Asus HD7970 Matrix Platinum reviews

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
If the 8970 is supposed to be 25% faster than the 7970, I hope that it's this one. If it's the original, then it's gonna have difficulty beating this card. :D
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
If the 8970 is supposed to be 25% faster than the 7970, I hope that it's this one. If it's the original, then it's gonna have difficulty beating this card. :D

If AMD HD 8970 can be 15 - 20% faster than HD 7970 while running at same clocks and reach similar max overclocks of 1.25+ Ghz then its ok. anything more like 25% its very good. Anything less than 15% is a real flop.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
If AMD HD 8970 can be 15 - 20% faster than HD 7970 while running at same clocks and reach similar max overclocks of 1.25+ Ghz then its ok. anything more like 25% its very good. Anything less than 15% is a real flop.

Fair assessment. I think it will be hard pressed to be 25% faster clock/clock.
 

Keromyaou

Member
Sep 14, 2012
49
0
66
AMD might want to make HD8970 a bit more power efficient than HD 7970, with 15-20% performance increase. AMD after all couldn't make HD7990 on its own this round. They might want to make HD8990 next round.
 
Feb 19, 2009
10,457
10
76
power_peak.gif


This just shows how bad AMD's reference 1.25vcore is and not representative of custom 7970s at all, even when they are OC higher.

http://www.guru3d.com/articles_pages/asus_7970_review_matrix_platinum_radeon,7.html
 
Last edited:

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
Good speed, bad price, bad efficiency. It's easy to clock a 7970 up to 1050MHz, and as mentioned use less voltage doing so.
 

3DVagabond

Lifer
Aug 10, 2009
11,951
204
106
This is an aftermarket card. It's not an AMD design. Asus decided to live with the power usage at that voltage and clocks. Besides, this reminds me of someone showing up in a shiny new Ferrari and people commenting, How many miles to the gallon does that thing get? :D
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
This just shows how bad AMD's reference 1.25vcore is and not representative of custom 7970s at all, even when they are OC higher.

http://www.guru3d.com/articles_pages/asus_7970_review_matrix_platinum_radeon,7.html

Ya, obviously not, not only because of the 1.25V BIOS, but also because the PCB/VRM components on the reference HD7970 card are not the same as used on after-market Ghz Edition cards. The 7970 card that was flashed to GE bios uses a 5+1 VRM design. You can't even buy any Ghz Edition card like that on the market.

Look at HIS HD7970 X Turbo that has 1180mhz clocks from the factory. Still uses less power than a GTX580 or the HD7970 Ghz Reference "non-existent" edition.

Untitled-1.png


:biggrin:

I've been saying this 4+ months now and it has been ignored by certain members on our forum. The Visiontek HD7970 Ghz Edition that's currently on sale on Newegg uses less power than a GTX680 despite 1.05ghz GPU clocks.

To add to reviews:

OverclockersClub - reached 1300mhz on air
HardOCP - reached 1310-1330mhz on air
UK Hardware.info - reached 1300mhz on air
VR-Zone - managed 1301mhz on air

This is no flaky Kepler. Core voltage can be raised up to 1.35-1.36V.
 
Last edited:

blastingcap

Diamond Member
Sep 16, 2010
6,654
5
76
power_peak.gif


This just shows how bad AMD's reference 1.25vcore is and not representative of custom 7970s at all, even when they are OC higher.

http://www.guru3d.com/articles_pages/asus_7970_review_matrix_platinum_radeon,7.html

Peak doesn't matter as much as average. Average is a better representation of how much power a card uses at load. Please read TPU's definitions of what peak, average, and max wattage mean. http://www.techpowerup.com/reviews/ASUS/HD_7970_Matrix/24.html

In turn, average may not even matter as much as one might think, if the PC is left idling most of the time and is on 24/7.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
If AMD HD 8970 can be 15 - 20% faster than HD 7970 while running at same clocks and reach similar max overclocks of 1.25+ Ghz then its ok. anything more like 25% its very good. Anything less than 15% is a real flop.

