Is Quadfire worth it?

alienware

Banned
The big Quadfire experiment

For years now I have sat and gazed lovingly at computers with either four graphics cards in them or two dual GPU graphics cards in them. I have been envious, however I have always been incredibly suspicious. Everywhere I look I have seen people defend their systems saying how fantastic and fast they are, yet on paper and in reviews it seems to state the contrary. Are these owners doing something that reviewers and the like are not?

Well, today we are going to find out.

If ever I was under the mistaken and most foolish impression that I might be able to part the red sea and perform some miracles this would have been it. There was a very thick air of caution however, I hardly spent anything to find out. What do I mean by that? well, us PC owners are among the most intelligent on the planet (especially those that know software as well as plugging things in). However, we can also be the thickest idiots on the planet at the same time. See, we all know best and we all know better than any one else. For most of us the hardware we have is better than any one elses, and our Radeons and Geforces are better than ones you can go and buy from the shop. Ours are special and personal to us. This can lead us into trouble sometimes, as sometimes we plainly ignore the writing on the wall and find out the hard way that it may have been a better idea to listen. However, with that said sometimes I feel you can truly only learn from your mistakes.

For today's tests we are using Quadfire. With Physx. Here is a pic of the test system used for these experiments.

AMD Phenom 9950 stock.

3gb DDR2 800 (2x512mb+2x1gb)

Asus M3A32-MVP WIFI motherboard (4x 16x PCIE full speed lanes)

2 Radeon 3870x2 1gb cards (512mb per GPU)

BFG Ageia Physx PPU PCI

yayforme.jpg


You see, right away it looks impressive. Those cards are large, masculine and guzzle down power. They are the muscle car of graphics cards and every man yearns to own a muscle car. So for looks alone Quadfire is a win. Sadly in computers (like in cars) it's what's on the inside that counts, and these muscle cars ain't no sleepers.

Today's test system comprises of a very cheap pile of parts. I spent £130 in order to get it all together and run Quadfire which isn't very much. So, without further ado let's get onto the goody shall we?

When I read about Quadfire I usually came up with the same comments - "Get it if you want to sit in the corner and pleasure yourself with 3Dmark scores" . I would imagine then that they do OK in 3Dmark?

Here is the results when running a single card.

ONEcard.jpg


Which TBH is a very respectable score for a £50 videocard. Moving on then, here is the score from both cards (Quadfire) with the Physx PPU helping along.

twocards.jpg


Very good eh? About an 80% increase. Now this I had read about, so was kind of expecting. I was slightly chuffed with this, given that it isn't a million miles away from my 5770s in Crossfire. However, the worst was to come. The first thing I did was load up Crysis. Now Crysis is a poorly coded dog load of old rubbish crap a very hard game to run at acceptable frames. I knew this long before trying. First test was a single card. Everything on high, no AA, standard settings, 1680x1050

crysiscrossfire.jpg


Which again is a very respectable result for a card knocking on three years old. Infact, it beats the crap out of a Geforce 8800 Ultra. Very impressive indeed. By now I was getting excited, thinking that maybe the second card would whack a homer out of the park and I would be celebrating with a nice ice cold Bud. Sadly it wasn't to be...

crysisquadfire.jpg


As you can see performance has literally fallen apart. I spent hours trying again, trying different configs and everything else but it just made no difference at all. I was still in denial when I went looking for CFG hacks and when I found one that apparently made it see four GPUs I could hardly contain myself. But, sadly, even with the hack nothing changed. It was also as erratic as a schizophrenic patient on steroids. Every time I ran the benchmark I got completely different results. The problem was not just the poor performance but the absolutely abysmal stuttering. It made the game (and I did try it, still in denial believing that it was the benchmark tool) utterly unplayable.

None of this would matter to me if this setup could eat Fallout 3. I hate Crysis any way. Because it's a poorly coded dog load of old rubbish crap

So I ran Fallout 3. A mixture of high and ultra settings. HDR, Bloom, 4xFSAA and 15xaniso.

fallout3crossfire.jpg


And now you can see why I like my 3870x2. It cost a meagre fifty pounds and can do that to my favourite game of all time. Again, beginning to drool I ran the same test on Quadfire.

fallout3quadfire.jpg


No, I have not posted the same picture. Take note of the Quadfire logo on the benchmark. Yup, I scored exactly the same results.

But Alien, that's not so bad, at least it wasn't worse like with Crysis !

No no, it was worse. What you can't see from those results is the stuttering pile of arse it became. It was so bad that I had to disable Quadfire just to get my hour in last night. Again, a total mess and something I was beginning to expect.

Last but not least it was time for Batman Arkham Asylum. Once again I was expecting terrible scaling. Let's see what I got shall we?

