Log in

View Full Version : Where's the bottleneck? Public system review for upgrade options.



Zzyzxx71
12-15-2010, 01:48 PM
My current system is a bit dated, but it's been a solid workhorse.

With the installation of Cataclysm my boxing performance has dipped every so slightly below my threshold of satisfaction.

I 5 box, I have no intention of running 10 anytime soon.

Hardware configuration
------------------------------
Current setup is a 2+ year old XFX nvidia based chipset mobo running a non OC'd Q6600 quad core (I believe at 2.4ghz) with 8g of ddr2 memory.
Center and right monitors are driven by a BFG 280 GTX (1gb memory).
The center monitor is a samsung 23" running at 2048x1156. This is the primary window for ISBoxer.
The right and left monitors are Acer 22" displays with native resolutions of 1920x1080.
The right monitor is used for research/mumble/vent/TS/browsing/porn/etc.
The left monitor is driven by an XFX 9800 series nvidia card with 768mb of memory. It supports the 4 slave/non active windows.
The PC (64 bit Win7 Pro) is kept as clean as possible. I have a laptop running next to me which handles all that other crap that typically slows your machine down.

Software configuration
------------------------------
Software used is ISBoxer using the 2 step method
I have 2 wow folders residing on a 64gb SSD, 1 for my tank, the other for the slaves.
The active window (center) has an FPS limit of 60.
Non active windows have an FPS limit of 15.

Questions/Requests for those who reply:
--------------
1. Is 2048x1154 too large to virtually be running 5 wow's at?
2. Are there any blatent flaws of this setup or is it just getting old? I see no reason why it wouldn't stand toe to toe with any of the current high end i3's and any of the i5's currently available.

3. If you feel so inclined, I'd like to see peoples opinions regarding hardware upgrade/replacement paths. Please list the top 3 items which would increase performance.

In my opinion currently it's:
1. Replace primary video card, replace secondary video card with current primary card.
2. Upgrade memory to 12/16gb - Is Cata really that much of an increase taxing my system that much more? (although I've seen in other threads WOW limits itself to 2gb max, nonetheless, I'd be at 10gb) and 3. Processor (either OC or replace - which would mean CPU/Mobo/Memory).

I'm sure there's some tweaking that can be done. Any input is appreciated.

d0z3rr
12-15-2010, 01:53 PM
The q6600 is not a true quad core, it is two dual cores slapped on one thing.

Upgrade to a i5/i7 platform, if you don't want to spend money try the x6 from AMD. Either way get off of that ancient q6600 and you'll see a huge difference.

Your vid cards are fine.

BrothelMeister
12-15-2010, 02:08 PM
Your set-up is identical to the one I JUST replaced 3 days ago. Same Processor, and monitor array (I too use the samsung 2048x1152 so that I can have 1920x1080 in windowed format, with its correct aspect ratio), with 2 extra monitors. I had the 8GB ddr2 @ 800 MHz.

I felt this system was getting dated as soon as patch 4.0 came out because the graphical requirements, even with all the slaves turned down to absolute lowest settings, and the main turned down below half. Framerates of 9-10 on slaves with 10-15 on the main window when traveling,and under 10 fps on the main window during action is just not acceptable.

I will be playing wow for the first time tonight with my new set-up, which is an i7 950, with 12 GB of ddr3 1600 MHz RAM.

As far as updating your set-up, go for a 1366 socket, i7. The i5 is only 4 core 4 thread, whereas the 1366 socket i7's are 4 core 8 thread, and as a multi-boxer, we can assign different instances of wow to different threads, meaning we get to use the full potential of the processor. The i5 is mainly for people that want a cheap, fast processor, but don't intend to be doing lots of multi-tasking.

The 1366 socket also means you will be buying your RAM in multiples of 3, not 2. so keep that in mind. I went with 6 sticks of 2GB, but you could probably be fine with just 3 sticks of 2GB (which would actually be a downgrade from the 8 you currently have which is why I decided to go to 6 sticks of 2)

As far as the botteneck with your current set-up, I would wager a strong guess it's your CPU, as that was my bottlenect as well. Im not certain how well your GPUs are working, but they are probably not the bottleneck. To check this, just open the task manager, and go to performance; your ram will probably be around 6GB, but if your CPU stays in the 85% or more range when you are doing good stuff in wow (like a large world pvp event, raid, or just running through a city) then your CPU is the bottleneck for certain.

d0z3rr
12-15-2010, 02:15 PM
I wish people would not use the "open taskmgr and look at CPU util", it's not a true benchmark at all.