The 15% rumor reference has to be related to HD7970 GE because GE is already 10-11% faster than the HD7970 card. Why would AMD bother releasing an HD8970 card which is just 4-5% faster than HD7970 GE?

At 1300mhz, this card will beat HD8970 if it's only 15% faster than 7970 GE.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Peak doesn't matter as much as average. Average is a better representation of how much power a card uses at load.

Average power consumption is not good enough when only 1 game is tested. Then we should look at 5-6 games like Legit Reviews does. The peak power consumption gives us a good idea of the 99th percentile power consumption usage a card would need a GPU demanding games like Metro 2033 or Crysis. If your PSU can't handle the peak power, you can't buy the card. Both measures are useful. The average is more useful for electricity cost calculations in games but peak is also useful for PSU selection or for people who use GPUs for things outside of games like Bitcoin mining or distributed computing that load the GPU to 99%. On average games won't have 99% GPU utilization but many programs do.

Even if you look at the average power consumption in games, Silverforce11's point is still 100% valid. HD7970 GE reference card has little correlation with what after-market 7970 GE cards use or other 1-1.07ghz HD7970 use. AMD should have never used that card in launch reviews especially since you cannot buy it. It was a pure marketing/PR blunder on their part. Even MSI Lightning with 1070mhz uses less power than HD7970 GE reference card.

power-consumption.jpg


Even with average power consumption, cards like Visiontek HD7970 Ghz edition still use less power than an HD7970 Ghz ED reference, even normal HD7970 reference or GTX680 reference. Shows how misleading it is to use AT's power consumption numbers from the review of the HD7970 Ghz edition.

power_average.gif
 
Last edited:

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
Hardocp has reviewed the ASUS HD 7970 Matrix Platinum. they have given a GOLD rating to a HD 7970 card after a long time.

http://www.hardocp.com/article/2012/10/16/asus_matrix_hd_7970_platinum_video_card_review/1

this card really is an amazing card. with 100% fan speed its still not loud and the temps are 60c at 1.31 Ghz.

The performance scaling with overclocks is good. In sleeping dogs particularly performance scaling is brilliant - almost linear scaling with overclocks.

2560 x 1600 High AA
HD 7970 Ghz - 37.3 fps
ASUS HD 7970 Ghz (1.31 Ghz) - 46.4 fps
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
The only thing that really stands out there is the 690, card is just crazy.

The power consumption on the GTX690 is awesome, but it costs $1,000. I'd rather take 2x 1200-1300mhz HD7970 Asus Matrix cards in CF any day (or GTX670 4GB SLI or GTX680 MSI Lightning SLI).

GTX690 is a very questionable purchase imo. If you have 2x GTX670/680 2GB cards, you can always sell them individually to people who have a 1080/1200P screens, where 2GB of VRAM is sufficient. As future games use more than 2GB of VRAM, that GTX690 could start hurting and its not going to be that useful for 1600P or multi-monitors. Who is going to want to buy it?

GTX690 is an amazing engineering feat, no doubt, but as an actual product for $1,000, it's not great to be honest.

For anyone who is counting, 1300mhz /7000mhz GDDR5 HD7970 is nearly 2x faster than HD6970 (and that means GTX570).

Single flops performance:
HD6970 = 2.7 Tflops
HD7970 1.3ghz = 5.325 Tflops

Double precision performance:
HD6970 = 0.675 Tflops
HD7970 1.3ghz = 1.33 Tflops

Pixel Fill-rate:
HD6970 = 28160
HD7970 1.3Ghz = 41600 * (but HD7970 has 50% faster real world throughput for pixel power, implying this is really more than 2x faster than 6970's fill-rate)

Texture fill-rate:
HD6970 = 84480
HD7970 1.3Ghz = 166400

Memory Bandwidth:
HD6970 = 176 Gb/sec
HD7970 7Ghz = 336 Gb/sec

The Asus Matrix HD7970 OCed on air effectively doubles the performance of a stock HD6970 in 1 generation, or HD7970 Matrix OC > HD6970 CF. Of course gamers who watercooled 7970s already had this level of performance in January but being able to hit 1300mhz on air rather regularly in reviews for a 7970 is pretty impressive to me as this card effectively makes all GTX680s slower and still costs has an MSRP of $480.