Everything maxed, no Physx, no AA, one card.

batmancrossfire.jpg


And then with Quadfire enabled. Now this time I had to pinch myself to actually believe this. It was now 2am and I was groggy, sweaty (my god these things get HOT) and ready to collapse.

batmanquadfire.jpg


Yes, those are actual real figures. I even ran it about ten times just so I could make sure it wasn't a misprint or a fluke. I would imagine that being that Batman was coded for multiple GPU use (being an Nvidia based game) and to see Physx (which wasn't running) that it can actually make some use of those 4 GPUs. However, sadly it was too little too late. My mind had already been made up on Quadfire.

Final Thoughts.

Do not, under any circumstances waste your money on Quadfire.Sorry, but it has to be said. It's crap, it's terrible and like all of the other Quadfire owners on the face of our lovely planet you will end up disabling Crossfire in the control panel and the second card will be a nice slot filler for the back of your case. It serves absolutely no other purpose than to absolutely destroy your gaming perfomance and lead you down the primrose path to fool's gold believing you can get it doing what every one else can not. Hint - you can't, it's f***ed and it just doesn't work.

There is no mythical driver update or fix that will resolve this because it isn't a driver problem. Crossfire/Quadfire will scale in every game you run it with. Sadly, one of the five or so games I tried it with could actually understand what was going on. And, even with those immense framerates it was still stuttery and I would not want to live with it. You can not do to a game what the designers simply have not done. So basically it won't make a rat's arse worth of difference how much of your time and energy you waste, it's simply not going to work.

I would love to promise you that in the future this will change but it will not. Game coders are not going to waste time and money coding in stuff for stupid people. Sorry if that sounds offensive, but take my stupidity, laugh at me and don't waste your valuable time or money. Now I knew this was going to be the end result and I have rehearsed this final part of the writeup over and over. And for £50 (because the Asus board will now be making a cosy new home for my 5770s and Phenom 2 940 with Noctua 14 thinger) it was worth it just to see, so that maybe I can stop silly people from wasting £600 on that extra 5970 that is only going to ruin their day.

The only possible way I could actually justify setting up Quadfire is for -

1. How it looks (because it is such a phallus enlarging look)

2. The Vantage scores.

But other than that my friends Quadfire simply has nothing going for it.
 
Speaking as someone with quadfire, I agree completely with your conclusion. I do differ on crysis though. The latest cat app profile has reduced the stutter and I can regularly get 90+fps with x fire (around 50% extra). Fallout 3 should scale well, and the benchies show it does but as you find, there is more stuttering than the all america amphetamine addict stammering contest.

I dare say there are a couple of games out there that aren't coded quite as badly. I'm going to run a few benchies myself on some 5970s to see if I can see the same results.

But really kiddies, don't do it.
 
Well Tom's hardware beg to differ. Infact, they reakon Crossfire 3870x2 beat 4890cf in Battlefield Bad Company 2.

http://www.tomshardware.com/charts/...h-quality/Battlefield-Bad-Company-2,2373.html

They also reakon that in 3dM06 they beat a 5970

http://www.tomshardware.com/charts/...charts-high-quality/3DMark06-v1.1.0,2392.html

I have yet to see any of it apart from in Batman. So, maybe the profile for Crysis eludes me? Please share your info oh wise one
biggrin.gif


I'm installing BFBC2 as we speak, this I have to see !

And what did you have to do to Fallout 3 to make it scale? I swear these app profiles don't work for me
angry.gif
 
I found out how to make FO3 work well. Disable Catalyst AI...only uses one gpu of the four...no stuttering though
biggrin.gif


I'll have to retest crysis, but the main problem was stuttering for me with xfire, I'm using warhead though, which is coded by the monkeys slightly cleverer older brother (still rubbish).
 
Scratch that. Now I have to actually remove a card to run without xfire (unticking the box, then running a game BSODs?!?!). I'm going to test 1 vs. 4 as I can't be bothered taking a card out, reinstalling drivers, and then reversing.

Crysis everything at max, 1920*1200

Min score I think is due to the checkpoint, but I'm not going back to retest

1 gpu (1/4 of total gpus) avg 23.267 min 9 max 30

2 gpu avg 39.566 min 12 max 56 (the once this worked)

4 gpu avg 55.245 min 1 max 90

So yeah, slight improvement in max and avg, the min is always the one to fail

Just cause 2:

1: avg 33.53

4: avg 100.16
 
HOLY ******* **** **** **** ******* !

Tom's hardware weren't joking. Dude, Quadfire *destroys* Battlefield BC2.

And, I think I have found my problems all along - 10.8. It seems 10.8 is equally as bad as 10.7 was, so by reverting to 10.6 (which I run on here) BF has come alive.

I can't believe this, I really can't. Who would have ever thought that a pair of £50 cards could actually beat a £450+ card in a brand new game? (EDIT. Seems I have that wrong, the 5970 wins but not by a huge margin !)

I cranked it to the full max with FSAA on 4x and Aniso on max with everything on as high as it'll go and it just chews it up and spits it out.

I really truly can not get over how these cards throw that game around. Now I must try something else
biggrin.gif
3Dmark06 for a starters !
 