Every CPU can easily be maxed out just by running calculations in Calculator. If an i7 is at 100% cpu just from using calc.exe, doesn't that mean it sucks? Nope. What really matters is how long it is pegged at 100%.

A q6600 and i7 can both be at 100% utilization from rendering a video. The real thing that matters is how fast do they render the video? Just because an i7 is faster and newer, doesn't mean that it's cpu utilization in taskmgr will be reduced while rendering a video......

To realise the benefit of switching from a q6600 to an i5/i7/x6, you need synthetic benchmarks that give you numbers.

HPAVC
12-15-2010, 03:23 PM
I wish people would not use the "open taskmgr and look at CPU util", it's not a true benchmark at all.

Every CPU can easily be maxed out just by running calculations in Calculator. If an i7 is at 100% cpu just from using calc.exe, doesn't that mean it sucks? Nope. What really matters is how long it is pegged at 100%.

A q6600 and i7 can both be at 100% utilization from rendering a video. The real thing that matters is how fast do they render the video? Just because an i7 is faster and newer, doesn't mean that it's cpu utilization in taskmgr will be reduced while rendering a video......

To realise the benefit of switching from a q6600 to an i5/i7/x6, you need synthetic benchmarks that give you numbers.

While I agree that task manager is not a very useful took, The OP can still find their bottleneck with windows 7's perfmon easily given a bit of time and help.

The multiboxing run time are not going to be anything near what the same as benchmarks results. Sure if you invest balls amounts of cache into making a system that is to the nines, you won't have any problems (minus compatibility isses). But that is a huge waste of money. Even if money isn't an option, money could be better spent on crazy toys than just pushing a useless benchmark number higher.

Ughmahedhurtz
12-15-2010, 04:43 PM
Upgrade to a i5/i7 platform, if you don't want to spend money try the x6 from AMD. Either way get off of that ancient q6600 and you'll see a huge difference.

To realise the benefit of switching from a q6600 to an i5/i7/x6, you need synthetic benchmarks that give you numbers.

You mean like these synthetic benchmarks (http://www.tomshardware.com/charts/2009-desktop-cpu-charts-update-1/Performance-Index,1407.html)?

Q6600 2.4GHz --> i5 2.4GHz = 20% performance bump.
Q9450 2.66GHz --> i5 2.4GHz = 5% performance bump. (Q9450 2.66GHz is roughly equivalent to a Q6600 OC@3.0GHz)
Q6600 2.4GHz --> i7 920 @ 2.66GHz = 38% performance bump.
Q9450 2.66GHz (or Q6600 @ 3.0GHz) --> i7 920 @ 2.66GHz = 25% performance bump.

So. If you got a nice heatsink and did some very minor overclocking, you could easily match an i5 for performance and get within about 20% of an i7 for 1/6th the cost.

I really wish people would tell the whole story and provide sources and numbers instead of just being a jackass and calling peoples' hardware "ancient" and sucky and outdated. Sounds like Paris Hilton doing a PC review.

Furthermore, this whole discussion of CPUs gets blown up when talking about ISBoxer. Due to the way it (usually) runs all clients at max resolution and reading the OP where they are running at 2048x1154, there is a much larger impact on VRAM and GPU than there is on CPU. Especially with the Cataclysm engine updates.

Personally, I'd look at doing a leapfrog update of the 9800 to a more recent card like a GTX 470 for about $250 which will at least _double_ or _triple_ the graphics performance of your 9800. The Q6600 is fine if you're only running 5 clients, though I would definitely try stepping the 4 alts down to 800x600 or 1024x768 or so. You really don't need to run them at max rez, IMO. Maybe there's a technical limitation (like losing the mouse broadcasting feature) with ISBoxer that necessitates that but I'd be surprised if so.


[edit] On a related note, see this page for a glaring example (http://www.tomshardware.com/reviews/radeon-hd-6970-radeon-hd-6950-cayman,2818-15.html) of how synthetic benchmarks like 3DMark can drastically skew numbers versus what you'll see in real-world game scenarios.

Ualaa
12-15-2010, 05:33 PM
My previous system was:

Q6600, stock speed
8GB, DDR2 at 800mhz
4870X2 Video (2GB)
Asus Motherboard
Raptor 10k rpm HD.

Wasn't happy with the performance, could lose follow in Dalaran with all toons mounted on ground mounts. Follow frequently broke in Alterac Valley or Wintergrasp (if mounted).
Instances and 5-boxing elsewhere were fine, but had a lot of the settings down.



