sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
I’m back – this time pairing my GTX 690 with several Kepler card variants to determine the best PhysX PPU for the buck. The test beds: Borderlands 2 and FluidMark.  From left to right: GTX 650 Ti, GTX 650, GT 640, and on the bottom, the GTX 680. Test Rig Details: i7 3820 @ 4.5 GHz (1.40v) / EVGA GTX 690 (1148 MHz boost, +400 Memory, 135% Power) / 32GB Corsair Vengeance DDR3 1600 MHz / 512GB Samsung 830 SSD / Seasonic X-1250 / CM Scout 2 Case  Total PhysX Overkill!  Windows 7 Ultimate with SP1 and latest patches (10/15/12) – x64 NVIDIA GeForce WHQL drivers 306.97 FRAPS 3.5.9 Game / Benchmark Details: Borderlands 2 / Steam / Updated on 10/15/12 All graphics details maxed (PhysX: High) – 1080p (1920×1080), 120hz monitor FluidMark 1.5.0 Testing Methodology: Configurations: EVGA GTX 690 EVGA GTX 690 + EVGA GT 640 (dedicated PhysX PPU) EVGA GTX 690 + EVGA GTX 650 SC (dedicated PhysX PPU) EVGA GTX 690 + EVGA GTX 650 Ti SSC (dedicated PhysX PPU) EVGA GTX 690 + GTX 680 (dedicated PhysX PPU) Game / Benchmark Scenario: Borderlands 2: Bloodshot Stronghold – first 3 waves of enemies (each configuration tested 3 times, results averaged). FluidMark: PhysX Benchmark Preset: 1080p (each configuration tested 1 time). NOTE: FluidMark is not SLI aware. Results:    FluidMark Score FPS GT 640 3265 54 GTX 650 4096 67 GTX 650 Ti 4711 77 GTX 680 4472 73 Borderlands 2 Average FPS Avg. FPS Gain Minimum FPS Maximum FPS No PPU 96.48 - 63.33 172.33 GT 640 102.29 6% 51.67 205.67 GTX 650 112.77 17% 62.33 210.00 GTX 650 Ti 106.82 11% 69.67 239.33 GTX 680 109.10 13% 67.00 185.00 Conclusion: Borderlands 2 continues to be a bit baffling. Throwing a GTX 680 at it as a PPU does not yield crazy results – but a GTX 650 does surprisingly well. My hypothesis? Clocks matter. The GTX 650 SC from EVGA that I used is clocked at 1202 MHz whereas the GTX 680 was at it’s default clock of 1006 MHz. Even with that said – then why doesn’t the 650 Ti beat the GTX 680 since it is clocked at 1071 MHz? The only data that makes sense from these tests are the results from Fluidmark. The cards improve in a linear fashion based on how they are positioned by NVIDIA. Interesting results…but overall I’m sticking with the 650 Ti in my system – as it is the best all-around performer: value, efficiency, and performance. As always, the full logs from my benchmarks are available to download by clicking here. Source: my blog - http://1pcent.com/
post edited by sk3tch - 2012/10/16 10:46:01
|
Rei86
CLASSIFIED Member
- Total Posts : 3675
- Reward points : 0
- Joined: 2010/02/19 17:54:56
- Status: offline
- Ribbons : 9
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 10:27:13
(permalink)
Great write up! Bravo.
I guess the conclusion here is that the fall off point is probably the 660/660Ti.
Best bang for the buck the 650.
RIG 1 - i7-4790K / Asus Z97 Gryphon / AMD Sapphire RX Vega 64 / 8GB*8GB G.Skill Trident X 2400Mhz CL10 / SeaSonic Prime Ultra 1000w Plat RIG 2 - Pentium G3258 / ASRock Z97 Extreme9 / AMD XFX RX580 GTS XXX x4 / 8GB*8GB Kingston HyperX Fury 1600Mhz CL10 / Corsair AX1200i 
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 10:29:20
(permalink)
Hey thank you.
