To SLI or not?

[Big.Wayne]

Obviously direct no ire towards me as I am just asking for "proof"!

Such a tiresome post, it sounds to me personally that ANY amount of data can be presented to you and you would still refuse to believe it

That's totally uncalled for Ejizz! . . . if there is a problem here that can be "proved" then there is scope for legal proceedings to be made under the trade descriptions act . . . please refrain from making further posts like the one you just made and please do not jump to assumptions about anything I say or anything I do? . . . I tire of people being unfriendly just because I ask for proof?

 

[An Exception]

Ummm, Micro-Stutter has already been proven, and no, there is no case for legal proceedings, unless your feeling lucky and have money to burn...
Oh and I think it's actually you who should refrain from posting, unless you can 'actually contribute' something...

 

[Big.Wayne]

Ejizz, you are a new member here on OcUK forums and I have said not one unkind word to you but yet somehow in your mind you feel "justified" to be unfriendly to me? . . . I'm not sure what forums you posted on before but please refrain from being unfriendly and unhelpful here? . . .

Your idea of what constitues "proof" and my idea of what constitues "proof" are obviously very different? . . . if there was a problem with good factual proof to show this do you think I would point people towards problematic hardware? . . .

Chill dude . . . we are on the same team!

 

[An Exception]

Ejizz, you are a new member here on OcUK forums and I have said not one unkind word to you but yet somehow in your mind you feel "justified" to be unfriendly to me?

The Op started a thread asking for advice, the Op was given lot's of different advice. After reading the advice and with strong evidence presented to the Op, he stated he was going to take the single GPU option.

Following this you to basically called the Op a sheep and suggested he wasn't capable of thinking for himself, yet you have the cheek to call me 'unfriendly'!

 

[An Exception]

Chill dude . . . we are on the same team!

No probs, I'm done now...

 

[Anewbe4u]

Ejizz, you are a new member here on OcUK forums and I have said not one unkind word to you but yet somehow in your mind you feel "justified" to be unfriendly to me? . . . I'm not sure what forums you posted on before but please refrain from being unfriendly and unhelpful here? . . .

Your idea of what constitues "proof" and my idea of what constitues "proof" are obviously very different? . . . if there was a problem with good factual proof to show this do you think I would point people towards problematic hardware? . . .

Chill dude . . . we are on the same team!

I often find myself agreeing with what you're saying, but I have to be honest, you come off as a bit arrogant some times in the way you put your points across.

Anyway, get what ever card you like!

 

[Duff-Man]

Thanks for writing that reply, the basic technical SLI stuff I already new but I haven't encountered an explanation of this "microstutter" thing before? . . . I just read the XS thread through from start to finish and I can't say there is any evidence presented in there that would stand up in a court of law!

All I'm doing is applying a simple and well known mathematical algorithm to quantify the degree of framerate variance in the output. It exists, and there are numerous ways it can be seen. My interest is in assessing the situation under as wide a range of conditions as possible, so I can better understand it.

I could just jump on the "microstutter" bandwagon with you guys and accept the explanation as given but allow me to take up the role of "Devils Advocate" and say the problem doesn't exist, your program is badly coded and the few users who complained had a faulty overclocked system and mushed up drivers?

Nothing wrong with playing devil's advocate. But let me set your mind at ease: The phenomenon is repeatable, and to a certain extent predictable. We understand where it comes from, and the situations in which it occurs. Anyone with a multi-GPU setup can see it for themselves. Just run a FRAPS benchmark, and check the output. It's not just a product of mushed up drivers.

If a genuine punter looked at some SLI benchmarks and saw 90FPS in a game he liked and then bought the exact hardware as used in the review, installed everything are you suggesting the performance wouldn't be the same? . . . all things being equal he should see approx 90FPS in the counter and he should have a smooth and enjoyable end user experience? . . . are you suggesting oherwise? . . . and if so can you prove it factually beyond the "Hearsay" of 1000 people?

The hardware would be 90fps. This is 100% true. The subtlety is that the 90fps be output irregularly, and so would look more like a game scene running at say 75fps, (for a 15% microstutter readout). The reason for this is that the human eye notices the gaps between frames as the basis for smoothness - it doesn't just count the raw number of frames like a computer. In your 90fps case the game would still seem very smooth, since 75fps from a regular output source would still look smooth...

Okay take it to the extreme, as an example, and consider a case where you have 60fps output, but two of the frames are output simultaneously. Or just a fraction of a millisecond apart. Your eye doesn't catch the first two frames, and although a solid 60fps is being output, you're looking at a scene that seems as if it is running at 30fps. This would represent 100% microstutter in the program I wrote. Reality is a much less extreme version.

