Getting back into WoW / Boxing, need hardware advice.

[spher0boom]

Hey Guys.

I took a break from WoW / multiboxing, but I’m hoping to get back into it again soon. I’m trying to plan out my hardware, and I’m hoping to get some advice on what my best options would be given the hardware I already have.

I originally boxed on 6 PCs each with a single dedicated monitor (3 Across by 2 high) using Multiplicity. I still have the monitors in that configuration, but have replaced the original 6 PCs with 4 new ones. I loved having all 5 clients full-screen with medium-to-high settings, but using 5 full-screens isn’t a requirement, I’d just really like to be able to use all of the slaves without having to minimize/maximize windows to do it.

Here is what I currently have for hardware:

Primary PC (Bottom 3 Monitors: 1680x1050, 1920x1200, 1680x1050)
- i7-2600K
- 32GB GSKIL F3-12800CL10Q-32GBZL
- Radeon HD 6870
- Intel 120GB SSD (Windows)
- Intel 80GB SSD (Mac OS X)
- 750 Watt (Corsair TX750 with 62A on a dedicated single +12V rail)

Secondary PC (Top Center Monitor: 1680x1050)
- i5-2500K
- 32GB GSKIL F3-12800CL10Q-32GBZL
- Radeon HD 4890
- Intel 250GB SSD
- 750 Watt (Corsair HX650 with 54A on a dedicated single +12V rail)

Tertiary PC 1 (Top Left Monitor: 1280x1024)
- i5-2500K
- 8GB Corsair PC3 12800
- Intel HD 3000
- 520 Watt (Corsair HX520HW with 3 x 18A +12V rails)

Tertiary PC 2 (Top Left Monitor: 1280x1024)
- i5-3750K
- 8GB Corsair PC3 12800
- Intel HD 4000
- 550 Watt (Corsair VX550 with 41A on a dedicated single +12V rail)

All 4 PCs are built on the Gigabyte Z68x-UD3H-B3 motherboard.

I also have the following video cards available:
- Radeon HD 4890
- 8800 GTS 640MB
- 8800 GTS 640MB
- 9600GT

I’m also more than willing to buy a new GTX 670 or 680 if it will accomplish what I’m looking for, as well as a wow-dedicated SSD.


Issues / Limitations:


(1) I’d prefer to 5-box with as few computers as possible (preferably 1) as the two tertiary PCs are currently dedicated to other tasks (video encoding, VMware, etc..).

(2) The physical setup of the PCs and the monitors limits my configuration. The secondary and tertiary PCs are in a different room connected by 50’ DVI cables. The distance requires I use DVI cables (VGA looks horrible) and the monitors have only a single DVI slot, so I’d have to add KVM/Monitor switches to use the DVI ports. The VGA ports are available and close to the main PC.

(3) I can go up to 4 monitors on my primary PC, but I don’t believe adding a second GPU to my primary PC is an option as the Gigabyte Z68x-UD3H-B3 motherboard has 1 16x and 1 8x 2.0 compliant slot, and if a second card is added the 16x slot runs at 8x, which I assume would hurt performance a lot more than it would help it?

(4) The Keyboard/Mouse control software I used in my old setup (Multiplicity) doesn't work well in multi-monitor setups, and is very awkward to use to navigate to the top left/right monitors, so it’s not ideal for fast reactions. I haven’t found a replacement that works well for gaming (mouse movement goes crazy when using Synergy and Share-Mouse), so a 1-PC or 2-PC solution using the bottom 3 and top center monitors (primary and secondary PCs only) would be best.


Here are my questions:


Everyone says that boxing 5 on a single PC is very doable with even a medium quality graphics card, especially if you use DX9. However I’m not sure if anyone is doing 5 large windows on a single PC with a single GPU, and I haven’t purchased the 670 yet to test. I saw a post from MiRai that said a 3/4GB card couldn’t do it in DX11. I saw a post from Ualaa saying he’s using running 4 slaves on a 1080x1920 secondary monitor (in portrait maybe?) in DX9, but I’m not sure what size they are.

I’m wondering if anyone is running a single 4GB card with a 1920x1080 High/Ultra window and 4 slaves at 1024x768 or larger with medium settings across multiple monitors?


MiRai said he had some issues using 3 monitors as his performance was choppy, but that most people didn’t have this issue. However, I attempted to get back into boxing 4-5 months ago tried it, and I did have similar issues when opening the 3rd client on the 3rd monitor with the 6870. However, that’s a 1GB card (I did run them in DX9).

Is anyone running 5 full-screen sessions? If so what is your setup? I might consider upgrading my motherboard if I could do that.


MiRai posted that running different resolutions on your monitors caused significant performance issues for the 4xx and 5xx cards. I didn’t see any posts that specifically listed different resolutions, but saw many listing the same, though I certainly could have missed them.

Does this issue still exist for the 6xx cards?


If you had the equipment listed above available to you, and had the option of buying an additional 670/680 and Game-dedicated SSD, how would you configure this boxing setup if it were yours?


Thanks in advance for any advice!
- spher0boom

[RSM72]

You should try running it all off your primary PC. I'm currently 5-boxing on an overclocked i7 950 (@3.6Ghz) with 12GB of system memory. I got two 24" screens hooked to that box and have one client fullscreen and the other 4 on the second screen. 50% downsampled 1920x1200 is close enough to the resolution of your tertiary screens (1280x1024) I think you wont lose much here. I'm running all clients in DX11 with one client on high settings and the others on low-medium, the slaves capped at 20fps.