Decided to upgrade a couple of summers ago.
Went with:

i7 920, stock speed
12GB, DDR3 at 1600mhz.
4870X2 Video (2GB)
Asus P6T Motherboard
Raptor 10k rpm HD.

Had the computer store do the upgrade for me; I don't like mounting motherboards, having cracked one during installation quite a few years ago.
They goofed up and put Vista 32-bit on, instead of 64-bit.
Even with only 2GB of ram usable (video card was using 2GB), the performance increase was massive; aside from crossing phase points, it was impossible to break follow on epic flyers unless the lead toon had Crusader Aura on (and even then it was difficult). On ground mounts, without getting them stuck on corners it was impossible to break follow.
And this was with only 2GB of the ram.

Got the OS fixed.
Had some kind of a conflict between the Video Card, and one of the upgrades (Processor, Motherboard or Ram).
One or two slave windows would randomly freeze (requiring a reboot); the freeze could occur at any point, but always did when taking a gryphon anywhere.
Had this issue with both Keyclone and IS Boxer, so pretty sure it was hardware related.
Eventually (through changing one part at a time), I isolated the issue to the video card.
Downgraded to a GTX275, which has been much more then sufficient for 5-boxing Warcraft (the 275 benchmarks comparable to a 1GB 4870 (not a 4870X2)).

As an aside loaned the motherboard, video card, processor and ram to a friend. The video card is working great for him.

d0z3rr
12-16-2010, 03:23 PM
You mean like these synthetic benchmarks (http://www.tomshardware.com/charts/2009-desktop-cpu-charts-update-1/Performance-Index,1407.html)?

Q6600 2.4GHz --> i5 2.4GHz = 20% performance bump.
Q9450 2.66GHz --> i5 2.4GH.

I consider it ancient. I have one, hell, it runs two of my characters in my 5man team. So I guess I'm a jackass to myself?

Also, why did you link that benchmark? The real benchmarks would be 3d games, which, clock-for-clock, the i5 slaughters the q6600. I have an i7, the performance increase over my overclocked q6600 was exponential - in fraps, vegas, acid 4.0.

I'd love to hear why you think the q6600 is such great architecture compared to an i5.

Ughmahedhurtz
12-16-2010, 04:18 PM
I consider it ancient. I have one, hell, it runs two of my characters in my 5man team. So I guess I'm a jackass to myself?

Also, why did you link that benchmark? The real benchmarks would be 3d games, which, clock-for-clock, the i5 slaughters the q6600. I have an i7, the performance increase over my overclocked q6600 was exponential - in fraps, vegas, acid 4.0.

I'd love to hear why you think the q6600 is such great architecture compared to an i5.

The benchmark I linked is the overall score. A cursory examination would have shown you the relative performance gaps between the two for individual test cases.

Here's a breakdown (http://www.tomshardware.com/charts/2009-desktop-cpu-charts-update-1/compare,1403.html?prod[3519]=on&prod[2638]=on) for each individual test. Fraps, vegas and acid aren't 3d games, so by your own argument why point that out? Your argument is inconsistent. First, it's (paraphrasing) "you need to use synthetic benchmarks because the numbers don't lie" and now it's "well not those benchmarks because those numbers don't tell the whole story." Thus, my comment about sources and links to benchmarks that back up your assertions being good for clarity's sake.

Also, "clock for clock" is meaningless as repeated benchmarks have shown.

You're making assertions that the i5 is light years ahead of the Q6600, yet the benchmarks I'm looking at don't seem to bear that out. And we're talking about answering the question of which component(s) in the OP's system would best be upgraded to help performance given his stated scenario, which is heavily stressing the graphics side versus the CPU side.

Finally, you mention having an i7. Which i7 do you have that is exponentially faster than an OC'd q6600?

I'll make a suggestion: point out empirical numbers that show the cost-effectiveness of upgrading his Q6600 to an i-series versus upgrading his 9800 video card to a GT470.

Sam DeathWalker
12-16-2010, 04:58 PM
Its not the cpu its the (well nvida calls them north and south bridge not sure what intel is calling them) X58 that makes the difference. Not sure what Intel used prior to the X58 but its the X58 that blows away everything else, with its QPI and that other fast buss.

Ughmahedhurtz
12-16-2010, 05:24 PM
Its not the cpu its the (well nvida calls them north and south bridge not sure what intel is calling them) X58 that makes the difference. Not sure what Intel used prior to the X58 but its the X58 that blows away everything else, with its QPI and that other fast buss.

