• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

7950 Crossfire vs 680 SLI at 5760*1080 (Eyefinity/Surround)

How come ? I'm finding crossfire 7950's a huge performance gain over a single 7970 especially at 120Hz. The only downside for me is they are quite noisy.

Not going to notice an extra 7950's worth of performance on a single screen - 120 Hz or not unless you're looking for it with an FPS monitor.

What volts are you using for 1700 on the memory?

1625. That was a figure plucked from the sky though. Didn't make any effort to reduce it.
 
Last edited:
I notice quite a difference games are a lot smoother especially bf3. With a single 7970 I noticed dips in performance especially in 3d. With the two 7950's it's silky smooth constant.

Sorry I didn't mean I would notice it full stop. I meant I wouldn't notice it for 90% of the time when frames are already decent.
 
Thanks for the time, effort and information Rusty. :)

I'm not too surprised by the outcome tbh, reviewers should take a good look at themselves and stop taking the cash under the table.

The likes of Kyle etc, that are supposed 'experts' can't give out honest info anymore is poor, very very poor indeed...

'As it is, two GALAXY GTX 660 Ti's together will cost $640 and that buys you a gameplay experience way superior to a highly clocked GTX 680, and competes neck-to-neck with Radeon HD 7950 CrossFireX in framerate and surpasses it in overall gameplay smoothness.':(

I mean, two good maxed 680's v oc'ed 7950 CF shouldn't be getting outgunned for the money spent.:mad:

Exactly Tommy - it's ridiculous.

I know that there are always going to be slight variations in results but the fact the 7950 is so much faster at triple screen resolution in BF3 is astounding for how much they cost. The cards aren't even in the same point hierarchically! And you have to bear in mind as well that my 680's were quite golden clockers and my 7950's are average (well one of them are).

To be fair I'm not sure they reviewed triple screen resolution and crossfire of anything above 7850/660 is always going to be slight overkill for a single screen anyway. The fact that reviewers try to say otherwise though without testing it properly is ridiculous.

I have no bias - I wouldn't have sold my 680's for 7950's otherwise - shame on all the people who thought I was a nVidia "fanboy" - (yuck at the word!) so my results may not be scientifically perfect they are what I believe to be reliable.

I have to state it again - I've got a knackered card as well. With a card identical to my first one I could easily push 1200+ on the core and be under 80c which would be my comfortable limit on these cards. The difference would be even more profound.

Anyway it's not all good news but on paper at least it's large.

My "but" isn't a deal breaker and it is somewhat subjective so it'll be up to people to decide individually. If I was staying triple screen, I would be keeping the 7950's (well keeping the one that works and getting another!) so based on that you can assume it's not a terrible thing.
 
Last edited:
I'm getting 3950 on Heaven which is still slightly lower than your result Rusty. Tbh I doubt I will get much more out of them as any increase over 1250 on the cores heaven artifacts quite badly.

Try the memory a little more - Heaven loves a memory overclock. To get 4100 (:eek:) I ran at 1150/1750 on both I think. Honestly I can't remember I've tried so many different clocks :D.
 
There were plenty Matt :)

What normally happens when you out argue someone while defending a view point which is favourable to nVidia you are labelled a fanboy. :D

Yes keeping the good 7950 and single screen.

I've sold both 680's and interest in the two monitors as well.
 
He's not big on BF3 IIRC so keeping the card for that game in particular doesn't really make sense. I play a lot of it so if it was me I'd hold onto it!

Mmm? I play a lot of BF3. I'm getting in real terms roughly equivalent performance to "the beast" in BF3 which is enough FPS.

I'd rather drop to 2*MSAA tbh than have a second 7950.

Two cards are terribly under utilised at 1080p as well. If they were both going at 99% most the time it would be a little easier to justify but they don't.

I wouldn't class it as a terrible experience with one card at all. Averages of 80-90 are more than enough IMO. Granted it's not optimum but £250 for the difference between that and whatever two cards give when most of the time you won't be able to tell the difference anyway is not worth it.
 
Last edited:
@ Rusty, Your GPU temps.

If one is running a bit to warm it could be because your picking up the hot air from the other in xFire.

I think you had closed pull push GTX 680's? they blow the hot air out the back of the case, if you want the same thing with AMD you need to get one with the reference cooler or the IceQ one. others do dump hot air into the case, one will suck in the hot air from the other if they are close together.

Failing that turn your intake and exhaust fans up to create more airflow.

Thanks humbug I am aware of how it works :D.

The hot card runs at 85c full blast with crossfire disabled so it's knackered. I would expect it to heat up by 5-10c with crossfire enabled which it does, it just heats up to 91-95c :D.

The other card in the same scenario (crossfire disabled) runs at 65-75c depending on the overclock. (1150 vs 1250)

I think I said it a while back but I haven't got a hope of getting above 1150 on the hot card. So like for like, it's a cool 20c hotter :(
 
Last edited:
http://forums.overclockers.co.uk/showpost.php?p=23073346&postcount=3

OK so the results are in and if you discount Borderlands 2 which is obviously going to be a lot slower due to the lack of PhysX processing on the GPU, it is practically a clean sweep for the 7950 over the 680. Hierarchically these cards shouldn't even be comparable - the GTX 680 is nVidia's top end part whereas the 7950 is AMD's second top end part behind the 7970.

There are couple of caveats which have to be stated first which actually would have increased the results in AMD's favour even more.

I had a dodgy 7950. One of them (which I had to move to the bottom slot to stop it spontaneously combusting) had a dodgy heatsink connection which resulted in thermal throttling would you believe it if it was in the top slot. This severley limited my overclocking capability and I had to bench at 1150/1700. Not a MHz more was available from the second card and to be honest this was pushing the limits already. I was happy to push the limits if I was touching 1250 but it just felt like a gigantic anti-climax to be pushing the card so hard at only 1150 MHz. A reasonably clocked 7970 can expect to easily add 10% onto myself and a 7950 matching my first can add on 5%.