Cats 12.9 Betas improved performance even more for Radeons. HD7970 came a long way from December 2011. Massive driver improvements.

PerfIndex.png
 
Last edited:

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
The Asus Matrix HD7970 OCed on air effectively doubles the performance of a stock HD6970 in 1 generation, or HD7970 Matrix OC > HD6970 CF. Of course gamers who watercooled 7970s already had this level of performance in January but being able to hit 1300mhz on air rather regularly in reviews for a 7970 is pretty impressive to me as this card effectively makes all GTX680s slower and still costs has an MSRP of $480.
I agree. ASUS really outdid themselves here - this is truly and enthusiast's card. 1.3GHz performance is no joke, and the [H] article does an excellent job defining how incredibly well these cards perform at that speed.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Hey Balla, what happened to NV's infamous performance advantage in BF3? Cats 12.9 Betas and 680 is losing.
1350337975kpLWqATHq4_5_5.gif


Sleeping Dogs - one of the few games that uses DirectCompute. HD7970 1.3ghz is beating the GTX680 by 53%.
1350337975kpLWqATHq4_4_5.gif
 
Last edited:

skipsneeky2

Diamond Member
May 21, 2011
5,035
1
71
This card looks pretty epic,i was all set on perhaps getting a evga ftw gtx670 but starting to rethink that plan after checking out these reviews.

Seems the major benefit of such a beast overclock is more effective when going pass 1080p,wonder in a worst case scenario this card does 1250mhz core,how would it compete against the gtx670 in such games as BF3 at 1080p?

I just freaking love how beast this card is,when the gtx580 Matrix came out i drooled over that one too and this card of course makes more sense over the gtx680 at this point.

Better rethink my case options,i was set on purchasing a antec 300 but at 11.2'' long,forget it.
 

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
Hey Balla, what happened to NV's infamous performance advantage in BF3?

I'm not here to defend Nv's mid-range card, but my guess as to what is happening is 30% more system power draw for 20% more performance using 4xMSAA @ 1600p.

What's really common with [H] reviews is the stock vs AM OC + OC in a review. If we step back to the GB SOC OC review we see the 680 getting 49.2 FPS in the same test with the same settings. Making the 680 still faster, so to answer your question, nothing has changed.

Even in that case the system draw of the overclocked 680 non reference OC is 350~ watts, while the 7970 OC is 100w more at 458w.

Unfortunately for Nv their mid-range card is simply far more limited by design tolerances than they are wattage, this looks to simply be a case of a high end design vs a mid-range design TDP wise.
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
What's really common with [H] reviews is the stock vs AM OC + OC in a review. If we step back to the GB SOC OC review we see the 680 getting 49.2 FPS in the same test with the same settings. Making the 680 still faster, so to answer your question, nothing has changed.

while comparing two reviews you could atleast check to see if the stock GTX 680 scored the same performance in both reviews. they don't. so a direct comparison isn't correct

http://www.hardocp.com/article/2012/08/29/gigabyte_gtx_680_super_overclocked_edition_review/5

2560 X 1600 Ultra 4X MSAA (the fraps graph is shown for 600 sec.)

Gigabyte GTX 680 SOC OC - avg 49.2 fps min 28 fps
Gigabyte GTX 680 SOC - avg 46.4 fps min 24 fps
HD 7970 Ghz - avg 43.8 fps min 27 fps
GTX 680 - avg 42.5 fps min 24 fps

http://www.hardocp.com/article/2012/10/16/asus_matrix_hd_7970_platinum_video_card_review/5