For the GPU you most likely want a non-SLI setup because SLI will give you only very low increase in performance but alot of problems and issues. My GTX 580 is already overkill for that setup since I'm running into CPU-limitations when LFR-ing rather that GPU-limits. I'd still replace that HD 6870 because WoW performs better on Nvidia cards and you want a card with 3-4GB of video memory for 5 clients.

Performance-wise its quiet good on a single box, fps on the main client is 25-40 in capitals. In 40-man BGs it dropped down to 20 in crowded areas but Blizzard solved that problem by removing /follow from BGs. In 25-man raids (LFR) it gets worse by dropping to 10-15 which is close to being unplayable (yet I still do good dps I just dont move). Problem here is the CPU (I see all cores close to 100%) not the GPU.

As for software - ISboxer supports multi-computer configurations I just cant comment anything about that because I never tried that (running of one box only). I think its recommended somewhere to run all clients at the same resolution when using ISboxer.

[MiRai]

(1) I’d prefer to 5-box with as few computers as possible (preferably 1)
This is what most people these days.


(3) I can go up to 4 monitors on my primary PC, but I don’t believe adding a second GPU to my primary PC is an option as the Gigabyte Z68x-UD3H-B3 motherboard has 1 16x and 1 8x 2.0 compliant slot, and if a second card is added the 16x slot runs at 8x, which I assume would hurt performance a lot more than it would help it?
It's highly doubtful you'll be able to saturate the PCIe lanes on 8x.

SLI 16x/16x VS 16x/8x - http://hardocp.com/article/2010/08/16/sli_cfx_pcie_bandwidth_perf_x16x16_vs_x16x8/2#.UZ4ft7XVB8E


(4) The Keyboard/Mouse control software I used in my old setup (Multiplicity) doesn't work well in multi-monitor setups, and is very awkward to use to navigate to the top left/right monitors, so it’s not ideal for fast reactions. I haven’t found a replacement that works well for gaming (mouse movement goes crazy when using Synergy and Share-Mouse), so a 1-PC or 2-PC solution using the bottom 3 and top center monitors (primary and secondary PCs only) would be best.
I know nothing of Multiplicity. Most people use Input Director because it's free and seems to do everything that Multiplicity does (except centralize audio).



Everyone says that boxing 5 on a single PC is very doable with even a medium quality graphics card, especially if you use DX9. However I’m not sure if anyone is doing 5 large windows on a single PC with a single GPU, and I haven’t purchased the 670 yet to test.
I would say that most people render each of their game clients at the resolution of their main client so that mouse broadcasting is 1:1. This really comes down to which software you'd be using because if you're going to be running something like Keyclone, then no matter what you do your game clients will render at the resolution they're shown on your monitor, whereas, if you use ISBoxer, then even though the game clients look scaled down, they're still rendering at the resolution you tell them to (by default -- this can be changed) so that mouse broadcasting accuracy is preserved.

This is on a person-to-person basis, though. It's hard to judge how a person truly has their clients set up via a screenshot unless they give a lot of details about it.


I saw a post from MiRai that said a 3/4GB card couldn’t do it in DX11.
...using the Ultra preset @ 1920x1080 x5. There are those that claim they can "run" their clients at Ultra @ 1920x1080 on 5x game clients using DX11, but everyone's definitions of "run" and "playable" are vastly different.


I saw a post from Ualaa saying he’s using running 4 slaves on a 1080x1920 secondary monitor (in portrait maybe?) in DX9, but I’m not sure what size they are.
Pretty sure he was rolling around with DX11 before he quit, but I could be wrong.


MiRai said he had some issues using 3 monitors as his performance was choppy, but that most people didn’t have this issue.
I can't imagine I said this recently? This sounds like something I said back when I was using a Q9550 and 2x GTX 260s... which was 2+ years ago.


MiRai posted that running different resolutions on your monitors caused significant performance issues for the 4xx and 5xx cards. I didn’t see any posts that specifically listed different resolutions, but saw many listing the same, though I certainly could have missed them. Does this issue still exist for the 6xx cards?
Significant performance issues? Are you talking about the issue where you run two monitors of different resolutions (monitor resolutions, not game resolutions) off of the same GPU, that GPU doesn't downclock?

If that's what you're asking, then that doesn't cause significant performance issues, but as far as I know, yes, the downclocking problem still exists; but if it really bothers you then you can manually change the P-State of each card through nVidia Inspector.


If you had the equipment listed above available to you, and had the option of buying an additional 670/680 and Game-dedicated SSD, how would you configure this boxing setup if it were yours?
Use the 2600K system and get a better GPU.



For the GPU you most likely want a non-SLI setup because SLI will give you only very low increase in performance but alot of problems and issues. My GTX 580 is already overkill for that setup since I'm running into CPU-limitations when LFR-ing rather that GPU-limits. I'd still replace that HD 6870 because WoW performs better on Nvidia cards and you want a card with 3-4GB of video memory for 5 clients.
This is something I've been meaning to address for awhile now because I've been running SLI for the last few months and the performance increase I'm seeing is amazing. I'm able to push video settings beyond what I normally would be able to using just a single card (or I would be forced to split the load between two GPUs on separate monitors).

Here's some quick proof I whipped up this morning: SLI VS NoSLI (http://imgur.com/a/WwJJD)
(hover over the image, click the gear icon, and choose to view in full resolution if you want to see the full image)

I've become quite attached to that layout and running all five clients on a single monitor, so I don't have any 1920x1080 comparisons for anyone.

