7870 or 660ti?

[xxhellraiser]

Hello, Guys i am wondering which one will be the best to buy and which one will out preform the other and buy how much % thanks

http://www.tigerdirect.ca/applications/SearchTools/item-details.asp?EdpNo=2316882&CatId=7387
or
http://www.canadacomputers.com/product_info.php?cPath=43_557_559&item_id=051482

 

[sam_p_lay]

Is my personal thought, nothing else comes to mind cause of the test methodology they follow, the hand picked/limited amount of games they chose to "prove" something. I'd expect you, at the first, would be approaching the issue with doubt and with a more scientific aspect, questioning the consistency and validity of that test. I showed you how inconsistent/confusing these tests can become and i can show more if you ask

There's actually a pretty decent selection of games there - THG don't typically test on many more games than that? As for previous tests showing inconsistency with both companies, that's exactly what I addressed in the first paragraph ^ 🙂 I certainly wouldn't suggest GeForces have always been smoother - it's a recent development.

As I say, my best guess is that nVidia's recent driver updates have fixed GeForce inconsistency, but it's anyone's guess. I'm 100% with you that more data is needed, and it is on the way. However, plenty of people are buying now and not waiting for new product lines or new benchmarks. For those people, we just have to go with the limited data we have.

iam2thecrowe is totally right about TR's bias (or lack thereof). You've read all their older roundups so you know they don't have a history of bias. Where nVidia previously delivered less consistency, TR didn't hesitate to highlight that. I really think people need to stop and think before immediately making allegations of bias, bribery etc. That's not something to be suggested lightly or without evidence.

 

[revro]

well i switched from ati to nvidia first time to evga 600 ftw. actually only cause i already had FC3 and wanted to buy AC3. problem i see with amds are they are underclocked and they want you to overclock. ftw is factory overclocked and has much bigger clocks than amds ain that region. i guess my 660ftw ~ 660ti cause i am really getting great fps with my q9550 and for 1080 its enough. skyrim is over 60, bf3 on 16p maps 50-60fps, on 64 35-60 but mostly averages 42-45. i guess its cause of cpu

pricewise 660ftw was 235eur, 660 ti are around 300 eur and 670 375+, so since ididnt wanted to drop any more money to my lga775 system i did, 35eur 4gb ram, 133eur q9550 second hand, 241,5eur evga 660ftw 3gb -50eur from sale of my old 5750 card

best
revro

 

[technoholic]

iam2thecrowe is totally right about TR's bias (or lack thereof). You've read all their older roundups so you know they don't have a history of bias. Where nVidia previously delivered less consistency, TR didn't hesitate to highlight that. I really think people need to stop and think before immediately making allegations of bias, bribery etc. That's not something to be suggested lightly or without evidence.

I over-reacted, that's about how seriously i take this, and anything other than this accusation is still valid in my eyes, including the bias. It is getting clearer when there is no other card, for the sake of comparison. And every single NV fanboy linking to that article and trying to claim it is the ultimate truth about GCN cards. Almost everything about this article vague. It is also still vague how much this affects gaming experience, especially if someone didn't have the opportunity to even play Skyrim on a 7950. The article managed to also show 7950 in fact worse performer at FPS side, which is also not true. Ignoring the potential of 7950 is something, saying that card is flawed is another thing. At the end no one can claim that Nvidia makes more quality hardware, this would be ridiculous and biased, considering the consistency of tests to prove it. Anyone wants to claim this? he must first play on a R7950 and than on a 660Ti and experience it himself

 

[sam_p_lay]

Well the experience thing they tried to show with the high-speed cameras. It's obviously a massively controversial article because it upset so many AMD fans (and especially 7950 owners). TR were under pressure to deliver proof, which they have.