As for proof, I can prove that it exists, and I can measure its effect. That's factual science that will stand up to any reasoning. As for anything else, it's all subjective. Adding a second card will still improve performance, but microstutter or not, that doesn't guarantee a "smooth and pleasant gaming experience".

In that thread there was a lot of assumptions and conjecture made? . . . has anyone done some very, very controlled testing, noting exact hardware configs (stock/overclocked etc), driver versions, made videos etc? . . . is this something that can be observed with the human eye or? . . .

I would love to have a whole suite of hardware to test out the effect on scientifically. Of course, I only have one rig and a couple of hardware sets to work with properly. So, I wrote a tool that other people could use to generate data in scenarios I can't replicate. I'd love for a review site like anand or HardOCP to include a formal analysis in their reviews. But they are familiar with average FPS benchmarks, and further complicating things is not an attractive proposition.

What I learned from XS and from Hardforums is a that it affects a broad range of configurations. That when vsync is enabled there is little to no microstutter (I had found this with my own testing but knowing it was the same for other setups was interesting). Also, I discovered that when the GPUs were not working at 100% capacity then microstutter was drastically reduced.

I read the bit about the "effect" being more noticable when the GPU's where 95%-100% loaded and less in effect when CPU limited . . . due to the GPU's being able to sync up their output? . . . but how do you know this in fact is not actually something to do with an overclocked system? . . . i.e when the system is overclocked something related to the graphics subsystem becomes a bit borked/out of range etc? . . .

As far as the CPU limitation issue. Don't think of it that way - the reduction of microstutter occurs whenever the GPUs are not working at full capacity. Whenever they have finished their workload and are waiting for data from another source they become synchronised to that source. In the case of vsync, the output is syncronised to the time the framebuffer clears and the GPU can start working again, which is in turn tied to the monitor refresh rate. In the case of CPU restriction, the output is syncronised to the output from the CPU, which is needed for the frame rendering to begin or proceed. In each of these cases, the output is regular.

It's just the mechanics of how the hardware works in conjunction. It's interesting to see that it holds true in the physical scenarios I've been able to test, whether directly or indirectly.

If someone runs two GTX 460's on a stock Intel Core i5 system and loads up a game and plays at 1920x1200 with 4xAA (or 8xAA etc) are you guys suggesting the end user won't have a lovely smooth experience? . . . assuming say the benchmark for this config was approx 60FPS? . . . would everything look nice and smooth etc?

I'm didn't say anything of the sort, and I don't have any interest in the performance of the card in any individual game.

All I'm trying to do is point out that this phenomena exists in multi-GPU systems, and until it is sorted people looking to buy a multi-GPU system should be aware of them. Adding a second card will still improve real-world gaming performance over a single card of the same type. That's for sure. But, the real-world improvement is not going to be as much as you might expect from average FPS benchmark scores.

It's no different to pointing out that minimum framerates might be worth considering as well as average scores. Or that the GPU memory use at different resolutions would be worth reporting. It's just another facet to the complicated work of computer hardware.

 

[Duff-Man]

That's totally uncalled for Ejizz! . . . if there is a problem here that can be "proved" then there is scope for legal proceedings to be made under the trade descriptions act . . .

Don't be daft

There is no case for legal proceedings

There is no law to force any GPU to perform any better by adding a second. Lots of GPUs are sold on the assumption, which is sound in almost all cases. Microstutter is just a subtlety of the technology. It doesn't mean that technology's invalid, it just means that measuring average-FPS benchmark outputs isn't the best way to assess it. But that's down to the end user.

If enough people are aware of it, who knows, maybe the driver teams at ATI and nvidia will start to take it more seriously as an issue. As it is, they make their sales based largely on average FPS benchmark performance. They have no real need to address microstutter.

 

[Big.Wayne]

The Op started a thread asking for advice, the Op was given lot's of different advice.

Situation might be interesting when 2 cards in SLI are cheaper and faster than strongest card on the market #5

When 2x mid-range cards are both faster and substantially cheaper in combined cost than the current top end single card. (Or poor availability of the top end card #6

Yeah this roughly my view on SLI aswell #10

So the general consensus is to go with a single GPU? #13

Yes, unless the performance of SLI lower ends beats a higher end equal to roughly the same cash #14

I don't have any favourites, whatever gives me best bang for my buck. #15

After reading the advice and with strong evidence presented to the Op, he stated he was going to take the single GPU option

"Strong Evidence"? . . . the only strong evidence presented in this thread was in post #17 which shows the "fact" that SLI'ed GTX 460's outperform most of the single cards by a decent margin, and in the case of the HD 5870 costs less ££ too?