Discussing BF3 first as it seems to be everybody's favourite benchmark: comparing average FPS reveals a 26% advantage for 7950 crossfire over 680 SLI at this resolution. To give this result some context: at 1920*1080 at the same clock speeds (1150/1700) the 7950 was 8% slower so that is a 34% swing in favour of the 7950 moving from 1920*1080 to 5760*1080. This is down to the 384 bit bus which really allows the card to stretch its legs at 5760*1080.

On the subject of microstutter and or any other visual discomforts: there was nothing major. If I was forced to pick something I would say the AMD set up was just a little bit less smooth feeling than 680 SLI but without any way to measure this it's difficult to really pin point or analyse so you can make of that subjective observation as you will. I would like to think it is the effort nVidia have made to eliminate it that produced that slightly more polished feel across 3 screens.

nVidia's pricing is an absolute mess. The fact that you could potentially (as I have) move from a 680 SLI set up to a 7950 CF set up and make money on the sell up and get better performance would be amusing if it wasn't true. nVidia just can't compete on paper at this resolution with this generation of cards and there is no question of what is a faster set up. I'm obviously not going to recommend everyone sell their 680's and move to 7950's (more of which in a second) just because of this but it is worth stating nonetheless that with the free games and the price of 7950's it really is a completely viable option.

I'm not going to talk about the numbers themselves because they do speak for themselves. You can take each result at face value and I have nothing else to add on top of what the numbers say. However there is a major, major problem which would seriously make me think twice about staying with AMD CF for triple screen and that is WORK AROUNDS. (sorry for the anti climax here! :D)

Now I'm going to list a load of things which are frankly a combination of sloppy, not as nice as nVidia or borderline broken. Some can be worked around (that phrase is going to crop up a lot) but that's not really the point and if you reply to me saying "yes, but you can do this..." then you're missing the point.

  • GPU usage flies about from 10-50% while idle in Eyefinity
  • Overclocking requires unofficial patches, unlocking, reboots - it doesn't sometimes work on both cards
  • Because all 3 monitors have to plug into one card - idle temperatures are high
  • No way to manually enter the number of pixels you want when bezel correcting
  • Connections on the cards leave a lot to be desired
  • The cards do run quite hot and need to have a fairly aggressive profile to keep them cool while overclocked which will ruin your eardrums with two custom cooled cards
  • Card 2 monitoring doesn't work properly unless you disable ULPS which means more idle noise
  • Occasional blue screens (though this could be down to needing to re-format)

I would say that overall I preferred Surround to Eyefinity. Eyefinity does the bits that Surround doesn't - like switching from triple screen to extended desktop quickly and it saves your profile but it also doesn't do some things out the box that Surround does.

The fact you can't idle properly in silence because the card has to clock up to drive all 3 monitors is major pain. I'm as insensitive to noise as the next guy who overclocks the hind parts off his machine but this was too much. I'm actually looking forward to getting the broken card out the machine and unplugging the two monitors... it was that bad overall but you have to understand my opinion here is skewed by dealing with a blatantly broken card.

Now I haven't researched enough or tried enough things to say whether these are permanent issues or things that can be "worked around" but my point is more that you shouldn't have to work around these things and they should work better out of the box. The only problem I had with Surround was switching to single GPU/single monitor which caused a hard reboot.

nVidia have triple screen almost completely nailed as they have two dual link DVI slots and a DisplayPort slot which makes it so easy to connect three screens to. For SLI users you plug 2 into DL DVI on card 1 and 1 into DL DVI on card 2. Cards can successfully clock down to idle speeds as a result. Obviously the fact they have such a robust process in place for it does make the card a little bit paradoxical with a 256 bit bus but that's already been debated above :).

If you already have a custom loop then the lure of 7950 becomes even greater. These could really fly with adequate cooling.

To summarise - on paper these cards are a fantastic option. I would probably get a custom cooled one like the Twin Frozr and pair it with one of the blow out the back types like the HIS IceQ. This would be a good combination to go for. That said you need to be prepared to do some leg work to get round the little qwirks which just don't exist on the equivalent nVidia side. Crossfire on its own was fine, it was crossfire paired with triple screen which caused the little niggles.

If you can get past these minor issues (and they are minor in truth) or they don't bother you in the first place then this is what you should look at.

What is there to say about performance costing £500 which is quite simply, much, much, much faster at this resolution than two of nVidia's top end cards which costs over £300 more?
 
Last edited:
Good review.

So all in all, a 7950 is a tiny bit off the slightly higher priced 670 at 1080p (Majority of users)

They'll be even I'd say. I don't think you'd notice the difference even with an FPS counter on.

It seems a bit stupid that it's not currently doable. What's went from nVidia's hack attempts to getting multidisplay gaming working on their GPUs over 2 cards, to an advantage for people who have to cards can only be a good thing.

Since you've got the option to have all displays from one card, or 2 on one and 1 on the other there's little to complain about with that now.

Agreed.

What do you plan to do now? Keep a single card and screen as you suggested?

The knackered one is already out and the monitors are down and ready to be boxed up. I couldn't bear to have them all plugged in and idleing around 50c :D.

A question about this chart:

It seems to say that in Metro 2033, at High preset your average FPS is 95. Which benchmark tool did you use?

Built in one - Frontline.

What am I missing?

I benched with MSAA off on High preset. That?

Good summary, thanks again :) seems like both nvidia and AMD need to work on their drivers for triple screens.

Apart from the SLI switching, the nVidia drivers are pretty robust.
 
Last edited:
Back
Top Bottom