jjb8675309
Splendid
- Feb 21, 2010
- 5,911
- 0
- 26,160
- Status
As far as WoW goes, yes its hard to benchmark. If we talk about just getting high framerates with high view from fliers etc, im not sure what to suggest. However if you actually raid, then most likely what you care about is high framerate with lots of people around.
My experience with this shows following rules:
- *IF* you play in 2560/1600 with AA like me, frame buffer matters - and over 1GB seems to help a bit.
- Other then that, any card past 5830 will most likely be CPU bottlenecked. Unless you run a heavily overclocked 980X, dont bother upgrading video card past lets say 5850 for wow.
- Crossfire has absolutely zero effect, and SLI has *maybe* marginal increase - im not too sure about that either. The game simply doesnt scale with multiple GPUs, which can be attributed either to the engine, or possibly to the point above - the game is CPU bound far more - and sometimes adding another card might even reduce FPS by increasing CPU overhead.
My experience with this shows following rules:
- *IF* you play in 2560/1600 with AA like me, frame buffer matters - and over 1GB seems to help a bit.
- Other then that, any card past 5830 will most likely be CPU bottlenecked. Unless you run a heavily overclocked 980X, dont bother upgrading video card past lets say 5850 for wow.
- Crossfire has absolutely zero effect, and SLI has *maybe* marginal increase - im not too sure about that either. The game simply doesnt scale with multiple GPUs, which can be attributed either to the engine, or possibly to the point above - the game is CPU bound far more - and sometimes adding another card might even reduce FPS by increasing CPU overhead.
hunter315
Champion
- May 12, 2009
- 10,652
- 3
- 51,960
Wonder what the results would look like without farcry 2 and SC2 both of which suffered obvious limitations that were not caused by the GPU. I also would have liked to see a % scaling on average for each type of card rather than everything based off of a 5830.
vvhocare5
Distinguished
- Mar 5, 2008
- 768
- 0
- 19,060
To tell the truth, I dont know why these writers do these articles. For any test series, 80% of the writers simply cannot fathom the issues being discussed, the other 20% say they arent in depth enough (which is true of any test). 99% come back with injecting cost into a technical issue (at least my decision was cheaper, faster, blah, blah). Completely irrelevant of course.
And did anyone think about the normal issues in producing articles? Someone calls in sick, your manager says you turn around an article for tomorrows publication. Its an oh cr*p moment. So you flip through your data and realize you have something.
Yes I have made known my views that this site is getting dumber and dumber as they chase volume for their click thrus, bu it was a nice simple article about simple scaling between the two big boys.
Leave it at that
And did anyone think about the normal issues in producing articles? Someone calls in sick, your manager says you turn around an article for tomorrows publication. Its an oh cr*p moment. So you flip through your data and realize you have something.
Yes I have made known my views that this site is getting dumber and dumber as they chase volume for their click thrus, bu it was a nice simple article about simple scaling between the two big boys.
Leave it at that
- Mar 7, 2003
- 12,012
- 0
- 40,790
[citation][nom]scrumworks[/nom]Yeah. It would show up even better if Tom would use a bit fresher drivers than 10.7 betas.[/citation]
Really? I'd be very interested in seeing some proof of that, scrum.
I believe you might be making a huge assumption, there.
Really? I'd be very interested in seeing some proof of that, scrum.
I believe you might be making a huge assumption, there.
- Mar 7, 2003
- 12,012
- 0
- 40,790
[citation][nom]RazberyBandit[/nom]Simulated results are just that - simulated - not the real deal.[/citation]
I disagree.
I've done extensive tests to see if disabling a CPU core in Windows delivers the same result as a cpu physically equipped with less cores -- it does. Results verified and cleared with AMD. Same thing with graphics cards.
If the results are verifiably identical, the method scientifically proven and the engineers back the method, I'm very OK with that.
Frankly, your claim that simulated is somehow a negative thing--a claim without any factual evidence to support it--bears a lot less weight with me than a scientifically verifiable method.
If you have any evidence that there's a problem with the data, i'll be glad to look at it. But I don't think that's the case here.
I disagree.
I've done extensive tests to see if disabling a CPU core in Windows delivers the same result as a cpu physically equipped with less cores -- it does. Results verified and cleared with AMD. Same thing with graphics cards.
If the results are verifiably identical, the method scientifically proven and the engineers back the method, I'm very OK with that.
Frankly, your claim that simulated is somehow a negative thing--a claim without any factual evidence to support it--bears a lot less weight with me than a scientifically verifiable method.
If you have any evidence that there's a problem with the data, i'll be glad to look at it. But I don't think that's the case here.
- Mar 7, 2003
- 12,012
- 0
- 40,790
[citation][nom]WheelsOfConfusion[/nom]Hey Cleeve, what's up with the drivers being two versions (+ beta) behind for both companies? I didn't think these benchmarks took that long.[/citation]
The article was handed in a while ago. Unfortunately publication was bumped a number of times due to scheduling issues and items that needed to take precedence for one reason or other.