2560 X 1600 Ultra 4X MSAA (the fraps graph is shown for 480 sec)

ASUS HD 7970 Ghz OC - avg 47.0 fps min 31 fps
ASUS HD 7970 Ghz - avg 42.4 fps min 29 fps
HD 7970 Ghz - avg 40.4 fps min 28 fps
GTX 680 - avg 38.9 fps min 24 fps

The ASUS HD 7970 Ghz OC scales around 16.3% from a stock HD 7970 Ghz (47.0 / 40.4 = 1.163) . If you apply that scaling to the HD 7970 Ghz score from the Gigabyte GTX 680 SOC review. 43.8 x 1.163 = 50.95 fps. so your conclusion is wrong. The MSI GTX 680 Lightning did 52.8 fps with the LN2 BIOS and voltage control allowing in game speeds of 1394 Mhz

http://www.hardocp.com/article/2012/07/30/msi_geforce_gtx_680_lightning_overclocking_redux/5

But seeing Nvidia's voltage locking and warranty voiding policy scare the s**t out of EVGA (who have removed EVBOT from GTX 680 Classified) and MSI there is no argument as to the fastest single GPU. Its the HD 7970 Ghz. Moreover in a demanding game like Sleeping dogs the HD 7970 ghz smashes the GTX 680 at 2560 x 1600 High AA. There is a linear scaling with overclocking in sleeping dogs. That is possible because the HD 7970 Ghz has huge bandwidth. The bandwidth and compute architecture of the HD 7970 Ghz really shine very well. We will see more of this in titles like Metro Last Light and Crysis 3 next year where the bandwidth will help the HD 7970 ghz pull away significantly at high AA and resolutions.
 
Last edited:

BallaTheFeared

Diamond Member
Nov 15, 2010
8,115
0
71
49.2 > 50.95 while using 100w less, not to mention by your own admission the 680 Lighting is even faster.

Sample size too small, nothing has changed.

There is nothing more demanding about Sleeping Dogs, it's performance hit comes for SSAA. Most people would argue anything past 4xAA is pointless, you're going into a realm of massive performance loss of minimal IQ improvements as if that's has any foot hold in reality.

It's almost as bad as saying AMD has a compute advantage because of the way they coded their other gaming evolved title Dirt Showdown, or saying that their compute knowns no equal in Sniper Elite v2 (gaming evolved) because it does better at blurring the entire screen with DoF compute shaders.
 
Last edited:

raghu78

Diamond Member
Aug 23, 2012
4,093
1,475
136
49.2 > 50.95 while using 100w less, not to mention by your own admission the 680 Lighting is even faster. Sample size too small, nothing has changed.

There is nothing more demanding about Sleeping Dogs, it's performance hit comes for SSAA. Most people would argue anything past 4xAA is pointless, you're going into a realm of massive performance loss of minimal IQ improvements as if that's has any foot hold in reality.

Sleeping dogs supports FXAA and SSAA . SSAA is more demanding and produces better image quality than MSAA. When HD 7970 wins like in Skyrim, Dirt Showdown, Alan Wake, Crysis Warhead, Sleeping Dogs, Metro 2033 the leads are very significant. In these games the bandwidth on the HD 7970 Ghz is the major factor for the significant performance lead.

About the relevance of these higher quality settings like MSAA 4x or MSAA 8x or SSAA some people would just turn it off and play with FXAA if present. But for that even a HD 7850 would do fine. Why pay 500 bucks to play with lower AA or no AA.

Games like Metro 2033 where in the last gen Nvidia ruled with GTX 580 are now faster on AMD HD 7970 Ghz. The same design choices which make GTX 680 power efficient also expose it in bandwidth hungry games. For a flagship chip costing 450+ bucks Nvidia is giving the user less than what they did last gen with the GTX 580.