My conclusion is the best bang/buck is the 650 Ti. Merely because it performs well enough in Borderlands 2 and really does well in FluidMark. I look at FluidMark as a better indicator of future performance as a dedicated PhysX PPU - because Borderlands 2 seems to result in some interesting numbers, to put it simply. :)
|
Gold Leader
CLASSIFIED Member
- Total Posts : 3939
- Reward points : 0
- Joined: 2009/05/30 03:06:17
- Location: Dirksland, The Netherlands
- Status: offline
- Ribbons : 62
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 10:38:28
(permalink)
Could you run these tests with Metro 2033 please? That is something I play a lot the past few weeks and since it has good PhysX support as well I just wonder how that would do here  Nice tests for the rest, the GTX 650 is overall the best solution, the 680 is indeed over the top, but even without a PPU the avr FPS doesn't change much.
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 10:44:30
(permalink)
Gold Leader
Could you run these tests with Metro 2033 please? That is something I play a lot the past few weeks and since it has good PhysX support as well I just wonder how that would do here
Nice tests for the rest, the GTX 650 is overall the best solution, the 680 is indeed over the top, but even without a PPU the avr FPS doesn't change much.
Right. Don't forget to look at the FluidMark numbers as a potential better indicator of future PhysX performance. Yes, Borderlands 2 resulted in the best performance by the GTX 650. However, I believe the GTX 650 Ti is the best overall for price and performance in both tests. I will not be testing Metro 2033. That game is irrelevant. There is a sequel coming out soon and you can use FluidMark to extrapolate synthetic performance that will give you a good idea of what to expect.
|
HeavyHemi
Insert Custom Title Here
- Total Posts : 15665
- Reward points : 0
- Joined: 2008/11/28 20:31:42
- Location: Western Washington
- Status: offline
- Ribbons : 135
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 10:54:05
(permalink)
Gold Leader
Could you run these tests with Metro 2033 please? That is something I play a lot the past few weeks and since it has good PhysX support as well I just wonder how that would do here
Nice tests for the rest, the GTX 650 is overall the best solution, the 680 is indeed over the top, but even without a PPU the avr FPS doesn't change much.
I've done some pretty extensive testing with various setups on Metro 2033. There is very little change in average FPS ~5% running with or without a dedicated PhysX GPU. However, the change in frame rate variation between min and max is significant. Meaning, Metro 2033 is perceptibly smoother with a PhysX PPU. I believe the highest PhysX GPU usage I noted was ~20%.
EVGA X99 FTWK / i7 6850K @ 4.5ghz / RTX 3080Ti FTW Ultra / 32GB Corsair LPX 3600mhz / Samsung 850Pro 256GB / Be Quiet BN516 Straight Power 12-1000w 80 Plus Platinum / Window 10 Pro
|
shahram_b
New Member
- Total Posts : 23
- Reward points : 0
- Joined: 2008/10/25 21:42:58
- Status: offline
- Ribbons : 0
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 11:02:35
(permalink)
Great work sk3tch! Thanks for sharing this.
do you know what kind of load you had on 680 card during your borderland test? was it always running at 1071? I assume the card was not boosting at all..
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 12:10:41
(permalink)
shahram_b
Great work sk3tch! Thanks for sharing this.
do you know what kind of load you had on 680 card during your borderland test? was it always running at 1071? I assume the card was not boosting at all..
Thank you! The 1071 MHz speed was the 650 Ti. The 680 was running at 1006 Mhz (stock) - I never saw it boost. Admittedly, I did not look up at the Precision X OSD a whole ton - but since the card was not taxed beyond 30%-ish I don't think boost came into play. I believe the 680 gave me the smoothest overall experience when it acted as a PhysX card, despite the numbers showing lower averages and highs...but obviously, it's not worth ~$450 for that smoothness. The 650 Ti gave a good approximation of that for less. :)
|
larsoncc
Superclocked Member
- Total Posts : 238
- Reward points : 0
- Joined: 2012/05/16 21:31:11
- Status: offline
- Ribbons : 1
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 12:47:19
(permalink)
Great tests! Will you be testing Arkham City? Anything to smooth out the frame rate in that game would be welcomed by me. (patches have made that game better, admittedly) This also answers another question that I had earlier - PhysX effects will still be shown in SLI, etc, but would lower frame rate. It seems that a dedicated card simply maximizes the frame rate, its goal isn't to show more effects.