For anyone that looks at those comparison images, know that SLI has never been a proven technology while running multiple game clients and its results may vary greatly from setup to setup or from game to game. The images I've posted above aren't saying, "SLI is a must! Go out and spend more money now!" because if you happen to run out to the store and purchase a second (or third or fourth) video card to SLI together and it doesn't work as awesome as you thought it would, I'm not responsible. :)

[remanz]

all I have to say is stay a way from SLI or dual chip cards. Not worth the hassle. Not worth the heat. get GTX 780 now~

[spher0boom]

I know nothing of Multiplicity. Most people use Input Director because it's free and seems to do everything that Multiplicity does (except centralize audio).


I saw this in one of Ualaa's posts last night, and went looking for it after that. A couple of forum posts identified WoW users who reported that when clicking on secondary screens InputDirector was also still clicking on the primary screen, so I didn't end up trying it. If that isn't the case for most users I'll go give it a try if I end up using multiple PCs.






MiRai said he had some issues using 3 monitors as his performance was choppy, but that most people didn’t have this issue.

I can't imagine I said this recently? This sounds like something I said back when I was using a Q9550 and 2x GTX 260s... which was 2+ years ago.


I believe the post was at least 18 months old. I wasn't really sure how much things changed over time, so I included it just in case it was still an issue. This is probably a result of me searching rather than systematically going back through the posts.




Significant performance issues? Are you talking about the issue where you run two monitors of different resolutions (monitor resolutions, not game resolutions) off of the same GPU, that GPU doesn't downclock?

If that's what you're asking, then that doesn't cause significant performance issues, but as far as I know, yes, the downclocking problem still exists; but if it really bothers you then you can manually change the P-State of each card through nVidia Inspector.

Yes, I'm talking about 2 different monitor resolutions off the same GPU. I'm not sure what the problem was, it wasn't identified in the post, I believe you were still looking in to it. I don't know what the P-State is ... I'll do some research, thanks!




Use the 2600K system and get a better GPU.

Done. GTX 670 will be here tomorrow.





It's highly doubtful you'll be able to saturate the PCIe lanes on 8x.

SLI 16x/16x VS 16x/8x - http://hardocp.com/article/2010/08/16/sli_cfx_pcie_bandwidth_perf_x16x16_vs_x16x8/2#.UZ4ft7XVB8E



This one is a 16x/16x to 16x/8x comparison, whereas I would end up 8x/8x. However, the article you found referenced a follow-up in the conclusion with a 16x/16x to 8x/8x comparison here:

http://www.hardocp.com/article/2010/08/23/gtx_480_sli_pcie_bandwidth_perf_x16x16_vs_x8x8/#.UZ7YV7WyB8E


In our evaluation last week, we tested at x16/x16 and x16/x8 and found that there were miniscule differences at 2560x1600 and 5760x1200 resolutions. However, in this week’s evaluation testing at x8/x8 and x16/x16, we see that having both video cards at x8 does somewhat impact performance, but only at the higher 5760x1200 Eyefinity/NV Surround resolution.

It seems that at 2560x1600, even with 4X AA, there was absolutely no difference between x16/x16 and x8/x8. This is good news if you game at x8/x8 on a single display configuration at 2560x1600 and below. You simply are not missing anything, and moving up to x16/x8 or x16/x16 will yield no performance improvements or gameplay differences, even on the fastest GTX 480 SLI.

At x8/x8 we didn’t find that it hurt our gameplay experience or caused us to change any in-game settings. All it did was cause around a 7% (at most) difference in the average framerate on one game, BF2.

Therefore, as we look into the future of gaming, if a game is more shader heavy, it will probably have less affect with PCIe bandwidth, but if a game uses insanely high texture sizes and AA settings, PCIe bandwidth may be a concern as we move into multiple display gaming configurations. And surely we are going to be seeing games with larger textures with the widespread adoption of multi-display gaming.


So that's good news for a lot of people. However, I wasn't really looking to do SLI, I wanted to add more full-screen monitors, not necessarily increase the performance of my primary screen. I had actually looked for some articles a while back, but I got discouraged quickly and stopped looking. I'm not sure where I was looking last time, because the answers I'm getting this time about single card 16x vs 8x performance say there is only about a 3% difference, so I'm encouraged to throw in a card and do some experimenting. I'm really hoping this works and doesn't impact performance because I would love to be able to use all 6 monitors. I used to be able to do this on the Z68x-UD3H-B3 using the on-board video, but once I moved up to 32GB it performed horribly and I had to stop.




This is something I've been meaning to address for awhile now because I've been running SLI for the last few months and the performance increase I'm seeing is amazing. I'm able to push video settings beyond what I normally would be able to using just a single card (or I would be forced to split the load between two GPUs on separate monitors).


I only have a 24" primary monitor, so my intent was to split the load. I'd rather sacrifice slave size/quality and have a large primary window with very high or ultra settings. Given my 8x/8x restriction an eyefinity display or something similar doesn't seem like it would perform well given the articles referenced above. My Preference would be to have two 23" 1080p monitors in portrait mode on either side of my primary and do 4 x 1080x810 windows for he slaves, but i'd have to buy 3 new monitors and end up limiting myself to 4 instead of 6 given my space, which doesn't work well with the other tasks the computers perform.

I have the extra 8800GTS cards. Would one of them come close to supporting two 1280x1024 WoW sessions on medium settings? I'd be happy with 4 1280x1024 windowed slaves. If not, what would be the minimum card I should look at as a secondary if I attempt go this route?




I would say that most people render each of their game clients at the resolution of their main client so that mouse broadcasting is 1:1. This really comes down to which software you'd be using because if you're going to be running something like Keyclone, then no matter what you do your game clients will render at the resolution they're shown on your monitor, whereas, if you use ISBoxer, then even though the game clients look scaled down, they're still rendering at the resolution you tell them to (by default -- this can be changed) so that mouse broadcasting accuracy is preserved.