And as I said before, the inconsistency of the 7950 is not the point. It's unexceptional - it's been demonstrated on other Radeons and on GeForces too. What's exceptional is how smooth the GTX660 Ti was. It's a first for either company. My best guess is that drivers are responsible (since the hardware hasn't changed) and that would suggest the entire range benefits (maybe even all ranges). But nobody can guarantee that for sure without more testing.

An angry AMD fan made a comment recently that you'd need 'cybernetic eyes' to spot the latency and that milliseconds are a timing faster than people can think/see. I can totally see where he was coming from and I think most people would at least consider this. Surely our eyes/brains deal in seconds and only electronic systems can deal in milliseconds?

But think about it. If our eyes/brains had a 'refresh rate' of one second, we'd all be gaming happily at 1fps. The fact that we need a minimum of 30fps (and many people would argue as much as 60fps) to create the illusion of fluid motion means our brains are operating at a minimum of 1/30th of a second, which is 33 milliseconds.

 

[technoholic]

The video is not a proof since at first it is a slow motion movie which can be affected by many many factors. At first, 240/120 FPS to 25-30 FPS transition was made. Heck it is recorded by a video camera sticked in front of a monitor, at last. In fact, when i look at the video, the only thing i can see is that GTX 660Ti skips frames more often (if it does oO) and HD 7950 does that less often. In fact, even if they put there 2 exactly same movies side by side (changing their brightness only) and start them 1 second apart from each-other, you probably wouldn't understand it. Selective perception, you see? You talk about physical perception, but there is the psychological side which affects physical side, you know. The "proof" can thus be perceived differently, even if it the test was done rightly. I mean anyone looking at that video may actually see HD 7950 as the better/smoother card. Frame latency might be relevant and even may be more important than FPS. But the nature of this test is harder to measure than that of FPS. Until a more relevant/solid test is established, this is still vague and about perception. And apart from the difficulties etc, if we want the different reviewers must do these tests

 

[sam_p_lay]

I know what you mean... I clearly saw large jumps forward with the Radeon and not with the GeForce, but maybe it was something to do with psychology. Somebody, grab your nearest non-gaming friend/family member and ask them which is smoother 🙂

As for the difficulty in understanding the test methodology, I think it's a pretty weak argument... TR have done an excellent job of explaining it in a very easy-to-understand way - it simply requires a reader to take the time to read a single page of explanation. We wouldn't have advanced very far as a species with the "too complicated - screw it" attitude 🙂

 

[technoholic]

As for the difficulty in understanding the test methodology, I think it's a pretty weak argument... TR have done an excellent job of explaining it in a very easy-to-understand way - it simply requires a reader to take the time to read a single page of explanation. We wouldn't have advanced very far as a species with the "too complicated - screw it" attitude 🙂

Mate, i didn't mean it. Arghhhh this English is hard sometimes as it is not my native language What i mean is about the standardizing of the process and using video as the proof. It is about testers side. Nothing wrong with this testing method, in fact frame latency must be standard and must be done by more and more reviewer, as long as it is made in a standard/common way. The gaming experience is about senses and experience. But scientific tests, if involves proving something, must be done in well established methods and with common process and wider range of data

 

[sam_p_lay]

Mate, i didn't mean it. Arghhhh this English is hard sometimes as it is not my native language What i mean is about the standardizing of the process and using video as the proof. It is about testers side. Nothing wrong with this testing method, in fact frame latency must be standard and must be done by more and more reviewer, as long as it is made in a standard/common way. The gaming experience is about senses and experience. But scientific tests, if involves proving something, must be done in well established methods and with common process and wider range of data

Sorry, looks like I misunderstood your point there! Yeah I completely agree about the methodology. If you look over TR's older articles you'll see they were previously looking at number of frames above a certain threshold, which was a bad system, since it actually penalised high frames (since of course faster cards render more frames in total over a given period of time). At a reader's suggestion, they switched to time spent above the threshold (as number of slow frames is less important than how long you're seeing the slow frames).

