TechReport GTX 690 Review (Measuring MS)

[3DVagabond]

TechReport's much anticipated (well, by some of us anyway) GTX-690 review. Including their "Frame Rendering Times" benches that attempt to quantify Micro Stutter (MS)

 

[Tempered81]

Nvidia's framerate metering is a success with the 690. The MS isn't as good as single gpu, but definitely not as pitiful as frame time with CFX/SLI. It's actually closer to single gpu frame time consistency than CFX/SLI/6990/590.

690 is an awesome card. But the 780 is still the best video card of 2012. Single GPU all the way, when are we gonna see High-end Sea Islands or GK110?

If you have tons of money, dont waste it on multiple 680's or 7970's, or a 590 or 6990. The only smart choice in a multi-gpu world of bad decisions is the 690.

Great review.

 

[boxleitnerb]

Funny how he writes in the conclusion, that he is not sure about his own measurements. It's the same as with all frametime measurements currently - not entirely reliable.

The flip metering aka frametime metering is employed on all Kepler cards btw, not just the 690.

 

[Jaydip]

So they did improve MS on 690.

 

[Alaa]

Micro stutter is the thickness of the ripple, right? Like the variation of latency between each frame and the next one?

 

[Silverforce11]

Nvidia's framerate metering is a success with the 690.

You got that from an article that uses a technique to determine MS that even NV has said is invalid?? That's pretty good extrapolation there sir.

 

[Grooveriding]

I'd step up to a 690 from one of my 680s and sell the other one, if it were not for EVGA pulling a dirty and not allowing stepups to the 690. :thumbsdown:

Honestly screw EVGA. You can get a 3 year or more transferable warranty from lots of other vendors; MSI, Asus etc. They're not offering much of anything to differentiate anymore imo. Never mind the $25 per card charge to upgrade your warranty to one that allows step-ups.

Which I partook of, only to get screwed out of stepping up to the 690. :thumbsdown:

 

[HurleyBird]

I would like to know how much latency the frame metering adds. No such thing as a free lunch after all.

That being said, with any luck Big Kepler will be a beast. I can't see myself going multi-GPU again for awhile.

 

[Lepton87]

Multi-GPU set-up can't render a frame faster than a single card using AFR so you get more frame per second by working on a few frames simultaneously. If you have 4 cards but your single card takes 50ms to render a frame you will get much more frames per second but that 50ms has to go somewhere. Assuming perfect scaling you will end up with 80fps and 50ms of input lag.

 

[Lonbjerg]

I like the fact that they are playing with a high speed camera now.
Other reviewers need to take notice...or realize their reviews are flawed, outdated...and worthless

 

[Silverforce11]

I like the fact that they are playing with a high speed camera now.
Other reviewers need to take notice...or realize their reviews are flawed, outdated...and worthless

When they produce their MS analysis with a high speed camera is when I take their results seriously. All they've been doing is present fancy graphs that has no basis in reality and end user experience. Not only that, if they enable vsync and triple buffering, their frame time would appear "fine" to fraps.

 

[SirPauly]

Thanks for the link -- nice topic.

 

[SirPauly]

Funny how he writes in the conclusion, that he is not sure about his own measurements. It's the same as with all frametime measurements currently - not entirely reliable.

The flip metering aka frametime metering is employed on all Kepler cards btw, not just the 690.

I believe that may be the case but couldn't find any examples or concrete data on this. It was like pulling teeth to find data here -- and did receive "well, we're not ready to announce anything new about hardware frame metering at this time!" in the live review chat at PC perspective.