This is on a person-to-person basis, though. It's hard to judge how a person truly has their clients set up via a screenshot unless they give a lot of details about it.

It never really occurred to me to use the single monitor for all clients. I recently upgraded to the 24" and am still getting used to it. I'll have to give it a try and see how it works.

The mouse broadcasting thing is actually something I completely forgot about. I used to use Octopus and HotKeyNet, which did relative movements, so my screens never had to be the same resolution. This time around I was going to try ISBoxer since everyone here seems to love it. I have a lot of learning to do around ISBoxer, but I see several forum member have put together nice guides, so I have a good place to start ;). Does ISBoxer offer any 'relative' mouse broadcasting capabilities if I chose to not use the same resolution on all clients?

Thank-you very much for your advice/input!

-spher0boom

[spher0boom]

all I have to say is stay a way from SLI or dual chip cards. Not worth the hassle. Not worth the heat. get GTX 780 now~
I read MiRai's post too late ... I had already ordered a new GTX-670 since all the posts on the forum indicated the same things, so we'll see how it goes. Lots of places (including Tom's and Ualaa's posts) indicated it was a great 'bang for the buck' card, and could easily be clocked up to the 680 speeds.

Thanks!

[spher0boom]

You should try running it all off your primary PC. I'm currently 5-boxing on an overclocked i7 950 (@3.6Ghz) with 12GB of system memory. I got two 24" screens hooked to that box and have one client fullscreen and the other 4 on the second screen. 50% downsampled 1920x1200 is close enough to the resolution of your tertiary screens (1280x1024) I think you wont lose much here. I'm running all clients in DX11 with one client on high settings and the others on low-medium, the slaves capped at 20fps.

This is very encouraging, thanks!



For the GPU you most likely want a non-SLI setup because SLI will give you only very low increase in performance but alot of problems and issues. My GTX 580 is already overkill for that setup since I'm running into CPU-limitations when LFR-ing rather that GPU-limits. I'd still replace that HD 6870 because WoW performs better on Nvidia cards and you want a card with 3-4GB of video memory for 5 clients.

I ordered a 670, so this is also very encouraging!



Performance-wise its quiet good on a single box, fps on the main client is 25-40 in capitals. In 40-man BGs it dropped down to 20 in crowded areas but Blizzard solved that problem by removing /follow from BGs. In 25-man raids (LFR) it gets worse by dropping to 10-15 which is close to being unplayable (yet I still do good dps I just dont move). Problem here is the CPU (I see all cores close to 100%) not the GPU.

Well, this part worries me a bit. I was hoping to keep 30+fps on the main worst case. Perhaps if I add a second GPU for two fo teh screens I can alleviate this issue.



As for software - ISboxer supports multi-computer configurations I just cant comment anything about that because I never tried that (running of one box only). I think its recommended somewhere to run all clients at the same resolution when using ISboxer.
I've never used it before, but it's on my list this time to try.

Thanks for the encouraging feedback!

[RSM72]

Well, this part worries me a bit. I was hoping to keep 30+fps on the main worst case. Perhaps if I add a second GPU for two fo teh screens I can alleviate this issue.
WoW is very cpu-intensive that is the main problem performance-wise. Without a six- or eight-core cpu you wont come close to any limit of your 670.

[MiRai]

I saw this in one of Ualaa's posts last night, and went looking for it after that. A couple of forum posts identified WoW users who reported that when clicking on secondary screens InputDirector was also still clicking on the primary screen, so I didn't end up trying it. If that isn't the case for most users I'll go give it a try if I end up using multiple PCs.
Not entirely sure on that. I know a lot of people use Input Director with ISBoxer (Synergy has shown to cause issues) when using two different computers to play. I only multibox on one machine, but I've had ID installed for years so I can use both of my machines at once. I can't say that I've heard of many ID issues as of lately.


Yes, I'm talking about 2 different monitor resolutions off the same GPU. I'm not sure what the problem was, it wasn't identified in the post, I believe you were still looking in to it. I don't know what the P-State is ... I'll do some research, thanks!
P-States are power states. Multiple monitors of different resolutions (or refresh rates) or having your cards in SLI will cause the main GPU to not downclock to its idle state. You can force it to downclock via third-party tools like nVidia Inspector, but if you don't remember to turn it back up when you begin to play you're going to realize very quickly that something is wrong. Years ago I was manually forcing power states on my cards so that they'd be idle when I wasn't playing games... that lasted about a week before I got tired of continuously forgetting to turn it back to its normal state and having my games load up at 3 FPS.

924



My Preference would be to have two 23" 1080p monitors in portrait mode on either side of my primary and do 4 x 1080x810 windows for he slaves, but i'd have to buy 3 new monitors and end up limiting myself to 4 instead of 6 given my space, which doesn't work well with the other tasks the computers perform.
1080x810 seems like an odd aspect ratio (1.33), you would get better visual quality on those screens if you tried to keep the AR as close to your main client's resolution as you can (1920/1080 = 1.78). Not to mention, if you ever felt like disabling ISBoxer's instant swapping and wanted the mouse cursors in each window to sorta line up, then the AR would need to be close to identical on each screen.


I have the extra 8800GTS cards. Would one of them come close to supporting two 1280x1024 WoW sessions on medium settings? I'd be happy with 4 1280x1024 windowed slaves. If not, what would be the minimum card I should look at as a secondary if I attempt go this route?
I think you need to just get your new GPU and see how it handles five clients @ 1920x1080 and then go from there.