I think the system they have now is sound. Although the thresholds chosen are arbitrary to some extent and affect those results considerably, they do now give the reader buttons to choose from 3 different thresholds. And the thresholds do line up with logical framerates anyway (50ms being the slow time to line up with 20fps, 16ms being the fast time for 60fps and 33ms for the 30fps equivalent).

Still, we're only looking at one half of their methodology/results here - I'd put most stock in their 99th percentile results. There's no decisions being made with those about thresholds or anything like that - it's purely a measured result.

 

[technoholic]

Standardized and well-established tests are always welcome. And more sources, different hands/aspects are needed to do these common tests, leaving little/no error margin and leaving no doubt at reader's side. Frame per second, as much as it is criticized, is the common tool for measuring the performance differences now. But i do wish that frame latency be added to the picture as a common method. Now if they conclude this with a video like this (meaning it is the ultimate evidence), it will always become open to criticize and people with different perceptions will get it differently. We shouldn't be discussing how the test is made, we should discuss the outcomes/results

 

[sam_p_lay]

In defence of Techreport, I don't think the video was ever the actual test. It was just used to help illustrate the results of the test. Maybe to provide additional evidence to back the frame latency benchmarking. But the video isn't the test itself, just a little extra. When the Radeon 8000 cards and GTX700 cards launch, I think there will be much more frame latency data on those for us to look at, from different sites.

 

[technoholic]

Yea video is the part of their evidence, proof to show what they mean by smoothness. But the point is that, it doesn't show what they mean clearly. I don't think anyone, without any gaming experience whatsoever, could distinguish between 2 sides. Not the best way to define what they mean i guess. As to what can be done/used to provide evidence for something like this, i didn't think about it nor i am an expert of these testing methods. I just want to see scientific evidence to prove something. Only that time, we can focus and discuss the outcome and how to cope with/avoid a certain issue

 

[sam_p_lay]

The forward skipping was pretty noticeable to me... did you watch the whole thing? There's a few seconds between each skip. Anyway I think the OP probably has enough to go on now 🙂 If they come back, they're gonna have some reading to do! I'm not sure there's much more to say about frame latency that would be useful now. As always though, feel free to PM me.

 

[technoholic]

I saw almost all video 240 fps and the 120fps one, yes there is slightly skipping forward i saw that at the both sides, difference being 7950 skipping once every few seconds (corrected minutes to seconds lol) and 660Ti doing this more often. Been questioning if i'm not so careful or miss something and if this slow-motion can be applied to actual game play. This is the special thing about these forums, we share our aspects and see different aspects of eachother and everyone can see it, nothing personal you know. If i have things to ask in my mind i will pm you, you know i do

 

[sam_p_lay]

Absolutely mate 🙂 Just this thread is getting pretty huge now... not sure if the OP is even coming back - they haven't posted a second post.

 

[whyso]

I think that the techreport frame latency thing should be ignored. How do they have the 7950 losing to the 660 ti in skyrim?

Anandtech has the 7950 beating the 660 ti 85 to 69 fps. http://www.anandtech.com/bench/GPU12/459

Toms has the 7950 beating the 660 ti 64.6 to 50 fps (this is with 8x MSAA) http://www.tomshardware.com/reviews/geforce-gtx-660-ti-benchmark-review,3279-9.html

Yet somehow it loses in techreport 72 to 62 fps.

 

[sam_p_lay]

Ignoring facts because you don't like them isn't an intelligent approach to life...

As for the differences in framerate, are they using exactly the same CPU etc? Are they using exactly the same graphics settings? And above all, are they testing the same part of the game? It's been very well demonstrated that one game can give very different results depending on what part you're in. For an extreme example, compare framerates in any game with a simple, empty room to a wide open area with loads of enemies and combat going on.

EDIT: The really crucial thing for fair testing is that the same spec, settings and test are used for each card. As long as nothing else changes, you've got a level playing field. That's why you can't compare results from different sites, unless all those variables are the same.