post edited by larsoncc - 2012/10/16 12:50:34
i7 3770, ASUS Sabertooth Z77, 16GB 1600 G.Skill, 3X SSD, 2x HDD, 2x GTX 670 in SLI, 3X 24" ASUS VE248h + 60" 3DTV ... Also - i7 3930K, ASROCK Extreme 6, 16GB 1600 Mushkin, 1 SSD, 1 HDD, 3x 760 in SLI Heatware My Video Game Web Site (online since 2002!): Stage Select - Video Game News and Information
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 13:05:46
(permalink)
larsoncc
Great tests! Will you be testing Arkham City? Anything to smooth out the frame rate in that game would be welcomed by me. (patches have made that game better, admittedly)
This also answers another question that I had earlier - PhysX effects will still be shown in SLI, etc, but would lower frame rate. It seems that a dedicated card simply maximizes the frame rate, its goal isn't to show more effects.
Thanks for your feedback. I will not be testing any other games. I simply did these tests because there was not much info out there and I wanted to make a decision on which card to buy for a dedicated PhysX PPU - primarily for Borderlands 2. I am not interested in any of the older games that use PhysX right now...but I'm looking forward to future games that support it. Take a look a the FluidMark results if you want an idea of the power/capability of any of these cards as a dedicated PhysX PPU.
|
TECH_DaveB
EVGA Alumni
- Total Posts : 4893
- Reward points : 0
- Joined: 2008/09/26 17:03:47
- Status: offline
- Ribbons : 46
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 14:25:15
(permalink)
Excellent write up, shed some good light on things. PhysX implementation in games is still a bit unpredictable some times results don't make a lot of sense, but at least we see where our bang for the buck is for PhysX cards.
Going forward, as you do have the hardware, you think you will run comparisons when other PhysX heavy games come out?
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 14:27:10
(permalink)
EVGATech_DaveB
Excellent write up, shed some good light on things. PhysX implementation in games is still a bit unpredictable some times results don't make a lot of sense, but at least we see where our bang for the buck is for PhysX cards.
Going forward, as you do have the hardware, you think you will run comparisons when other PhysX heavy games come out?
Thank you. Although I would like to be able to do that, I mostly did this research so I could determine which card I should buy for dedicated PhysX purposes. I was going to send the GT 640 and GTX 650 back based on my research. Perhaps I will reconsider...but it's tough to invest $200 into cards just being used for testing purposes. :)
|
PhantomApollyon
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2010/01/27 23:49:20
- Status: offline
- Ribbons : 1
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 21:09:04
(permalink)
I suppose I'll pick up the 650ti SC then but do I let the PLX run both my 680 & and the 650ti or do I split them on the EVGA Z77 board. The reason I ask is because I see better performance(500-800 points in 3dmark) running my 680 in slot 2(PLX) by itself rather than running it in slot 1(direct CPU). So I'd either do slot 2&4 for the 680 and 650 or slot 1&2.
ASUS Maximus z690 Hero // i9 12900K // 4090 FE // 32GB 6000Mhz DDR5 // EVGA 1600 T2
|
EVGA_JacobF
EVGA Alumni
- Total Posts : 16946
- Reward points : 0
- Joined: 2006/01/17 12:10:20
- Location: Brea, CA
- Status: offline
- Ribbons : 26
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 21:13:52
(permalink)
When the GTX 680 was used as a PhysX processor, was it maintaining its base clock or was it throttling?
|
PhantomApollyon
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2010/01/27 23:49:20
- Status: offline
- Ribbons : 1
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 21:17:19
(permalink)
EVGA_JacobF
When the GTX 680 was used as a PhysX processor, was it maintaining its base clock or was it throttling?
Any input on appropriate slot configuration for running my 680+ a 650ti(physx)?