The mouse broadcasting thing is actually something I completely forgot about. I used to use Octopus and HotKeyNet, which did relative movements, so my screens never had to be the same resolution. This time around I was going to try ISBoxer since everyone here seems to love it. I have a lot of learning to do around ISBoxer, but I see several forum member have put together nice guides, so I have a good place to start ;). Does ISBoxer offer any 'relative' mouse broadcasting capabilities if I chose to not use the same resolution on all clients?
Yes, you can disable instant swapping and scale your mouse cursor to fit to the windows.

[spher0boom]

1080x810 seems like an odd aspect ratio (1.33), you would get better visual quality on those screens if you tried to keep the AR as close to your main client's resolution as you can (1920/1080 = 1.78). Not to mention, if you ever felt like disabling ISBoxer's instant swapping and wanted the mouse cursors in each window to sorta line up, then the AR would need to be close to identical on each screen.
I didn't think about that ... I'll scale them to the same aspect ratio. I'm using 16:10 (1920x1200) so I'll get a little extra vertical space out of it any way.



I think you need to just get your new GPU and see how it handles five clients @ 1920x1080 and then go from there.
Well, Amazon delivered me a used card today instead of the new one I purchased. I'm a little hesitant to install it and play with it while waiting for the new one to arrive since I don't know if it was returned because it was defective. I'm not very happy about this since I had planned on trying it tonight :(

My 6870 couldn't do a single 1920x1200 session with all ultra settings without it being 'jerky' when I ran. In the human starting area I was only getting 45fps.

This site (http://www.videocardbenchmark.net/high_end_gpus.html) lists the passmark ratings, and the 670 is rated at 5,349 compared to the 6870's 2,555 rating, so I'm hoping for a vast improvement. The 680 was only rated at 5,682, which is why I didn't grab it ... seems like a small improvement for $100.


Yes, you can disable instant swapping and scale your mouse cursor to fit to the windows.
I started going through the ISBoxer tutorials last night after I posed. You did a very nice job on them, and make the settings easy to understand, so thank you. My 7-day trial was activated today so I can play with it tonight no matter which card I have installed.

Thanks!

[MiRai]

This site (http://www.videocardbenchmark.net/high_end_gpus.html) lists the passmark ratings, and the 670 is rated at 5,349 compared to the 6870's 2,555 rating, so I'm hoping for a vast improvement. The 680 was only rated at 5,682, which is why I didn't grab it ... seems like a small improvement for $100.
If you can tell me what those numbers mean then I won't entirely discredit that site. How do those numbers pertain to FPS in World of Warcraft while multiboxing? Actually, how do those numbers pertain to FPS in any game?

The answer is... They don't. I'm not being a dick or calling you out, but too many people rely on that site to make their next big purchase and it really doesn't tell you much of anything except that the GTX 670 gets 5,349 points. Points for what?

No review site that I know of uses Passmark in any of their benchmark compilations -- I would suggest sticking to looking at gaming benchmarks at high resolutions (1920x1080 and up) in order to make future purchases.


I started going through the ISBoxer tutorials last night after I posed. You did a very nice job on them, and make the settings easy to understand, so thank you.
Thanks. My video tutorials are mostly out-of-date at the moment because ISBoxer 41 just came out and a lot of things don't match up, but I'm working on it.

[spher0boom]

If you can tell me what those numbers mean then I won't entirely discredit that site. How do those numbers pertain to FPS in World of Warcraft while multiboxing? Actually, how do those numbers pertain to FPS in any game?

The answer is... They don't. I'm not being a dick or calling you out, but too many people rely on that site to make their next big purchase and it really doesn't tell you much of anything except that the GTX 670 gets 5,349 points. Points for what?

Feel free to call me out, and I wouldn't consider you a dick in doing so. You are trying to help me meet my gaming needs. If i did something dumb you should tell me. A dick wouldn't have replied 4 times with pages of text :)

I agree that those are not real world numbers, and I would never make a purchase on the passmark score. I only looked there to try and get a rough quantifiable idea of how much better the 680 was compared to the 670, given the 680 was 25% more as far as cost. The primary decision point in selecting that card were the reviews I read and the forum posts that said the 670 was close to a 680, and that you could OC a 670 to reach 680 performance easily if you bought the right card. However, most of the data I could find as far as 670 vs 680 was forum posts and not actual game reviews from reputable sources, so I was looking for some confirmation. I found the performance of my 6870 very disappointing, and just wanted some quantifiable reference to the new cards as the reviews that I found for 6870 were done at different resolutions, so I had a harder time comparing performance.

My intent is to do exactly what you indicated. Get the card installed and try it for my self. I'll start with full 1080p and work my way down if necessary. I've actually decided to try the card amazon sent, which could work out well. Tomorrow I'll have 2 cards for a few day, so I can try SLI and a variety of configurations across my two primary workspaces. I also have a standing workstation with a 24" Landscape and 23" portrait monitor, which I hadn't thought about using for gaming. However, it may work out better for boxing in the end. It's attached to the secondary workstation with the i5, so I grabbed an ivy bridge i7 on the way home tonight based on the results I saw in your hyper-threading numbers and your reccommendation to use the 2600K system.


Thanks. My video tutorials are mostly out-of-date at the moment because ISBoxer 41 just came out and a lot of things don't match up, but I'm working on it.
Ok, I'll start complaining to you tomorrow then :)

I do appreciate the time and effort you put into posts on this site, it has helped me immensely, so thank you very much!

-spher0boom.

[Ualaa]

I've never used Input Director or Multiplicity.

In general, when boxing a lot of copies of a game, the limiting factor is the amount of video ram.
So having a lot of video ram gets you the option for more eye candy enabled, and/or a greater number of clients compared to a card with less ram.

