 6970 scores 45 fps in heaven in the 6970 review, but scores 32fps in heaven in the 7970 review?

 

[dj-electric]

So... much... power....

For reference, Stock HD5850 - about 7700~