ASUS Maximus z690 Hero // i9 12900K // 4090 FE // 32GB 6000Mhz DDR5 // EVGA 1600 T2
|
surfnet
Superclocked Member
- Total Posts : 151
- Reward points : 0
- Joined: 2007/01/16 10:04:04
- Location: Queen Creek, AZ
- Status: offline
- Ribbons : 0
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 21:27:36
(permalink)
Running a GTX 650 tI and 2 GTX690's. Just ran Mafia 2 and did get about 12 FPS more running the 650TI for Physx as opposed to the 2 GTX690's handling everything. Take these benches with a grain of salt..not much time to delve deeper but looks very promising... And the 650 TI is small enough not to block the cooling for the 690's 
post edited by surfnet - 2012/10/16 21:30:25
|
EVGA_JacobF
EVGA Alumni
- Total Posts : 16946
- Reward points : 0
- Joined: 2006/01/17 12:10:20
- Location: Brea, CA
- Status: offline
- Ribbons : 26
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 21:29:53
(permalink)
PhantomApollyon
I suppose I'll pick up the 650ti SC then but do I let the PLX run both my 680 & and the 650ti or do I split them on the EVGA Z77 board. The reason I ask is because I see better performance(500-800 points in 3dmark) running my 680 in slot 2(PLX) by itself rather than running it in slot 1(direct CPU). So I'd either do slot 2&4 for the 680 and 650 or slot 1&2.
I would suggest slots 1&2 for full performance, or 1 and 4 for better airflow.
|
Sajin
EVGA Forum Moderator
- Total Posts : 49168
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
|
Sajin
EVGA Forum Moderator
- Total Posts : 49168
- Reward points : 0
- Joined: 2010/06/07 21:11:51
- Location: Texas, USA.
- Status: offline
- Ribbons : 199
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 21:39:13
(permalink)
EVGA_JacobF
PhantomApollyon
I suppose I'll pick up the 650ti SC then but do I let the PLX run both my 680 & and the 650ti or do I split them on the EVGA Z77 board. The reason I ask is because I see better performance(500-800 points in 3dmark) running my 680 in slot 2(PLX) by itself rather than running it in slot 1(direct CPU). So I'd either do slot 2&4 for the 680 and 650 or slot 1&2.
I would suggest slots 1&2 for full performance, or 1 and 4 for better airflow.
@PhantomApollyon, How is that possible since direct link to CPU should be better? @Jacob, Don't you mean slot 2&4 for better performance? Running in slots 2&4 allow 16x Gen3 on both cards, while running slot 1&2 allow 8x Gen3 on both.
|
EVGA_JacobF
EVGA Alumni
- Total Posts : 16946
- Reward points : 0
- Joined: 2006/01/17 12:10:20
- Location: Brea, CA
- Status: offline
- Ribbons : 26
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 22:01:35
(permalink)
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/16 22:34:25
(permalink)
EVGA_JacobF
When the GTX 680 was used as a PhysX processor, was it maintaining its base clock or was it throttling?
Base clock - no boost.
|
Gold Leader
CLASSIFIED Member
- Total Posts : 3939
- Reward points : 0
- Joined: 2009/05/30 03:06:17
- Location: Dirksland, The Netherlands
- Status: offline
- Ribbons : 62
|
jaafaman
CLASSIFIED Member
- Total Posts : 2188
- Reward points : 0
- Joined: 2008/02/23 15:04:11
- Status: offline
- Ribbons : 13
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 01:08:00
(permalink)
sk3tch larsoncc Great tests! Will you be testing Arkham City? Anything to smooth out the frame rate in that game would be welcomed by me. (patches have made that game better, admittedly)
This also answers another question that I had earlier - PhysX effects will still be shown in SLI, etc, but would lower frame rate. It seems that a dedicated card simply maximizes the frame rate, its goal isn't to show more effects.
Thanks for your feedback.
I will not be testing any other games. I simply did these tests because there was not much info out there and I wanted to make a decision on which card to buy for a dedicated PhysX PPU - primarily for Borderlands 2. I am not interested in any of the older games that use PhysX right now...but I'm looking forward to future games that support it. Take a look a the FluidMark results if you want an idea of the power/capability of any of these cards as a dedicated PhysX PPU.
