GeForce 680 & 670 vs. Radeon 7970 & 7950 Gaming Performance @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,601
GeForce 680 & 670 vs. Radeon 7970 & 7950 Gaming Performance - Wondering how the GeForce GTX 680 and GeForce GTX 670 compare to the Radeon HD 7970 and Radeon HD 7950 at stock frequencies as well as overclocked? You ask for it and we have done just that. Hold onto your hard earned cash and take note of just how the new GeForce GTX 670 compares with the rest of the competition.
 
When the 670 reviews came out, I immediately thought that the 7970 and 7950 needs another price drop.
 
Very comprehensive review.

I'd like to see the workbench with all those cards stacked up....:eek:

Looks like the nvidia GTX 670 is a show-stealer when overclocked.....price-wise.

(been "just looking" for a number of years here......always enjoy the GPU reviews, thanks for the comprehensive review):D
 
The OC vs OC comparison is relevant to me. I don't remember the last time I've had a stock running GPU or CPU. In my opinion, it's also a more "fair" comparison of hardware performance because the 6 series dynamically overclocks. Without overclocking the 7 series, you are effectively comparing an OC card to a non-OC card. This levels the comparison field.

Unfortunately, the comments about xfire support is completely valid and this will be the last time I do a multicard set up with AMD.
 
The delta in power is NON trivial as well...for an avid gamer, it could save them $30/year in electricity alone per card.
 
Trepidati0n said:
The delta in power is NON trivial as well...for an avid gamer, it could save them $30/year in electricity alone per card.
Click to expand...

I am all for saving as much electricity and money as possible, but when it comes to $2.50/month worst case the difference between cards doesn't matter from a price point. Now building an argument off of heat output from the lower consumption card I can see.
 
Trepidati0n said:
The delta in power is NON trivial as well...for an avid gamer, it could save them $30/year in electricity alone per card.
Click to expand...

It seems odd to spend $400-$500 on a card, or double that and then worry about $30 a year in electricity. Especially as the gamer is also likely buying multiple monitors, etc. Give small energy differences the same weight you do the adapters packaged in with cards, because having to buy those can cost $25-$50 as well.

Of course, few people alter their purchasing decision based on the bundle that comes with video cards. The same applies for these types of energy savings. It's just a talking point in the red. vs. green dynamic.
 
My 670 came in today, can't wait to get my SB CPU and get this bad boy running!
 
Overall a good review, but with two little points.

In the stock tech specs I strongly believe it would be more accurate to list the NVidia cards at their boost speeds rather than stock speeds if those cards/chips tended to run at boost speed for the great majority of each test.

Second, any time there is an FXAA based review there also needs to be a standard MSAA review because FXAA is NVidia created and NVidia favoring AA technology.

For the AMD side of things. Thank you for again grilling them about drivers. I am still stuck running the release and I want a stable, installable update that can let crossfire work properly. Your continued gripping on this point may actually change things for the better for the broader consumer base. It also good that you clarified a 1260 overclock on the current batch of 7970s is on the high side.
 
My question is: Is the perf the same at eyefinity resolutions? We have seen in the past where some cards hold higher framerates better into the higher resolutions. (5760x1200) In other words two card have similar perf at say 1920x1080 but when you jump to the eyefinity res one card jumps way ahead in perf.
 
Last edited:
I'd be more than happy to have any one of those cards to replace my 470! Lovely article, though. This is exactly what so many people, including myself, have been looking for. Thanks for your contributions, Brent, Kyle, and the rest of [H]ard!
 
Teenyman45 said:
Overall a good review, but with two little points.

In the stock tech specs I strongly believe it would be more accurate to list the NVidia cards at their boost speeds rather than stock speeds if those cards/chips tended to run at boost speed for the great majority of each test.

Second, any time there is an FXAA based review there also needs to be a standard MSAA review because FXAA is NVidia created and NVidia favoring AA technology.

For the AMD side of things. Thank you for again grilling them about drivers. I am still stuck running the release and I want a stable, installable update that can let crossfire work properly. Your continued gripping on this point may actually change things for the better for the broader consumer base. It also good that you clarified a 1260 overclock on the current batch of 7970s is on the high side.
Click to expand...

First, I think the MHz are really of no importance. Could have written the entire article and not mentioned those beyond "stock" clocks. This is about performance not about IPC or MHz.

Second. Your opinion is noted.

Agreed.
 