The benchmarks I've seen for the 670 placed it around 80-90% of the performance of the 680 (links were from Tom's Hardware).
When I got my 670, it was basically 66% of the price of a 680.
The Tom's Hardware video hierarchy recommendation of best bang for the buck was a strong factor.

My EVGA GTX 670 (4GB, Superclocked) could barely handle five clients on Ultra with DX 11, in and around Orgrimmar.
The game was playable on those settings, but in any kind of a challenging environment (raids and/or mass pvp) it wouldn't have been enjoyable.
The card itself wasn't pushed that far, in terms of graphical power, but it did use up almost all of the 4GB of its ram.
With DX9, the card didn't have any issues; I had reduced the settings to approximately medium/high and medium/low, as I was planning on 10-boxing on the one machine... never got there, with the Follow removal from battlegrounds.

[spher0boom]

The benchmarks I've seen for the 670 placed it around 80-90% of the performance of the 680 (links were from Tom's Hardware).
When I got my 670, it was basically 66% of the price of a 680.
The Tom's Hardware video hierarchy recommendation of best bang for the buck was a strong factor.
That's exactly how i ended up with the card as well.



My EVGA GTX 670 (4GB, Superclocked) could barely handle five clients on Ultra with DX 11, in and around Orgrimmar.
The game was playable on those settings, but in any kind of a challenging environment (raids and/or mass pvp) it wouldn't have been enjoyable.
The card itself wasn't pushed that far, in terms of graphical power, but it did use up almost all of the 4GB of its ram.
With DX9, the card didn't have any issues; I had reduced the settings to approximately medium/high and medium/low, as I was planning n 10-boxing on the one machine... never got there, with the Follow removal from battlegrounds.
Were all clients lowered to those settings, or just the slaves? I'm looking for high/ultra on the main and medium-ish on the slaves. I'll take a look at the eye candy differences between 9/11 and see if the differences will matter to me, but its very encouraging to know you had no issues in dx9.

Thanks!

[Ualaa]

With DX9, and five accounts, Ultra across the board was playable at a decent performance.

However with DX11, and the same five accounts, while the game would load and I would consider if playable, it would have been less than fun in congested locations.
Orgrimmar, around the auction house (wasn't terribly populated at the time), was noticeably spikey in performance with FPS jumping all over the place.
A mass battleground combat would not have been smooth; I'd assume the same would be true for a 25-man type raid.

I pretty much only cared about the pvp aspect, towards the end (last couple years) of boxing Wow.
So, as far as settings went, I wanted maximum for view distance of characters and objects.
But didn't mind so much if the settings (eye-candy) was lower for pretty much everything else.
I essentially put the main on Higher settings, and then disabled Shadows, Water effects, Reflections, and turned down things like Spell effects and Weather.
The slaves did the same, but started on Medium settings, and then had view distance type effects cranked upwards to near maximum.

With IS Boxer, setting up a mapped key to run for the current window (and another for all other windows), which runs on any client switch was fairly easy to set up.
And the closer the settings were for between the two, the faster the swap was...
I lowered the mains settings a fair ways, to enable absolutely instant swaps between any two toons.
If I didn't mind a 1.5 seconds of... the game is adjusting effects and I cannot do anything... the main window could have been on much higher settings; I'd have likely gone that route, if PvE instances were still my focus and instant swaps weren't as essential.

Another consideration for me, was the intention to 10-box on the one machine.
I built the machine with that in mind.
So I lowered the settings, quite a bit lower than I'd actually need for smooth 5-box performance, with the intention of not having to change the settings once the extra five accounts were added in.

Warcraft is more CPU dependent than Video dependent, for most of the graphics and effects in the game.
Video ram seems to be the limiting factor, when attempting to launch a boatload of accounts.
I'd imagine both CPU and Video card power, would determine how enjoyable the play experience is... but enough raw video ram is required to even launch the clients.

I was using a 6 core processor, a GTX 670 with 4GB of ram and 32GB of system ram.
GPU-Z showed the video card at 60-70% capacity on DX11 x5 Ultra settings; it also showed the ram at 3.5+ GB, and that in a lower graphic area (Orgrimmar, as opposed to a newer zone with better graphic effects and such).
The cores were not pushed hard at all.
And the system ram wasn't even slightly an issue.
The dual SSDs, in raid 0, weren't being pushed hard.
Here's my system:

Ualaa boxes with:
OS: Windows 7, Professional 64-bit
Case: Antec 1100 Gamer
PSU: Corsair Modular 1050 Watt
MB: Asus Sabertooth, X79.
CPU: i7 3930K (6 Cores + 6 Logical Cores)
OC: @4.5GHz, on Air with a Noctua NH-D14
Memory: 32GB Corsair Vengeance DDR3 1600MHz, CL8, Low-Profile Ram
Video: eVGA GeForce 670 GTX, 4GB Superclocked
OS Drive: OCZ Vertex3, 120GB SSD x2 Striped (Raid 0)
Storage: Seagate Barracuda, 3TB x2

Streaming on Twitch.TV with XSplit Broadcaster
Shaw Broadband 250 (250Mbp/s Download, 15Mb/s Upload)

[spher0boom]

With DX9, and five accounts, Ultra across the board was playable at a decent performance.

However with DX11, and the same five accounts, while the game would load and I would consider if playable, it would have been less than fun in congested locations.
Orgrimmar, around the auction house (wasn't terribly populated at the time), was noticeably spikey in performance with FPS jumping all over the place.
A mass battleground combat would not have been smooth; I'd assume the same would be true for a 25-man type raid.