Games like Batman:AA, Batman: AC, Metro2033 and even Mafia2 aren't quite that "irrelevant" in a PhysX test, however, as they represent various implementations of the APEX SDK. Unlike games with more "generalized" PhysX support, these games tend to break things like SLI at times. Going into the future, we can only hope for improved support ala Borderlands2, but as long as people like Warner Brothers are releasing games then APEX implementations may hold a higher significance than you assume when considering dedicated PPU support... EDIT - Just so you know, I've been doing the same thing for the last three generations of nVidia engines so I'm not coming into this from far afield. Here you can see some of the mistakes I made in the first round by following the conversation...
post edited by jaafaman - 2012/10/17 01:16:34
|
PhantomApollyon
Superclocked Member
- Total Posts : 212
- Reward points : 0
- Joined: 2010/01/27 23:49:20
- Status: offline
- Ribbons : 1
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 04:52:55
(permalink)
EVGA_JacobF
PhantomApollyon
I suppose I'll pick up the 650ti SC then but do I let the PLX run both my 680 & and the 650ti or do I split them on the EVGA Z77 board. The reason I ask is because I see better performance(500-800 points in 3dmark) running my 680 in slot 2(PLX) by itself rather than running it in slot 1(direct CPU). So I'd either do slot 2&4 for the 680 and 650 or slot 1&2.
I would suggest slots 1&2 for full performance, or 1 and 4 for better airflow.
Sajin
EVGA_JacobF
PhantomApollyon
I suppose I'll pick up the 650ti SC then but do I let the PLX run both my 680 & and the 650ti or do I split them on the EVGA Z77 board. The reason I ask is because I see better performance(500-800 points in 3dmark) running my 680 in slot 2(PLX) by itself rather than running it in slot 1(direct CPU). So I'd either do slot 2&4 for the 680 and 650 or slot 1&2.
I would suggest slots 1&2 for full performance, or 1 and 4 for better airflow.
@PhantomApollyon, How is that possible since direct link to CPU should be better?
@Jacob, Don't you mean slot 2&4 for better performance? Running in slots 2&4 allow 16x Gen3 on both cards, while running slot 1&2 allow 8x Gen3 on both.
Good deal, I'll get my 650ti on order and pop it in slot 4 to get things rolling with some improved physx goodness, thanks for the input guys!
ASUS Maximus z690 Hero // i9 12900K // 4090 FE // 32GB 6000Mhz DDR5 // EVGA 1600 T2
|
dyceskynes
Superclocked Member
- Total Posts : 153
- Reward points : 0
- Joined: 2009/05/22 16:45:44
- Status: offline
- Ribbons : 0
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 07:06:27
(permalink)
I assume since there was a performance increase on SLi 690s that adding a 650ti to SLI 680 Classifieds will still show a significant performance increase?
Intel Core I7 3930K @ 4.5 | 16GB Corsair Vengence 1600 MHZ | EVGA X79 Dark | EVGA 1080 FTW | 256GB 850 Evo SSD x1 | 500 GB WD x2 | EVGA 1000 G2 | XSPC Raystorm, 360 + 180 Rad, MCP655 | Win 10 Pro
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 07:57:03
(permalink)
jaafaman
Games like Batman:AA, Batman: AC, Metro2033 and even Mafia2 aren't quite that "irrelevant" in a PhysX test, however, as they represent various implementations of the APEX SDK. Unlike games with more "generalized" PhysX support, these games tend to break things like SLI at times.
Going into the future, we can only hope for improved support ala Borderlands2, but as long as people like Warner Brothers are releasing games then APEX implementations may hold a higher significance than you assume when considering dedicated PPU support...
EDIT -
Just so you know, I've been doing the same thing for the last three generations of nVidia engines so I'm not coming into this from far afield. Here you can see some of the mistakes I made in the first round by following the conversation...
For my purposes - they're irrelevant. There is data out there on these games that you can go find and get a good idea regarding performance. For Borderlands 2 and Kepler cards as dedicated PhysX in general - there was not data (or as much as I would have liked to see to choose a card). So I went out and did some tests and put these results up. Those games are also old. I don't care to play them. But, I do understand others would...I'm just a dude with a blog and not a professional so I'm merely doing this for my own fact finding and as something to share info with people. The cards are going back today, with the exception of the 650 Ti. :) dyceskynes
I assume since there was a performance increase on SLi 690s that adding a 650ti to SLI 680 Classifieds will still show a significant performance increase?