Well from personal experience even @ 1080p with 4xAA in the titles mentioned especially BF3 680 pulls way ahead for me. I own a 7970 also, I play at 1080p all ultra and 7970 with 4xAA was a no go for me but with 680 it feels much better and fps is higher. I wish these reviews would try and reflect that also. FXAA in some titles really looks like shit IMO. FXAA makes BF3 looks nasty on 7970 and 680, but in say batman and deus ex it looks fine.
 
Rattle said:
Well from personal experience even @ 1080p with 4xAA in the titles mentioned especially BF3 680 pulls way ahead for me. I own a 7970 also, I play at 1080p all ultra and 7970 with 4xAA was a no go for me but with 680 it feels much better and fps is higher. I wish these reviews would try and reflect that also.
Click to expand...

What exactly do you want us to "reflect" about your personal experience in our reviews? I am a bit lost on what you are asking for. Please be specific.
 
Rattle said:
I know these cards are geared a lot towards 2560x1600 and 3 monitor setups but I think there's quite a few people like me with 1x 1080p that play with all the goodies up on max.

I would like to see more 1080p with 4xAA comparisons. For me the 680 in BF3 with 4xAA is better than the 7970. If you use them both with FXAA in the same situation they are much more EQUAL.
Click to expand...

I got 80-90fps Ultra 4xMSAA in BF3 @ 1200p with a 680 (up to 120 avg. and 180fps max in MP), hardly worth reviewing the card at that resolution.
Unless it's a 120Hz review, it's already known the 680 will crush all at 1080p 60Hz.


Great Review Brent and Kyle.
 
Awesome review. Hopefully all the BS about OCed 680 vs OCed 7970 will come to an end.
 
Last edited by a moderator:
On topic please. The trolling posts have been removed.
 
Bastards!

Always getting the newest, coolest stuff in your rigs before everyone else!

Teenyman45 said:
...Second, any time there is an FXAA based review there also needs to be a standard MSAA review because FXAA is NVidia created and NVidia favoring AA technology...
Click to expand...

I thought even tho nvidia came up with it, it actually helps both brands, assuming the game engine supports it?


Rattle said:
... FXAA in some titles really looks like shit IMO. FXAA makes BF3 looks nasty on 7970 and 680, but in say batman and deus ex it looks fine.
Click to expand...

Again, I think it depends on the game engine. I've forced FXAA on in the gpu control panel.
 
Last edited:
Wow... now I really am wanting to upgrade to a GTX670. Have to wait a few months, but how green compares to red is eye-opening. What I like about the article is basically this:

You spend "X" amount of dollars you get "Y" or "Z" performance based on going green or going red.

The best bang for the buck looks to be the 670. Like Kyle and Brent said in the 670 review, this beast is the "new" Ti4200. :)
 
Its pretty clear from this review that in general on overclocked 7970 is the same as an overclocked 680, and both are more powerful than the 670. The 670 is a heck of a card for the money though, and would be my choice if I was running a single 1080p monitor.
 
Hey Kyle,

(excuse my ignorance... this isn't something I've actually looked for myself)

Is there any sort of DB out there that users of both AMD / Nvidia cards are using to post their highest / stable clock speeds on the cards they purchased? I think you'd be doing your user base a much greater service by OC'ing your cards to those values vs the awesome clocks in this review.

I've read about more then a few people on this forum not even being able to hit much higher then 1.1ghz on their 7970's. With that said though... this isn't the first time I've been wrong about something.

Once again... Thanks for the awesome reviews.
 
Seraphical said:
Hey Kyle,

(excuse my ignorance... this isn't something I've actually looked for myself)

Is there any sort of DB out there that users of both AMD / Nvidia cards are using to post their highest / stable clock speeds on the cards they purchased? I think you'd be doing your user base a much greater service by OC'ing your cards to those values vs the awesome clocks in this review.

I've read about more then a few people on this forum not even being able to hit much higher then 1.1ghz on their 7970's. With that said though... this isn't the first time I've been wrong about something.

Once again... Thanks for the awesome reviews.
Click to expand...

This is a community activity that is ongoing...

GTX 680 Overclock Thread.
http://hardforum.com/showthread.php?t=1681453

GTX 670 Overclock Thread.
http://hardforum.com/showthread.php?t=1692660
 
I'd like to see a non-reference vs non-reference comparison. Most non-reference 7950's can hit 1100 easily , even people are hitting 1200. Your XFX DD 7950 hit 1195. Maybe I'll compare the 2 scores and post em here. I guess it wouldn't be fair until that Galaxy 670 gets review and you can compare that non-reference
 
Seraphical said:
Hey Kyle,

(excuse my ignorance... this isn't something I've actually looked for myself)

Is there any sort of DB out there that users of both AMD / Nvidia cards are using to post their highest / stable clock speeds on the cards they purchased? I think you'd be doing your user base a much greater service by OC'ing your cards to those values vs the awesome clocks in this review.