3D Mark 06 runs much worse (scorewise) than vantage for me. With vantage I can easily crank 33K+, 3d mark I'm stuck in the mid 20s. Synthetics are really built for multi gpu solutions at least.
 
I'm interested to see how close to the Tom's score I can get in 3dm06 tbh.

From the top down they have -

5850CF - 25,549

5870CF - 25,473

3870X2 CF - 25,293

5970 - 25,088

I will be interested to see how it works. I'm running Vantage again now with 10.6 cats to see if anything changes. Then Ill install 06. Got it free with my motherboard
biggrin.gif


I just really can't get over that Battlefield result though dude. It was worth the £128 to see that as smooth as it is. It's like butter !
 
I'm halfway tempted to go out and by BFBC2 (£15 tesco), but the fact that I didn't start practicing when I was still in the womb means I'll be filled with more holes than a triple *something*ed *something else* in 1860 London.
 
Amazing game. Seriously, amazing game. Get,now.

And TBH having this Quadfire thing is like sex when you're married.. Doesn't happen very often when it does best thing ever ETC ETC.

After paying what you have it's worth it dude, just to see your cards rip it to pieces.
 
May well, although I have to see what my cash reserves are like. I've just run 3dmk06 25601... what the fiddlysticks is wrong with it? Obviously it doesn't like the new drivers or the new hardware. Even my old 280GTXs could beat that.
 
Nothing wrong with it I dont think, I only scored 14,895.

The trick (apparently, I am trying it now) is to go into CCC, go to 3D, go to ALL (at the end) and where it says "wait for vertical refresh" take it all the way left to performance. This disables Vsync via the CCC for 3DM.

I'll let you know how I get on with it like that, but give that a go
smile.gif


I also looked at it and for some reason it seems to only be using one card. I'll look into it more if the Vsync thing doesnt fix it.
 
Not checked yet. But let's say many
biggrin.gif


It's as smooth as butter.

Benchmark results in. And it came as absolutely no surprise whatsoever.

Level - the one with the submarine with all of the falling concrete debris and bridges (cos it's a real killer).

Settings. Everything on as high as it will go. HBAO on (mad lighting that eats your GPUs) 4xFSAA (8 crashes the game, I think it's a problem their end). 1680x1050 DX 10.1.

I used FRAPS with a 2 minute benchmark run, just to make sure.

Same hardware as before (check up top).

Frames, Time (ms), Min, Max, Avg

7016, 90000, 54, 95, 77.956

And the entire two minute run of FPS.

FPS

73

60

71

76

71

60

77

73

63

74

77

75

80

81

77

60

76

80

81

79

78

78

75

74

75

76

79

75

81

80

81

82

80

73

71

65

79

83

84

86

84

86

85

91

88

86

81

78

75

76

82

79

79

83

84

59

74

77

74

77

82

89

76

89

88

88

81

78

73

75

79

83

77

82

84

72

90

78

79

89

88

87

84

81

78

84

77

65

64

58

So as I said, Quadfire (when working to the fullest) is an absolute animal. I will try and make a video later just so you can see how silky smooth it is, but I think from the FPS counts you will have a good idea of what to expect.
 
Sorry for the double post Tom, but the last one was longuns.

I just scalped Dirt 2 with the Quad. Man, utterly amazed tbh. Really is hard to believe this is £100 worth of GPU on a crusty old 9950
biggrin.gif


Settings. Everything on max, 1680x1050 4XFSAA Battersea stage (daytime)

Frames, Time (ms), Min, Max, Avg

5496, 90000, 36, 107, 61.067

And the complete run down
smile.gif


FPS

40

39

43

38

41

59

67

76

69

57

55

57

67

53

52

59

57

56

49

58

60

62

67

91

100

80

49

44

42

50

53

56

58

59

59

62

64

65

69

60

65

75

76

62

62

60

57

56

56

63

73

64

60

66

70

72

54

50

50

45

45

44

52

53

51

49

54

56

58

62

63

70

79

86

75

62

55

56

56

60

64

62

59

63

64

67

70

88

105

89
 
36 min isnt all that amazing tbh for dirt 2, just ran the benchmark on ultra @ 1920x1080 on this single stock 8800gt and my dualy at stock (ive been trying to work out why my mate gets such bad fps in games and he has a e7200 @ 3.33ghz & 8800gt thus running stock) and i got 4531 total frames - 51.5 average - 43.5 minimum
 
fair enough, ill load it up again in a min and give that track a go, lowest ive seen on anytrack so far is 38 tho
 
Turns out battersea is the bench mark level but here is the results using fraps in an actual race on battersea.

Min 43 - Max 68 - Avg 54.990

EDIT: shoulda edited my first post but i forgot about it
unsure.gif
, if a mod wants to merge them or something feel free
 
FRAPS automatically makes you lose a few FPS. I haven't tried it with Vsync on yet (as I was pushing for the upper limits) but I will try that too
smile.gif
 
Back
Top