Not sure if you're aware, but a GTX 690 is basically SLI 680s. Mine was overclocked to beyond stock Classified speeds (I own 4 of them, I should know, ha). Borderlands 2 and FluidMark both do not use more than ~1GB of VRAM. So, the tests above will provide a good approximation of what you could expect if you were to add a dedicated PhysX GPU/PPU from the list to your system. SLI 690s is quad-SLI - 4 GPUs. :)
post edited by sk3tch - 2012/10/17 08:04:58
|
dyceskynes
Superclocked Member
- Total Posts : 153
- Reward points : 0
- Joined: 2009/05/22 16:45:44
- Status: offline
- Ribbons : 0
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 08:48:32
(permalink)
Indeed I did know, but I was not sure if the extra slot not being pci 3 gen would be an issue. Thanks for all your research!
Oh and a 1G 650 Ti SCC is what you used or was it the 2G?
Intel Core I7 3930K @ 4.5 | 16GB Corsair Vengence 1600 MHZ | EVGA X79 Dark | EVGA 1080 FTW | 256GB 850 Evo SSD x1 | 500 GB WD x2 | EVGA 1000 G2 | XSPC Raystorm, 360 + 180 Rad, MCP655 | Win 10 Pro
|
sk3tch
iCX Member
- Total Posts : 278
- Reward points : 0
- Joined: 2007/07/11 12:33:45
- Status: offline
- Ribbons : 3
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 08:54:59
(permalink)
dyceskynes
Indeed I did know, but I was not sure if the extra slot not being pci 3 gen would be an issue. Thanks for all your research!
Oh and a 1G 650 Ti SCC is what you used or was it the 2G?
Ah good. Sorry. I think the only consideration would be your motherboard, case, and platform (X79, Z77, P67, etc.) - if you can fit the card, do it. If it will be cool enough, etc. Yes, only the 1GB card. So far, in my testing I have found memory to be pretty irrelevant for PhysX. Only around 200MB is ever used on the dedicated PPU (max).
|
dyceskynes
Superclocked Member
- Total Posts : 153
- Reward points : 0
- Joined: 2009/05/22 16:45:44
- Status: offline
- Ribbons : 0
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 09:39:42
(permalink)
Thanks again. In my mind this card was free with the -$50 off for the Assassins Creed game and -$100 for the Classy. Somehow I do not think my wife will view it that way.
Intel Core I7 3930K @ 4.5 | 16GB Corsair Vengence 1600 MHZ | EVGA X79 Dark | EVGA 1080 FTW | 256GB 850 Evo SSD x1 | 500 GB WD x2 | EVGA 1000 G2 | XSPC Raystorm, 360 + 180 Rad, MCP655 | Win 10 Pro
|
lehpron
Regular Guy
- Total Posts : 16254
- Reward points : 0
- Joined: 2006/05/18 15:22:06
- Status: offline
- Ribbons : 191
Re:PhysX PPU / GPU Shoot Out: GT 640, GTX 650, GTX 650 Ti, and GTX 680
2012/10/17 11:39:17
(permalink)
sk3tch
Throwing a GTX 680 at it as a PPU does not yield crazy results – but a GTX 650 does surprisingly well. My hypothesis? Clocks matter. I'd inquire how each game uses the graphics card for both PhysX and general game rendering. It is easy to assume that performance scales with CUDA cores since the CUDA cores themselves are doing the actual rendering and computation, but their frequencies matter; that's why I multiply the numbers and scale performance that way. But, it is just an "up to" because games don't use all of the capability depending on how much was GPU-intensive, or resolution or settings. What if because the game uses only some of the CPU cores and not all to run traditional physics, then not all CUDA cores get used? Hypothetically, as the number of CUDA cores goes up, the extra shouldn't make a difference, like having a game on a 6-core or 8-core x86 CPU in general performance; thus having a GTX680 as PPU should have the same performance as the nearest slower card that added the last chunk of gain. That said, I could explain the drop in performance due to the GTX680 as possibly too much physics, bottlenecking the main card. A perfect way to test that theory is to make the 680 PhysX and make the slower cards do the rendering. Anything is possible, but we need more data, i.e. perform more tests with more programs, to find a pattern. It is a bit dangerous to generalize based on 1 or 2.
|