You then suggested that the "facts" presented were somehow not real? . . . you presented no evidence apart from your fingers on a keyboard? . . . even the link from XS forums does not constitute proof that the benchmark "Facts" are in someway not real? . . . if they are not real then why are they published? . . .why does every professional website not mark the figures down to represent this "rule of thumb" your trying to impose on someone asking for advice?

It will not do to "attack" someone because they do not point blank believe you . . . I asked you politely and now somehow you are miffed you haven't been able to backup something you have said?

Following this you to basically called the Op a sheep and suggested he wasn't capable of thinking for himself, yet you have the cheek to call me 'unfriendly'!

The O.P stated the following?

So the general consensus is to go with a single GPU? #13

This is basically going with the herd no? . . . going with consensus is basically not thinking for yourself? . . . right or wrong?

What makes it even more strange is that the "consensus" happened to state that they would all go with SLI if the conditions were right? . . . I'm just baffled!

I would ask you kindly *not* to make trouble for the sake of it, Rhodan is perfectly able to let me know himself if he has taken offence . . . which I hope he hasn't because it wasn't intended . . .we are just looking at the technical facts and you really don't need to be so unfriendly just because I don't take everything you say as gospel! . . .

You would prefer he choose an OC'd 5850 or cheapest 5870 for whatever reasons you have and seem to taken offence to both Radeons appearing to be inferior to SLI'ed GTX 460's in the benchmarks? . . . . its really no reason to get personal with me . . . really no reason at all!

 

[Big.Wayne]

[Off Topic]

I often find myself agreeing with what you're saying, but I have to be honest, you come off as a bit arrogant some times in the way you put your points across.

You know, you really should avoid making "assumptions" and pointing your finger at your fellow forum members? . . .

We should be talking about the topic, not me:

To SLI or not?

I would prefere if people avoided making person attacks because they cannot deal with technical conversations . . . .

 

[Anewbe4u]

[Off Topic]


You know, you really should avoid making "assumptions" and pointing your finger at your fellow forum members? . . .

We should be talking about the topic, not me:

To SLI or not?

I would prefere if people avoided making person attacks because they cannot deal with technical conversations . . . .

OFF-TOPIC

Are you asking me a question?

You kind of prove my point though in this post so I'll end it there with the disgusting personal attacks, as far as I'm concerned, this is finished, if your posting style continues to annoy, I can just ignore. No harm to anyone then.

ON-TOPIC

Get a 480 - This one. Should be cool and quiet...

http://forums.overclockers.co.uk/showthread.php?t=18196201

 

[Metroid]

-When 2x mid-range cards are both faster and substantially cheaper in combined cost than the current top end single card.

I second this rule.

 

[Big.Wayne]

Hey Duff-Man

Nothing wrong with playing devil's advocate.

indeed . . . it actually makes us better and stronger as a group although I get the feeling some people don't understand this!

If one doesn't agree with someone else on a technical GPU matter these days . . . before you know it your doing anything but talking about technical GPU matters!

If someone runs two GTX 460's on a stock Intel Core i5 system and loads up a game and plays at 1920x1200 with 4xAA (or 8xAA etc) are you guys suggesting the end user won't have a lovely smooth experience? . . . assuming say the benchmark for this config was approx [90]FPS? . . . would everything look nice and smooth etc?

I'm didn't say anything of the sort, and I don't have any interest in the performance of the card in any individual game

Ok this is a bit confusing . . . because you stated this? . . .

The hardware would be 90fps. This is 100% true. The subtlety is that the 90fps be output irregularly, and so would look more like a game scene running at say 75fps, (for a 15% microstutter readout). The reason for this is that the human eye notices the gaps between frames as the basis for smoothness - it doesn't just count the raw number of frames like a computer. In your 90fps case the game would still seem very smooth, since 75fps from a regular output source would still look smooth...

To me what is being "suggested" is that there is a users-experience performance "hit" that is shadowy and generally offputting to anyone considering going the bang-forbuck SLI route! . . . what is strange about this is if what you are suggesting is true it is a kinda FPS "Stealth Tax" which is hard to gauge actual performance . .

