Ok, well, I've done my best to clean up this thread. What a mess you guys have made
Sorry for the cases where I deleted posts that were non-offensive, but the reason I did so in most cases was because they contained quotes of offensive/ToS-violating posts.
As someone who's a bit of an expert on the subject of physX (and I think most who know me will acknowledge my expertise) I'm going to leave you with my own words of wisdom
First off, Dagan, Birdman,and Thrasher have the right idea, and Johnny Doe, I'm sorry, but you're wrong in much of what you've said , and your posting style of declaring people "WRONG" at the beginning of many posts, and ridiculing others on the boards ... is simply not how we interact with one another here on OCN
(for anyone confused about this statement, understand that much of what JD has posted has since been deleted).
The simple fact of the matter is that PhysX calcs and graphics are NOT separated.
As MANY posters have tried to explain to JD, the reason that increased rendering power in your 'primary' GPU's affects the amount of physX 'power' you need in your dedicated physX card is very simple ... it's because with more rendering power, you expect more FPS.
IOW, the more rendering power you have in your main card(s), the higher the FPS, and hence, the faster your dedicated physX card needs to be in order for it to be able to 'keep up', i.e. to provide the physX calcs as quickly as your rendering cards need them to be done.
Ironically, the very FPS diagram JD posted provides a perfect illustration of the exact opposite of what he himself is claiming.
First off, unfortunately, they did not do 3x260SLI + 260 for physX in this test. That basically renders the 3xSLI results entirely irrelevant, because we lack apples/apples comparison between the 240 and the 260 in terms of physX perf with tri-SLI 260's.
So let's look at the results here that actually DO provide illustrative data, which is the 1xSLI vs 2xSLI data, comparing the 240 against the 260 for physX.
You need only observe the following numbers, because these are all that are relevant from this diagram:
Single 260 + GT240 = 51fps
Single 260 + GTX260 = 51fps
Difference in performance going from 240 to 260 for physX, w/single GTX260 = 0%
SLI 260 + GT240 = 80fps
SLI 260 + GT260 = 88fps
Difference in performance going from 240 to 260 for physX, w/SLI GTX260 = 10%
These results 100% and absolutely back up the point that Birdman and the others are making. The more powerful the primary card(s), the more benefit is derived from having a more powerful physX card. And this is because physX is NOT separate from rendering. The faster that your primary cards are rendering, the more physX power you need in order to 'keep up'.
If JD's basic claim was true that physX is entirely separate from the rendering task, and that the GT240 is ALWAYS enough physX power, we simply WOULD NOT SEE the increase in performance for the GTX260 vs. the GT240 when we went from a single 260 to SLI 260's. But WE DO.
It is PRECISELY because the FPS was raised between a single 260 and SLI260 that the performance of the GT240 for physX begins to 'fall off' vs using the GTX260 for physX.
And I can 100% guarantee all of you readers that if there were results on this graph for Tri-SLI 260's + GTX260 for physX, we would see that the improvement for the 260 vs the 240 would be EVEN MORE than the 10% improvement seen in 2xSLI. Because the FPS would higher, and the physX performance of the GT240 would fall even further behind the performance of the GTX260.
Now ... moving forward, I expect the dialog in this thread to remain CIVIL, and profanity-free. Also, the personal attacks/name-calling between members needs to STOP. Otherwise, I will be closing this thread. Do we understand each other?Edited by brettjv - 11/2/11 at 12:36am