I've read about more then a few people on this forum not even being able to hit much higher then 1.1ghz on their 7970's. With that said though... this isn't the first time I've been wrong about something.

Once again... Thanks for the awesome reviews.
Click to expand...

There are threads going on you can easily find in AMD/nVidia section.

About the 7970 OC, I have personally never seen a non-defective card that cannot reach 1.125Mhz. While some did claim they cannot, but if you pay attention to their previous post you will find something quite funny. :D
 
Thanks for the review -- No doubt the 670 /680 cards rock. I guess I'll be keeping my 7970's for a lil bit longer though. Maybe AMD will release some cat-daddy drivers soon giving the 7970 a nice performance boost. One can always hope... :)
 
heatthegreat said:
I'd like to see a non-reference vs non-reference comparison. Most non-reference 7950's can hit 1100 easily , even people are hitting 1200. Your XFX DD 7950 hit 1195. Maybe I'll compare the 2 scores and post em here. I guess it wouldn't be fair until that Galaxy 670 gets review and you can compare that non-reference
Click to expand...

When I read the OC'ing thread it seems pretty clear to me that the average overclocks on Kepler are ~1250MHz core and ~3300-3400MHz memory (2x effective). I don't think temperatures or power delivery is the bottleneck but I could be wrong. A non-reference full custom Kepler would make for an interesting review...
 
If you read the OC thread on the 7950, most of reaching 1100 quite easily with some reaching 1150/1200 respectively. If I were buying now, I'd go with a 670, but since I spent the same amount for my gigabyte wf3, and also am able to reach 1150, I see no reason to switch, as I'm probably beating a 670 in most tests.
 
Awesome cards for sure. Now if we can just get something for under $400...
 
Good article. Thanks.

If one card is giving me 75 FPS in a title and another is giving me 90 FPS I'd go for the 75 FPS card if it "pauses" a lot less when I'm running around in a game.

I'm saying feel and smooth action is much more important to me than whatever FRAPS may be reporting.

Raw average FPS reports don't always tell the whole story.
 
Last edited:
I *SWEAR* to god there is more to come from the 7000 GPUs. Just have that gut feeling. (from the drivers, I'd give until Cat 12.6 before I would consider it 'case closed' on that front.)

All the same excellent article as always.

Hardocp LISTENS to it's readers. That's why it's no.1 in my book.
 
So OC'd single card go AMD, for everything else there's MasterCa, err Nvidia
 
spinejam said:
Thanks for the review -- No doubt the 670 /680 cards rock. I guess I'll be keeping my 7970's for a lil bit longer though. Maybe AMD will release some cat-daddy drivers soon giving the 7970 a nice performance boost. One can always hope... :)
Click to expand...

I would keep an eye out for Cat. 12.5 then. I've been hearing good things about the beta and 2 card XFire

night_2004 said:
When I read the OC'ing thread it seems pretty clear to me that the average overclocks on Kepler are ~1250MHz core and ~3300-3400MHz memory (2x effective). I don't think temperatures or power delivery is the bottleneck but I could be wrong. A non-reference full custom Kepler would make for an interesting review...
Click to expand...

If it isn't temps. or power delivery, then .... hang on a second it must be one or the other.

Blacklash said:
Good article. Thanks.

If one card is giving me 75 FPS in a title and another is giving me 90 FPS I'd go for the 75 FPS card if it "pauses" a lot less when I'm running around in a game.

I'm saying feel and smooth action is much more important to me than whatever FRAPS may be reporting.

Raw average FPS reports don't always tell the whole story.
Click to expand...

Is this really an issue for single cards? What fraps says is going to be true with 1 card, surely?
 
Stoly said:
For the price? not even close.
Click to expand...

Well in defense of you coming out and basically calling bullshit on what we wrote, let's get a correct quote to respond to.

And given the GTX 670 card's overclocked performance, shown on page 7, encroaching into stock clock GTX 680 territory, we would suggest that we have somewhat of a "new Ti 4200" on our hands.
Click to expand...

Now, if you would like to comment on what we actually wrote, please go for it.
 
I believe the oced gtx680 looks a bit better than the oced gtx670 mainly because the memory was able to oc much higher on your sample. they both use 6000mhz memory chips and its the luck of the draw as to whether the memory will oc that high on the gtx680. basically the gtx680 would be less than 5% faster clock for clock than the gtx670 from what I am seeing from all reviews combined so that extra bandwidth probably really helped.
 
You must log in or register to reply here.
Back
Top