I pretty much only cared about the pvp aspect, towards the end (last couple years) of boxing Wow.

I was hoping for better, so I may have to return he card and get a better one. My interest is in 5 boxing my own instances, or doing 10-mans with a friend who also 5 boxes. PvP or 25 mans don't really interest me any more. If I do 10 mans I can use 2 computers and i'll be fine in dx9 in really low mode on the second one.


So, as far as settings went, I wanted maximum for view distance of characters and objects.
But didn't mind so much if the settings (eye-candy) was lower for pretty much everything else.
I essentially put the main on Higher settings, and then disabled Shadows, Water effects, Reflections, and turned down things like Spell effects and Weather.
The slaves did the same, but started on Medium settings, and then had view distance type effects cranked upwards to near maximum.

With IS Boxer, setting up a mapped key to run for the current window (and another for all other windows), which runs on any client switch was fairly easy to set up.
And the closer the settings were for between the two, the faster the swap was...
I lowered the mains settings a fair ways, to enable absolutely instant swaps between any two toons.
If I didn't mind a 1.5 seconds of... the game is adjusting effects and I cannot do anything... the main window could have been on much higher settings; I'd have likely gone that route, if PvE instances were still my focus and instant swaps weren't as essential.

My main is the only one i care about about the advanced effects for, and my intent is that it will be in a full-screen, non-swappable window on my primary 1980x1200 screen. A decent view distance is mostly what i care about for the slaves. I'll leave on what i can, but it wont kill me to turn off most things. Smooth game play is my primary concern.

My second monitor is a 1080p in portrait mode. I intend to have two slave windows at 1080 x whatever, in rows 1 and two, and two 540 x whatever windows side by side in row 3. Row 1 (healer) will not be swappable like the main. Rows 2 and 3 (dps) will be the only swappable windows. However, i haven't played with ISBoxer to see if this configuration is supported, so it may change if i can't do that. If i can i can easily segregate the quality settings only for the swapping windows as you described. However, the delay would be fine for me.



Another consideration for me, was the intention to 10-box on the one machine.
I built the machine with that in mind.
So I lowered the settings, quite a bit lower than I'd actually need for smooth 5-box performance, with the intention of not having to change the settings once the extra five accounts were added in.

Warcraft is more CPU dependent than Video dependent, for most of the graphics and effects in the game.
Video ram seems to be the limiting factor, when attempting to launch a boatload of accounts.
I'd imagine both CPU and Video card power, would determine how enjoyable the play experience is... but enough raw video ram is required to even launch the clients.

I was using a 6 core processor, a GTX 670 with 4GB of ram and 32GB of system ram.
GPU-Z showed the video card at 60-70% capacity on DX11 x5 Ultra settings; it also showed the ram at 3.5+ GB, and that in a lower graphic area (Orgrimmar, as opposed to a newer zone with better graphic effects and such).
The cores were not pushed hard at all.
And the system ram wasn't even slightly an issue.
The dual SSDs, in raid 0, weren't being pushed hard.


I have two i7s with 32GB and SSDs. I'll make sure to stick with a 4GB card if I need to upgrade, which it sounds like i may (i get the feeling I'm on the border). Any other upgrades would require me to build a new PC, and I'd rather buy 2 video cards and use the computers i have.

Thank-you very much for your post, the performance numbers give me a much better feeling about where I'm starting. I finally get time to play around with everything tonight, so i should know tomorrow if i need to get a new/additional card for the second PC.

[spher0boom]

I finally got to play around with this tonight, and it looks like the GTX 670 will do everything I need.

My Setup:
- i7-3770K (4 cores, hyperthreading)
- 32GB RAM
- Intel 240GB SSD
- ASUS GTX 670 4GB
- Main Monitor: 1920x1200 Windows Fullscreen @ Ultra with 4x Multisampling, DX11
- Secondary Monitor: 1080x1920 (Portrait) with 4 slaves @ 1024x768 with 1x Multisampling

With the slaves at ultra movement was a big jerky ... the video was mostly smooth, but the main seemed to stutter-step while running through Durotar (100+fps) and Orgrimmar (35-ish fps). Changing the slaves to high gave me smooth movement throughout Durotar (forgot to get FPS) and Orgrimmar (35-45-ish fps). Orgrimmar was not busy at the time. These are all level 1's, so I don't have a lot of areas to check ... I haven't activated my main accounts yet. The Video Card never got above 90% utilization, and the CPU never got above 75%. While not very scientific, it seems that the card should provide very playable results for me since I have a lot of leeway left in the video settings.

These settings were WoW native. I was not using IS Boxer as I couldn't get it to work correctly - it would load the first game but never hook it, and it gave me no errors in the console to track down. After 2+ hours digging through the forums and wiki i gave up and just loaded the games manually. I'm hoping i can get ISBoxer working, as WoW won't let me load anything bigger than 1024x768 on the second monitor because of it's resolution, so I can't create a 16x10 version and scale it using hotkeynet.

[spher0boom]

I finally got ISBoxer working with my initial configuration. ISBoxer not hooking the clients was due to a DirectX issue ... the installer must not have completed successfully during the ISBoxer install for some reason. Once I re-installed DirectX it ISBoxer worked great.

All my 'tests' are a set of level 1's running around Orgrimmar and Durotar.