According to a new friendly forum user: While the FPS of 460 SLI is impressive, with multiple GPU's the perceived smoothness doesn't translate well in terms of FPS like a single GPU does.
As a rule of thumb, you can take SLI/Xfire Fps and reduce it by 10-30% to get an indication of real world performance... #19

I hope you can understand why this is a problem Duff-Man? . . . if its true (which I'm not convinced yet it is) then it means any performance chart like what is in post #17 is not valid . . .

There is no case for legal proceedings

If a product does not function in the way it is meant to function and produce the results its meant to then there is a case? . . . for example referencing the bench results in #17 Alien vs Predator (DX 11) 4xAA would you say anyone could tell the difference betwen the SLI'ed GTX 460's and the HD 5870 in actual use? . . . or would this "SLI Stealth Tax" make it seem slower . . .

There is no law to force any GPU to perform any better by adding a second.

 

[Big.Wayne]

If you can't wait, get some 460's is you arn't sensitive to micro stutter, and are not bothered by the odd SLI issue.

Micro stutter?

^^Skipping slightly, not smooth

If I was someone that didn't know any better and I read what both Ejizz & arc@css wrote above I think I would make the same decision as Rhodan . . .

"Scaremongering" is manipulation . . . .

I've decided to go with a single GPU.

 

[An Exception]

If I was someone that didn't know any better and I read what both Ejizz & arc@css wrote above I think I would make the same decision as Rhodan . . .

"Scaremongering" is manipulation . . . .

Ouch! Soooo unfriendly!

 

[Big.Wayne]

Oh my good god! . . . . LOL I don't believe it!

Is that it? . . . is that what you guys are calling "Micro-Stutter" on the jerky bumpy HD 5970 video playing in the right hand pane!

That's classic . . . I first noticed that years ago when I had a Radeon X1950 Pro . . . really bugged the **** out of me, I was like WTF is going on here . . . it turned out to be either CPU-z in the background or perhaps FRAPs interfering with the rendering! . . . turned them all off, rebooted and was smooth as butter!

Same thing happened to me when I got a Radeon HD 4850, one minute it was smooth and the next day . . judder, judder, judder . . . again turned out to be a combination of ATI-Tray Tools, FRAPS and RIVATuner . . . Turn all that junk off, reboot and you should be smooth as silk again!

 

[An Exception]

^^^
Originally post wrong vid...

 

[coupe69]

[Off topic]

I tire of people being unfriendly just because I ask for proof?

Awwww Poor little lamb, my heart bleeds for you Hardly helps that you come across as an arrogant, slightly patronising person though does it
I could possibly understand your passion (or is it madness ) more if you'd actually got some of the hardware and benchmarked it yourself.
The glorified pie charts and smart alec remarks would perhaps then seem a bit more worthwhile......maybe?


Remind us all again, what is it you're running? 4850 is it? 512MB or 1GB? Athlon dual core? I do seem to remember you're OS is XP though.....Very retro no DX10 or 11 for you then
So what we have here is somebody that cannot contribute with anything other than fancy graphs and bar charts "print screened" from various websites, bravo fine sir, thanks for the effort!
When you've actually got some real world experience with the hardware you're pretending to be an expert on then just maybe I'll start to take you seriously.

[Back on topic]

I've gone down the SLi route a couple of times, it's good but I'd still rather have a single great card than a couple of average ones.
Plenty of horsepower with minimum fuss = more time playing games instead of messing around with SLi/CF profiles, drivers etc etc.

"Scaremongering" is manipulation . . .

LOL, how do you know it's scaremongering when you've never experienced the hardware in the real world?

 

[An Exception]

Damn.... now that's what I call 'unfriendly'!

 

[Big.Wayne]

• I don't want something that is going to be noisy and hot (GTX 480).
• I'd appreciate your advice
• especially those of you that have GTX 460's in SLI

Hey Rhodan,

as your not buying your graphic solution for a while yet I think you have time to get some homework done and weigh up all the pro's and con's . . . your in a good place right now to be considering your GPU options . . . all I ask is that you consider the "facts" and make your own mind up about which solution is right for you and not be driven down a path that suits others instead of what suits you!

Anyway I totally can see why you would not want a GTX 480! . . . It's not often you see a more affordable set of GPU's beating out a more expensive and older GPU in some cases by up to 50% lol!

Not sure spending the extra on the SLI'ed 1024MB cards is the best bang-for-buck atm but anandtech benchdata doesn't have any 768Mb results so . . .




SLI'ed GTX 460's . . . the choice of any "serious" gamer in OCT 2010 who has deep pockets but still wants value for money £££