I'm probably more unhappy about it than anyone else. Having said that, I play with these drivers on a regular basis and I haven't seen anything to indicate that the brand new betas would make any measurable difference. (barring AA in sc2 as mentioned above, assuming it makes a difference. Even if it does it's something so new it wouldn't have made it into this article anyway, usually takes at least a couple weeks to be written, edited, posted, and scheduled).
The article was handed in a while ago. Unfortunately publication was bumped a number of times due to scheduling issues and items that needed to take precedence for one reason or other.
I'm probably more unhappy about it than anyone else. Having said that, I play with these drivers on a regular basis and I haven't seen anything to indicate that the brand new betas would make any measurable difference. (barring AA in sc2 as mentioned above, assuming it makes a difference. Even if it does it's something so new it wouldn't have made it into this article anyway, usually takes at least a couple weeks to be written, edited, posted, and scheduled).
I understand benchmarking takes time, but why use such old drivers? I wanted to read this article until I saw the old drivers, which made this article DOA because of the major increases is scaling, especially on the ATI cards. Oh well, good try I guess.
RazberyBandit
Distinguished
- Dec 25, 2008
- 2,303
- 0
- 19,960
Out of all the comments, you chose to respond as such to that one? Why not re-run the test using newer, non-beta drivers and offer some clear proof yourself? After all, you're the one with all the hardware whose results were questioned...
I guess someone pointing out that testing any graphics card using dated driver software is unrepresentative of actual current-day results isn't sitting too well with you, then. If it were, you would have refuted it with facts, not a question or challenge of your own.
Edit... I see you replied to others as well. Suppose you did that as I was typing. I'll take your word from AMD, as I find you reliable as a source, but only when it comes to the concern over CPUs. In regards to GPUs, I'm not convinced, especially in the case of the ASUS Ares since a dual-GPU board uses only one interface, whereas an actual CrossFired config would use two.
- Mar 7, 2003
- 12,012
- 0
- 40,790
[citation][nom]RazberyBandit[/nom]Out of all the comments, you chose to respond as such to that one? Why not re-run the test using newer, non-beta drivers and offer some clear proof yourself? [/citation]
I'll be happy to explain.
Primarily, a new driver iteration doesn't make a big difference except in extremely rare cases. Almost never. Anyone who pays attention to actual performance differences in drivers knows that performance stays quite similar over long periods of time, the exception being when there is a really bad performance issue that needs to be fixed in a specific title.
I have seen cases where the driver responsibility myth has even been perpetrated by reps from the graphics companies, who at the last miniute demand that we use a new beta driver and retest days of work. We comply in good faith to discover ***ZERO*** difference in results... none. Nada.
Yet, after all of this, after countless articles on the web demonstrating that driver updates so rarely make any difference at all, it's always the first thing people shout if the results don't jive with their favorite sports team/graphics company/etc. "It's the drivers, that's why my preferred manufacturer didn't win!" It's an easy out.
Those of you know know better already realize that it takes a lot of driver iterations to make a lick of difference, and even then a difference is not guaranteed.
On the other hand, folks more interested in attaching their self esteem to one graphics company over the other--instead of simply allowing themselves the open-mindedness to actually learn something--may continue to place all of their woes on the driver boogeyman.
So, to answer your question, I am not interested in wasting time re-testing hours of work based on a complaint with absolutely no evidence that a newer driver will make a hair of difference.
I'll be happy to explain.
Primarily, a new driver iteration doesn't make a big difference except in extremely rare cases. Almost never. Anyone who pays attention to actual performance differences in drivers knows that performance stays quite similar over long periods of time, the exception being when there is a really bad performance issue that needs to be fixed in a specific title.
I have seen cases where the driver responsibility myth has even been perpetrated by reps from the graphics companies, who at the last miniute demand that we use a new beta driver and retest days of work. We comply in good faith to discover ***ZERO*** difference in results... none. Nada.
Yet, after all of this, after countless articles on the web demonstrating that driver updates so rarely make any difference at all, it's always the first thing people shout if the results don't jive with their favorite sports team/graphics company/etc. "It's the drivers, that's why my preferred manufacturer didn't win!" It's an easy out.
Those of you know know better already realize that it takes a lot of driver iterations to make a lick of difference, and even then a difference is not guaranteed.
On the other hand, folks more interested in attaching their self esteem to one graphics company over the other--instead of simply allowing themselves the open-mindedness to actually learn something--may continue to place all of their woes on the driver boogeyman.
So, to answer your question, I am not interested in wasting time re-testing hours of work based on a complaint with absolutely no evidence that a newer driver will make a hair of difference.
RazberyBandit
Distinguished
- Dec 25, 2008
- 2,303
- 0
- 19,960
Don, while I understand your stance and appreciate you taking the time to explain it, I find myself respectively disagreeing.
