How fast are my GTX460 Cyclones at 900Mhz?

[Jacky60]

Quite fast!

 

[apoppin]

You probably made an accurate assessment. i am also not understanding AA's "respectful" comments. i just disagree with his assessment of SLI. i am right now - for the past week and ongoing for the next month - benching SLI vs CF.

Frankly i am *amazed* at the scaling that multi-GPU achieves. Anyone basing their assessment on past results is simply not up to date.

There are some games that do not scale (period). And generally, single GPU is sufficient - as with Wolfenstein. So it is not such a priority with Nvidia nor with AMD.

FACT: Wolfenstein is not popular. But i like it and include it in my benching as it represents OGL (along with ET:QW, also does not scale). Yet they both run fine on single GPU.
-- and for Wolfenstein 23>43 and 41>76 FPS average is a pretty good boost by adding a second GTX 560 at 25x16 and 19x12.

Most of my 29 test games scale. The older ones and the newer ones and generally, scaling is good (imo over 75% is good)

Some games scale well over 90%. No surprisingly, they are the ones that AMD and Nvidia optimizes for as a priority as they are POPULAR games.

I hate to butt in, but I think you two (Apoppin and ArchAngel777) are talking past one another in a sense.

ArchAngel777's point appears to be: "Sometimes SLI works very well, sometimes it works okay, and sometimes it doesn't work at all". He points to Wolfenstein as a simple token that represents "working okay".

Apoppin points to a larger sample size (than Anandtech's review data that Archangel777 uses) and argues: "On the whole, SLI works well. There's a game here and there where it doesn't work as well as it might but across this large sample of games it works well. By the way, Wolfenstein's poorer SLI scaling is dwarfed by the fact that SLI works well in many other games (also, nobody cares about Wolfenstein anyway)."

You're talking past each other in a way because it's sufficient for ArchAngel777 that at times, SLI is not as great as it could be. Even if you (apoppin) were to show him that in 95/100 games the scaling was 85% and above that probably wouldn't convince him. He just doesn't care for it as it stands now.

I think ArchAngel's entire point was to argue against happy_medium's 90% scaling quote. On the whole, they obviously do not scale 90%, end of story. Whether or not 80% scaling makes SLI worth it or 85% etc., is up to the consumer. But I agree with the sentiment of ArchAngel777's post that the data is important before we start tossing around "90% scaling" as if it's the norm. It isn't, though it can happen. Go figure that this all starts with some sensationalist post.

 

[happy medium]

I think ArchAngel's entire point was to argue against happy_medium's 90% scaling quote. On the whole, they obviously do not scale 90%, end of story.

No , ArchAngel used a release beta driver sample from a Anantech review, the review I posted from Tom's used a later driver and showed 90% scaling.
Please don't try to point out that someone is wrong till you get the facts straight.

But I agree with the sentiment of ArchAngel777's post that the data is important before we start tossing around "90% scaling" as if it's the norm. It isn't, though it can happen. Go figure that this all starts with some sensationalist post.

I didn't toss around, a review site did, and they used a more up to date driver.

 

[ArchAngel777]

No , ArchAngel used a release beta driver sample from a Anantech review, the review I posted from Tom's used a later driver and showed 90% scaling.
Please don't try to point out that someone is wrong till you get the facts straight.



I didn't toss around, a review site did, and they used a more up to date driver.

Let us just wait and read the review from ATB when it is finished. Then we can discuss the numbers. I have never been a fan of Tom's Hardware. A lot of their benchmarks were suspect in past. However, if AT did test with a poor driver, that would have merit as well. So, lets wait for ABT's review and we can discuss the subject further.

Edit ** FYI I find that the way you phrased yourself, Happy, is that you accused me of attempting to find a review that used a beta driver. I don't think that was fairly stated. I simply used AT's review, which happened to use 258.80 (BETA) driver. Tom's Hardware used 258.96 which would have been BETA at the time, and later released as WHQL. Please don't try and slant the discussion by these types of accusations. If nVidia had such a large difference in performance from 258.80 to 258.96, they would have asked AT to redo the tests.

On a side note, we are well beyond both drivers sets used for both of those reviews, so let us wait for ABT's review, which should have the updated drivers. Thanks, and please try to remain fair minded.