My Setup:
- i7-3770K (4 cores, hyperthreading)
- 32GB RAM
- Intel 240GB SSD
- ASUS GTX 670 4GB
- Main Monitor: 1920x1200 Windows Fullscreen @ Ultra with 4x Multisampling, DX11
- Secondary Monitor: 1080x1920 (Portrait) with 4 slaves @ 1080x675 with 1x Multisampling

I initially had performance issues when using ISBoxer. I went through the Wizard and used the same settings as running wow normally, but performance wasn't very good, even though the numbers all looked good (FPS, RAM, VideoRAM, Memory, CPU). Movement was very jerky, especially for the main toon. ISBoxer was launching all my slaves on my main monitor and then moving them to the secondary. I found an old post in the forums that said you need to launch the game on the monitor you will use it on for WoW, and doing that seemed to have fixed the issue. With the main on Ultra and the Slaves on high the game was very fluid and swaps on toons 3,4 and 5 were instantaneous (1 and 2 were excluded from swapping.

Can anyone tell me if launching and playing on the same monitor still a best practice to avoid performance issues? The post I found was quite old, and I wondered if I might have done something else to fix the problem and not realized it?

I was unable to test all 5 clients at 1920x1200 in Ultra as MiRai suggested (since he said most users box all clients at the same resolution), as I couldn't figure out how to get it to work. I had 3 different swap groups defined, and since the main toon/monitor was excluded from swapping I think ISBoxer just sized the rest of the clients to the largest window (1080x675) in the swap group, rather than the largest being used. Therefore my setup is using 2 different resolutions.


This is the second setup I tested:
- i7-2600K (4 cores, hyperthreading)
- 32GB RAM
- Intel 120GB SSD
- ASUS GTX 670 4GB
- Main Monitor: 1920x1200 Windows Fullscreen @ Ultra with 4x Multisampling, DX11
- Secondary Monitors: 2 slaves @ 1680x1050 fullscreen, 2 slaves @ 1280x1024 on a 1680x1050 monitor, with 1x Multisampling

With the slaves at Ultra or High, movement was what I would consider jerky, particularly on the main. High was better, but Good finally gave me nice smooth movement across all the toons. When I set it up in ISBoxer I again initially had issues. My primary CPU was topping out frequently. Changing the CPU distribution in ISBoxer fixed the issue and gave me nice smooth movement again.

I added an old BFG 8800GTS 640MB OC card in to add two additional 1280x1024 monitors for the last two clients so each slave gets their own monitor. I was surprised it could handle the two clients, but it does without issue. The two on the 8800 had to stay at good, but this allowed me to up the other two clients on the 670 to High and still have nice fluid movement. However, the noise and heal will probably force me to upgrade 8800.

With both cards and each client full-screen on it's own monitor I get the following resource utilizations at 2am in Orgrimmar:
- GTX 670 - GPU load of 60% or less, with an MCU that was never over 25%.
- 8800GTS - GPU load of 75% or less.
- RAM - 1GB for the main, 725MB for each slave.
- CPU - Averages 45%, peaks at 75%.

The CPU was by far the biggest Constraint. I tried several CPU distributions to try and keep each CPU instance from topping out, which caused performance issues, but no matter what I did some were always peaking, and some were really low. In the end I assigned the main to all 8 instances, and then assigned each slave to 1 instance on two different cores. This worked well and really smoothed out the CPU graph and pretty much eliminated the topping out.

After playing WoW on 5 PCs for several years before taking a break it just shocks me that I can do 5 full-screen monitors / large windows on a single video card with very reasonable settings. A single PC and ISBoxer will make multi-boxing much more enjoyable than my previous setup.

Thanks for all the help guys, especially Ualaa and MiRai. Your posts really helped me narrow down my issues quickly and got me up and running fast with very little hassle. I really appreciate it!

-spher0boom

[Lax]

Can anyone tell me if launching and playing on the same monitor still a best practice to avoid performance issues? The post I found was quite old, and I wondered if I might have done something else to fix the problem and not realized it?
If there was a post that said it should be on the same monitor for performance reasons, it would have been referring to limitations described here: http://isboxer.com/wiki/GPU_Management. Same monitor should only matter if you're using Windows XP (or earlier) OR the monitors in question are handled by different GPUs. For example after you added your second GPU, you wouldn't want one to cross from one GPU to the other, because it would incur a significant performance penalty (documented under GPU Management, and also documented by Microsoft at a link provided in my GPU Management article).

As far as your CPU being so much of an issue, did you consider the effect of Addons? Depending on your WoW Addons, disabling them could smooth out your CPU usage a bit

[spher0boom]

Thank you for the link, I missed that. I should have checked the ISBoxer site first, since you generally document things very well ... my fault :)


As far as your CPU being so much of an issue, did you consider the effect of Addons? Depending on your WoW Addons, disabling them could smooth out your CPU usage a bit

The only add-on I had enabled was the ISBoxer add-on. I was debating on a CPU upgrade, but turning the other add-ons back on may make that decision for me :)

[spher0boom]

If there was a post that said it should be on the same monitor for performance reasons, it would have been referring to limitations described here: http://isboxer.com/wiki/GPU_Management. Same monitor should only matter if you're using Windows XP (or earlier) OR the monitors in question are handled by different GPUs.

Ok, after reading this I see I must have inadvertently done something to fix the issue and not realized it.

I'm trying to get a newer second card (something quieter and cooler) to handle the last two monitors, and I need to use the VGA inputs on them. They are the monitors I use least, so after reading the link you posted it seems like I should keep the 4 monitors I use most on the same GPU for the best performance, Correct?

To do that I have to use the HDMI output of the 670 with an HDMI-to-DVI cable. I'm wondering if I'll see any kind of performance penalty on that output/monitor? I wouldn't think so since that output won't be used for audio, so the PC shouldn't be sending any or have any conversion to do, but I thought I'd ask for confirmation.

Thanks!