It seems an easy cop-out to blame fanboyism and some past experience as the reason why you didn't use more-recent drivers, let alone an official release driver in the case of the AMD cards, in this series of tests. As I pointed out, the new nVIDIA 260.89 driver notes lay claim to GTX 460 performance increases of 5-19% in four of the titles/resolutions specifically used within this test. While a 5% performance increase within a title in which a card can already reach 45-50fps may be hardly noticeable, a 19% (or close to four times the difference) increase in such a title is a rather significant performance boost. You could be talking about a 5-20fps increase depending upon the starting point within the title in question.
I find your point about people crying foul when you use older drivers another lame-duck excuse. The entire argument could easily be avoided by simply updating the drivers. It's not like doing so is some hugely complex process. And, by not doing so, you're just throwing out open invitations for the questions myself and others have raised.
I don't find myself siding with either camp. I want the best performance for my dollar, and I use products from both accordingly. As someone who wants to see a truly fair and objective comparisons, something as simple as the driver version used should never need be questioned. Why? Because they should simply be updated.
It seems an easy cop-out to blame fanboyism and some past experience as the reason why you didn't use more-recent drivers, let alone an official release driver in the case of the AMD cards, in this series of tests. As I pointed out, the new nVIDIA 260.89 driver notes lay claim to GTX 460 performance increases of 5-19% in four of the titles/resolutions specifically used within this test. While a 5% performance increase within a title in which a card can already reach 45-50fps may be hardly noticeable, a 19% (or close to four times the difference) increase in such a title is a rather significant performance boost. You could be talking about a 5-20fps increase depending upon the starting point within the title in question.
I find your point about people crying foul when you use older drivers another lame-duck excuse. The entire argument could easily be avoided by simply updating the drivers. It's not like doing so is some hugely complex process. And, by not doing so, you're just throwing out open invitations for the questions myself and others have raised.
I don't find myself siding with either camp. I want the best performance for my dollar, and I use products from both accordingly. As someone who wants to see a truly fair and objective comparisons, something as simple as the driver version used should never need be questioned. Why? Because they should simply be updated.
- Mar 7, 2003
- 12,012
- 0
- 40,790
I don't think you understand my stance at all, then.
My stance is that there is likely no huge performance difference that would justify the massive time investment, not that fanboyism prevents me from re-testing. More, newer tests are always inbound. If the claims are true they will be vindicated in subsequent testing. I don't believe that will be the case--not to the extent the driver notes claim-- but it'd be great to be proven wrong.
You can call it a cop-out if you like, but for me it's simply a matter of paying attention to my experience and a choice not to waste hours of valuable time. I invite you to investigate similar driver claims. If you don't feel like doing that investigation on my say so, then I can't blame you... you're no more guilty than I am.
- Jul 26, 2008
- 4,041
- 684
- 23,840
The true scaling / power of SLI is once you start using 3D vision. Having to render each frame twice requires a god awful amount of GPU power. Almost twice as much as single frame rendering. I would love to see Toms do some benchmarks with the high end SLI setups using typical scenarios and then turning on 3D-Vision. Would be interesting to see how SLI scales when it has to render more then just bigger frames.
RazberyBandit
Distinguished
- Dec 25, 2008
- 2,303
- 0
- 19,960
And my point is that your "is likely" reply is by no means a "no difference at all" answer. "Is likely" calls for assumption and speculation, something I thought of all people here at Tom's, you'd have the integrity to stand above, Don.
- Mar 7, 2003
- 12,012
- 0
- 40,790
To me, integrity doesn't mean you have to waste time just because someone tells you too.
To me, integrity means you stand behind what you've done, and I'm more than happy to do that. If someone can point out a flaw with the tests as carried out, I'd be very happy to investigate. I simply don't agree that using the previous driver is a flaw. I'm not going to retrofit every article I've ever written every time a new driver becomes available. I don't think that is reasonable.
I guess we'll just have to agree to disagree then... on the meaning of integrity, and a bunch of other stuff, too.
I'm ok with that. Peace!
blackmancer
Distinguished
- Nov 29, 2008
- 429
- 0
- 18,810
ruffopurititiwang
Distinguished
- Aug 26, 2009
- 65
- 0
- 18,630
rutoojinn
Distinguished
- Nov 5, 2009
- 47
- 0
- 18,530
Considering how much earlier ATi released their current gen cards I find it amazing how well they are able to keep up with Nvidia's card.
If you have more time to develop something the outcome usually is better performance.
This gen has been great to both ATi and Nvid since the performance is quite similar. Prices are the only thing to consider when purchasing a card.
If you have more time to develop something the outcome usually is better performance.
This gen has been great to both ATi and Nvid since the performance is quite similar. Prices are the only thing to consider when purchasing a card.
spentshells
Judicious
- Apr 1, 2006
- 8,555
- 294
- 33,990
- Status
TRENDING THREADS
-
News TSMC founder says Intel should focus on AI, not advanced process technologies- Started by Admin
- Replies: 30
-
-
-
-
Facebook
Twitter
Instagram