This is where I was trying to lead this: people aren't comparing apples to apples and are just throwing out anecdotal hyperbole.

Zzyzxx71
12-17-2010, 11:25 AM
OK - I've thrown together a list of options.

1. i7 cpu/mobo/8gb memory, no video upgrade ($550?)
2. AMD 6 core/mobo/8gb memory, new video card ($600)
3. New video card, new CPU chiller - overclock the shit out of it (Cheap)

I'm tempted to try #3 first.

MiRai
12-17-2010, 12:12 PM
OK - I've thrown together a list of options.

1. i7 cpu/mobo/8gb memory, no video upgrade ($550?)
2. AMD 6 core/mobo/8gb memory, new video card ($600)
3. New video card, new CPU chiller - overclock the shit out of it (Cheap)

I'm tempted to try #3 first.
This (http://www.dual-boxing.com/showthread.php?t=33360) is why I cannot recommend Option 2 with AMD's current state unless you can use the 6 core usefully in some other way.

Zzyzxx71
12-17-2010, 12:17 PM
Ouch - I knew there was a performance hit, I didn't know there was that much of a delta.

d0z3rr
12-17-2010, 04:13 PM
Its not the cpu its the (well nvida calls them north and south bridge not sure what intel is calling them) X58 that makes the difference. Not sure what Intel used prior to the X58 but its the X58 that blows away everything else, with its QPI and that other fast buss.

Yup, the bridges help too. And guess what? The q6600 cannot utilize those new bridges. Hence the term "architecture", the i5/i7 have better/newer architecture, both within the CPU and what they run on.

d0z3rr
12-17-2010, 04:15 PM
I'll make a suggestion: point out empirical numbers that show the cost-effectiveness of upgrading his Q6600 to an i-series versus upgrading his 9800 video card to a GT470.

Uh nope, I just realised I'm arguing with someone who thinks a q6600 is still a viable processor even though the i5/i7 are out. I'll agree to disagree.

Here is my opinion and experience:

I upgraded from a q6600 to an i7. I noticed far better performance in everything. If you don't believe me, that's kewl.

Zzyzxx71
12-29-2010, 09:50 AM
I did some more testing, turns out prior to Cata with5 clients running my cores would hover in the 70-80% range all the time.

After cata they're pegged at 100% all the time - I think it's the extra video rendering overhead that comes with cata.

When I raid with my guild (1 client) I turn up the grafx to Ultra and don't have an issue. It's definitely a processor issue.

Next paycheck, new board/cpu/memory will be on the way - I doubt I'll have to do anything with my video configuration.

Thanks for all the input.

Zzyzxx71
01-10-2011, 11:02 AM
To put some closure to this, I upgraded the cpu/mobo/memory yesterday. I went to a i7 2600k, 8gb 1333mhz memory on a new asus mobo - Will be upgrading to 16gb next check.

Same video cards, same hard drives.

Single client framerates fluctuate between 150-200 in Orgrimmar with 1 client running. With all 5 it caps out, 60fps on the primary window with video effects 2 steps up from the minimum, and 30fps with video effects 1 step up from minimum.

CPU utilization hovers around 20-30% at all times - it absolutely positively was the CPU. Thanks for all the input.

d0z3rr
01-18-2011, 04:44 PM
To put some closure to this, I upgraded the cpu/mobo/memory yesterday. I went to a i7 2600k, 8gb 1333mhz memory on a new asus mobo - Will be upgrading to 16gb next check.



Are you using triple channel? I went with 6gb so I could do triple channel - 3x 2gb sticks.

MiRai
01-18-2011, 04:47 PM
Are you using triple channel? I went with 6gb so I could do triple channel - 3x 2gb sticks.
2600K does not use triple channel.

Zzyzxx71
01-18-2011, 04:48 PM
Are you using triple channel? I went with 6gb so I could do triple channel - 3x 2gb sticks.

Nope - the socket 1155 boards all have 4 slots, so dual channel is what I'm stuck with. I went with 2x4gb, and will be getting another 8gb next payday.

An update to the story, Since buying it, I basically have spent the last week troubleshooting every iteration of blue screen known to man.

Eventually I started eliminating pieces of hardware and returned/replaced the motherboard on sunday. I haven't had an issue since then, so apparently is was the mobo. I currently have it running at a base speed of 3.8ghz, and it's peaked at over 4.5ghz under heavy load, that's with no perceivable increase in temps.

Once it's burned in I'll start playing with the overclocking, but it's not really necessary - the performance I'm getting right now is more than acceptable.