 

[happy medium]

Happy, is that you accused me of attempting to find a review that used a beta driver

not true, its a shame that most review sites only used the release drivers to test though.

 

[OCGuy]

That setup is much faster than a 580.

 

[Aristotelian]

No , ArchAngel used a release beta driver sample from a Anantech review, the review I posted from Tom's used a later driver and showed 90% scaling.
Please don't try to point out that someone is wrong till you get the facts straight.

First of all, I'd appreciate it if you read what I wrote.. Tom's using x driver and showing y scaling is not sufficient to conclude '90% scaling' per se. His review did not prove 90% scaling for a large sample of games, so it's misleading (and I'm being very charitable here) to utter the statement "90% scaling".

I didn't toss around, a review site did, and they used a more up to date driver.

Regarding what I just said (reading what I wrote):

I think ArchAngel's entire point was to argue against happy_medium's 90% scaling quote.

So, attempting to imply that I've made a mistake (where none was made) is precisely what you accuse me of doing to you. I had already stated that you were using a quote, looks like you missed that. Yet I am noting a mistake in that review (this isn't about you, other than you quoted a review that just doesn't do the work to support the claim you quoted). For those interested, here's the proof:

Tom's usage of 'all games' on his charts is an obvious reference to "all games in this review" (an enormous total of 5), and you can't conclude that because the scaling is high in those 5 games that it will be that high in all games. I hope this is clear to people, as seen through the differing levels of SLI scaling (that apoppin refers to in his post as fluctuating between 75% and higher than 90%, and I'll take his word over any other reviewer's any day of the week).

This is not an esoteric point, it's a simple point about entailment. I look forward to seeing ABT's review to see if the claim "90% scaling" holds more generally, but I bet that they have writers smart enough to write something like "and in these games the sli setups show scaling of up to blah" and not something stupid like "Because SLI scales so well, giving two GeForce GTX 460 1GB graphics cards a 90% performance boost over a single card..." without qualifying the statement appropriately.

In sum: Tom's review did not do the work to infer that from a set of X games that have high scaling this scaling will hold true for a larger set of games Y. Or even more simply: Testing 5 games does not give a reviewer the grounding to make an unqualified claim such as "The scaling is 90%". Rather, it should say "The scaling averaged to 90% in these games." Why? Because we end up getting one line quotes from the conclusion, taken in isolation, that do not have the relevant context. It's misleading, as I said before.

So, my facts are straight, my logic is tight, and the point I just made here about necessary entailment should be obvious. Note that the entirety of what I said above will be compatible with ABT's review when it says "SLI scaling is great, frequently above 75% and sometimes well over 90%", and even if it says "the average is over 90% in many games" - why? Take a wild guess as to the major difference between their reviews and other sites'.

 

[happy medium]

'.

OK then just say that no dual card setup has 90% scaling because the use of different games , drivers, cpu bottlenecks, and detail settings are all different.

If I were to guess at Apoppins results, I would guess the best scaling is expected with the latest more graphically intense games and 75% with the older ones.

 

[Aristotelian]

OK then just say that no dual card setup has 90% scaling because the use of different games , drivers, cpu bottlenecks, and detail settings are all different.

If I were to guess at Apoppins results, I would guess the best scaling is expected with the latest more graphically intense games and 75% with the older ones.

No, I'm sure there are games where the scaling is 90%, but to say it's 90% across the board just reads like cheap rhetoric to me when someone is basing that on 5 games, isn't my point clear?

I'd guess so too, so perhaps it's more important to have better scaling on modern games. I do take a bit of issue on the need to 'prioritize' the scaling optimizations with popular games, but I guess that's just a logistical problem.

 

[happy medium]

No, I'm sure there are games where the scaling is 90%, but to say it's 90% across the board just reads like cheap rhetoric to me when someone is basing that on 5 games, isn't my point clear?

I'd guess so too, so perhaps it's more important to have better scaling on modern games. I do take a bit of issue on the need to 'prioritize' the scaling optimizations with popular games, but I guess that's just a logistical problem.

yes, and the next time anyone uses the point that any x-fire/sli setup scales 90%, I'll just tell them no it doesn't, especially if you add in games that are 5 years old, that no one plays anymore, and can be played with one card at 90 fps, where its not even relevant to use sli/x-fire.

In short no dual card solution is going to ever scale 90% accross 29 games on average. Why? because by the time you test them, the older games get forgotten and lose performance to benifit new games in newer drivers.


I expect you to do the same when people use the 90% theory, that goes for the 69xx series also.

 

[apoppin]

Let us just wait and read the review from ATB when it is finished. Then we can discuss the numbers. I

Thanks. Sorry for any misunderstanding over what i wrote about SLI and CF.

Just realize that i am quite impressed; but then i have liked multi-GPU since HD 2900 Crossfire. There is some bias here on the side of pure performance.

However, it appears that i may have paid some HW price for it. My Thermaltake 775W PSU shorted out yesterday and shut down after benching and gaming for hours with GTX 560 SLI. The PSU is not underpowered as my system only peaks around 625 W from the wall but it appears that my MB is unable to supply enough power to the PCIe slots and my 12V (yellow) wires in my 24-pin ATX connector got hot enough to melt the plastic in the connector.

Testing is suspended until i contact Gigabyte about my x58 MB.
:'(

 

[Aristotelian]

I expect you to do the same when people use the 90% theory, that goes for the 69xx series also.

And I expect you to think about the reviews you quote a bit more carefully in the future, in addition to reading posts more clearly to boot.

 

[happy medium]

And I expect you to think about the reviews you quote a bit more carefully in the future, in addition to reading posts more clearly to boot.

Nice selective quote, I guess the rest of my post was too much for you to handle.


This is uncalled for.

Please read and reflect upon the following comments taken directly from our AnandTech Forum Guidelines:

We want to give all our members as much freedom as possible while maintaining an environment that encourages productive discussion. It is our desire to encourage our members to share their knowledge and experiences in order to benefit the rest of the community, while also providing a place for people to come and just hang out.

We also intend to encourage respect and responsibility among members in order to maintain order and civility. Our social forums will have a relaxed atmosphere, but other forums will be expected to remain on-topic and posts should be helpful, relevant and professional.

We ask for respect and common decency towards your fellow forum members.

^ those words are in our posting guidelines for a reason. We expect our forum members to approach the community with decorum and respect.

Moderator Idontcare

 

[Aristotelian]

Nice selective quote, I guess the rest of my post was too much for you to handle.

You snipped my 1000 word proof, and you're going to tell me that I'm the one with selective quotes? This is the third accusation you've made in this thread (towards me) and you've been wrong on all 3 (I've proven that, scroll back up).

A 5 game test does not make for meaningful support to "90% scaling". It's that simple. Anyone with a brain gets it, and your attempt to undermine my point by saying "well nobody plays older games any more" is frankly stupid. I'm not going to write another proof for you to discard it with some one liner fallacy that isn't beneficial for anyone.

Note that my posts are full of arguments, and yours are full of mere assertions. What do you have to say about my "cheap rhetoric" then?

 

[SHAQ]

yes, and the next time anyone uses the point that any x-fire/sli setup scales 90%, I'll just tell them no it doesn't, especially if you add in games that are 5 years old, that no one plays anymore, and can be played with one card at 90 fps, where its not even relevant to use sli/x-fire.

In short no dual card solution is going to ever scale 90% accross 29 games on average. Why? because by the time you test them, the older games get forgotten and lose performance to benifit new games in newer drivers.


I expect you to do the same when people use the 90% theory, that goes for the 69xx series also.

And there is most likely a huge CPU bottleneck as older games are predominately single core and CPU scaling isn't even close to GPU scaling the last 5 years. I can't think of any game older than Crysis that needs more than 1 high end video card from the last 2 years.

 

[Aristotelian]

And there is most likely a huge CPU bottleneck as older games are predominately single core and CPU scaling isn't even close to GPU scaling the last 5 years. I can't think of any game older than Crysis that needs more than 1 high end video card from the last 2 years.

http://alienbabeltech.com/main/?p=23490&all=1

1. Ctrl-f "The Benchmarks"
2. The list there: do you see many (any) 5 year old games?

I agree with your general point, but my point is to discredit the Tomshardware review, not to claim that SLI performance doesn't fluctuate from game to game (or become better over time as it is right now with the 6XXX and 5XX series)

 

[happy medium]

And there is most likely a huge CPU bottleneck as older games are predominately single core and CPU scaling isn't even close to GPU scaling the last 5 years. I can't think of any game older than Crysis that needs more than 1 high end video card from the last 2 years.

exactly, and Toms review used 5 newer games that push the cards.

http://www.tomshardware.com/reviews/geforce-gtx-460-sli-geforce-gtx-480,2694-2.html

 

[Aristotelian]

exactly, and Toms review used 5 newer games that push the cards.

http://www.tomshardware.com/reviews/geforce-gtx-460-sli-geforce-gtx-480,2694-2.html

And ABT used 4 of those 5 games plus about 20 others, making their approach far more informative to people who play more than 5 games.

 

[happy medium]

And ABT used 4 of those 5 games plus about 20 others, making their approach far more informative to people who play more than 5 games.

No arguments there, but this is the reason I used Toms review and believe 90% scaling to be accurate for more modern games.

Much more accurate then the Anand review with beta release drivers that Archangel used.

That was my point.

 

[Aristotelian]

No arguments there, but this is the reason I used Toms review and believe 90% scaling to be accurate for more modern games.

Much more accurate then the Anand review with beta release drivers that Archangel used.

That was my point.

This is what I mean about 'talking past one another'!

If Tom's review had said "In these 5 modern games the scaling is 90%" that's one thing, but he wrote "the scaling is 90%" - that's what I mean by unqualified. The way you wrote it in your post is something I wouldn't really have a problem with (as it seems that in modern games the sli scaling is getting better and better).

Personally I'm looking forward to ABT's review, as ArchAngel777 said this'll help us get a better picture for 'numbers on the whole' (and also with the numbers specifically for more modern games, that's the joy of their comprehensive reviews).

 

[RussianSensation]

I am not sure why you guys got into this heated argument so much. I already linked Xbitlabs' review where a 900mhz 460 was shown to be faster than a single 470. In addition, TechReport also showed 2x 810mhz GTX460s again beating 580 with ease. There is almost no question that when the 2 cards scale, 2x900mhz 460s will be a lot faster than a single 580. Of course when they don't scale, it's another story (but most people already know this anyway).

The bottom line a single 580 has 512 SPs with 48 ROPs and 64 TMUs, with GPU clock of 772mhz. 2x 460s have 672 SPs, 64 ROPs and 112 TMUs, with GPU clocks of 900mhz (as per OP). There is no way a single 580 can compete with that. That would be like stating that a single GTX580 is faster than 2x GTX470s in SLI.

Anandtech's own review already showed 2xGTX470s beating a single 580:
http://www.anandtech.com/show/4061/amds-radeon-hd-6970-radeon-hd-6950/13

 

[LiuKangBakinPie]

Here are my 2 Tallons

 

[Rifter]

my 460's @ 860Mhz are faster than a single 580 from most of the benchs i have seen. Obviously SLI scaling is not perfect in all games but as mentioned above in games old enough to not have good scaling a single 460 is more than enough to get 100FPS+ anyways.

 

[Serradifalco]

Here are my 2 Tallons

What are those Tallons clocked at?

 

[ArchAngel777]

Thanks. Sorry for any misunderstanding over what i wrote about SLI and CF.

Just realize that i am quite impressed; but then i have liked multi-GPU since HD 2900 Crossfire. There is some bias here on the side of pure performance.

However, it appears that i may have paid some HW price for it. My Thermaltake 775W PSU shorted out yesterday and shut down after benching and gaming for hours with GTX 560 SLI. The PSU is not underpowered as my system only peaks around 625 W from the wall but it appears that my MB is unable to supply enough power to the PCIe slots and my 12V (yellow) wires in my 24-pin ATX connector got hot enough to melt the plastic in the connector.

Testing is suspended until i contact Gigabyte about my x58 MB.
:'(

Mark,

No problem. I may have misunderstood your tone. I apoligize if I seemed, sensative. Moving past that, I hope you can get your benchmark rig back up and running. The results you post will have me re-consider my position. Currently, you can purchase 2 460's stock and after rebate you will have paid around $320. That is a pretty good price to be able to obliterate a $500 GTX 580.