Associate
- Joined
- 1 Aug 2011
- Posts
- 81
hi guys was wondering which of the above settups would give best fps gaming at 1080p budget is £270 for cf i would have to get a new psu tho, thanks
-
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.
6970 vs Gtx 570 Vs 6870CF
- Thread starter aceofspadesrob
- Start date
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,248
- Location
- Billericay, UK
I would be tempted to get 2 6870's but then you get the issues with stuttering, drivers which are common place on multiple video card setups. I don't know how bad it is so prehaps talk to some people who run this sort of setup.
Until there exists anyone who can prove that 1GB cards in SLI/CF can do metro2033benchmark max without many lag spikes, I do not recommend putting 1GB cards into SLI/CF.
Single card: GTX570
Multiple cards: 6970 (may add another one for CF in the future)
Single card: GTX570
Multiple cards: 6970 (may add another one for CF in the future)
Soldato
- Joined
- 25 Sep 2009
- Posts
- 10,248
- Location
- Billericay, UK
Ignore the foolishness above. Play Metro 2033 on 2 6870's and you can expect nearly double the performance at 1920x1200.
If your video card is running out of VRAM it can use your system ram as temporary measure, its only once you start to go massively over you get inconsistent performance.
If your video card is running out of VRAM it can use your system ram as temporary measure, its only once you start to go massively over you get inconsistent performance.
- Joined
- 1 Aug 2011
- Posts
- 81
lol advantage with cf is i can buy 1 card this month and 1 card next month with new psu seen ad there only £130 @ scan for the 6870
Bang on budget, without a doubt, if you can afford a 3 slot gpu, I would get this:
http://www.overclockers.co.uk/showproduct.php?prodid=GX-250-AS&groupid=701&catid=1914&subcat=
OC it to hell and back and it will win more than it loses!
http://www.overclockers.co.uk/showproduct.php?prodid=GX-250-AS&groupid=701&catid=1914&subcat=
OC it to hell and back and it will win more than it loses!
If you don't want a ban, you better edit your post to something like this:
You can't mention any other competitors here.
You can't mention any other competitors here.
Bang on budget, without a doubt, if you can afford a 3 slot gpu, I would get this:
http://www.overclockers.co.uk/showproduct.php?prodid=GX-250-AS&groupid=701&catid=1914&subcat=
OC it to hell and back and it will win more than it loses!
this card is just beautiful.To the OP: if you decide to follow a fool like this, then better not to look into the following comparison:
It is indeed, but my twin custom 6950>70's in CrossFire look mental
(Here's 1 of them): 
Which I named them as(shall give you the full title
) the: 'Sessanta nove cinquanta/settanta l'edizione di gemelli brutale!'Or in English, the: '6950>70 the brute twins edition!'
Last edited:
Swings and roundabouts.
The 1GB cards may have more spikes in the framerate but the 2GB cards have a whole section where they drop below the frame rate of the 1GB cards.
1GB cards ain't so bad after all.
Last edited:
This is why I post such comparison - many people still heavily rely on traditional benchmark results of average fps and adhoc min fps, not even having to mention that most of those fps numbers are not instantaneous fps but averaged numbers over 1-second time intervals.
When talking about vram limitations, I'd like to count the total percentage of time with instantaneous framerate below a certain threshold e.g. gaps above 33ms. This defines the smoothness of the play.
Also keep in mind that this Metro 2033 benchmark does not rotate the camera quickly or switch any scene. If the benchmark does, as the Crysis benchmark / Shogun 2 benchmark, I expect more serious lag spike / stuttering problems from 1GB cards.
Last edited:
Like with the 2GB cards where they drop below 10FPS for several seconds?
That's a lot of gaps of 33ms.
Let me put it this way, in order to discuss the smoothness, I'm saying it's necessary to carry out a statistical measure over the whole benchmark session. Pointing to a local section is like looking into adhoc min fps. In the section you picked, both the 1GB cards and the 2GB cards got lots of gaps > 33ms. Focusing on this local region is a bad pick to differentiate these two setups. In order to tell the difference, simply look over the whole benchmark session, and I believe it is not difficult to see the difference.
What I'm interested is to see how 6870 CF works. If we still see lots of lag spikes / stuttering, just like what I saw in 5870 1GB CF and 560 Ti 1GB SLI, then it's evidence of vram limitation, due to the latency of swapping between dedicated video memory and system main memory through PCIE communication.
Last edited:
To the OP: if you decide to follow a fool like this, then better not to look into the following comparison:
What? Everything about the 1G cards is better on that comparison, higher max framerate, even higher min framerate. Average rate is higher as well.
An important point to make is that neither setup is actually playable. The benchmark is running below 20fps for most of the "action section" which is simply not enough. This is a pure example of GTX560TI SLI not being powerful enough rather than purely running out of memory.
Settings would have to be dropped to bring up the minimums in both cases.
The same would be true for any of the setups the Op is considering.
You are missing the main point I'm trying to tell. This comparison is to show people how vram page fault causes lag spikes / stuttering, and how traditional avg/min fps fail to detect such phenomena. Suppose there is a game with exactly the same load as the last 10 seconds of the benchmark, you can clearly see average fps above 30 (playable) while the 1GB setup suffer from lag spikes.
There are many other games which can benefit from more than 1GB vram without having to drop the IQ in order to achieve 60 fps. Unfortunately to capture the lag spikes / stuttering caused by vram limitations, we need to use Fraps or whatever to record the frametimes, and run data analysis. This is a pain because it is already difficult to find the same GPU with different vram size, and even more difficult to beg different users to use Fraps to capture frametimes and upload. Metro 2033 benchmark is convenient - it already makes plots based on frametimes, which is in highest resolution of showing lag spikes.
Last edited:
Metro 2033 is hardly the best comparison but I think it's generally true that it's a bad idea to go SLI/CFX with 1GB cards.
You do need to go look at fps graphs of games on [H] - one of the only review sites that shows you how fps changes over a period of time, rather than most sites that just plop a meaningless number down for you. Even with a lower 'average FPS', some card setups might be more playable than cards with a higher average which suffer from heavy spikes and stuttering every 5 minutes.
You do need to go look at fps graphs of games on [H] - one of the only review sites that shows you how fps changes over a period of time, rather than most sites that just plop a meaningless number down for you. Even with a lower 'average FPS', some card setups might be more playable than cards with a higher average which suffer from heavy spikes and stuttering every 5 minutes.
Exactly. This is why Duff-Man made a dedicated program to show people how to measure microstuttering.
Unfortunately [H]'s reviews only use fps sampled every 1 second, i.e. they average the instantaneous fps over each 1-second time interval, smoothing out most lag spikes. To detect lag spikes, it is vital to plot based on frametimes, not average numbers.
Last edited: