Pages: 1 [2] :: one page |
|
Author |
Thread Statistics | Show CCP posts - 5 post(s) |
|
CCP Lingorm
C C P
|
Posted - 2008.10.03 12:58:00 -
[31]
Originally by: Khaleb Flosk Hi Guys,
Sorry for not posting in the Hardware Section but it just fits here perfekt. I was wondering about your EvE Ingame Settings for running multiple Clients. IŠve got a brand new Computer: - Intel E8500 @ 3,6Ghz on a G35 Chipset - 4GB DDR2 @ 888Mhz (testet stable) - ATI Radeon 4870 512MB @ catalyst 8.7 - Vista 64Bit Sp1
IŠm capable of running one Premium Client in maxed out 8xAA 1900x1200 without stressing my machine at all. But at the time i start a second client, my machine gets on his knees beging for stoping (means whole windows is stressed to 100%, stuttering). Funny thing about it is, that this allready starts when the second client shouldnŠt produce any graphics load at all (means login client) so it should be a memory related problem.
Whole Thing started two or three patches ago...and i tried pretty much everthing that came in my mind. - installing in a dedicated directory - running defrag - experimenting with my catalyst settings.
is there any "must do" Parameter / cache setting that allows you to run multiple clients smooth.
thank you.
cheers
We have had some reports of this and I am currently running some extended benchmarks to see if we can isolate the problem and fix it.
CCP Lingorm CCP Quality Assurance QA Engineering Team Leader
Originally by: Lord Fitz Eve is to WoW as Wow is to an 8 player game of Unreal Tournament.
|
|
Thebro Nobrunder
Schrodinger's Renegades
|
Posted - 2008.10.03 13:49:00 -
[32]
Originally by: Yaholo A TRUE 64 bit app does offer a speed boost over a native 32 bit app, so this is a great reason beyond just the increased memory. However it would likely be easier to make the game better utilize multiple cores and make it where a quad core can be really leveraged for it's full power. This should require, I would presume, less of a re-write and actually give a nicer performance boost.
64bit programming only results in a speed boost in certain cases. Video decoding for instance when you can pack extra data in the 64bits and perform twice the work with one operation.
The main thing which reduces performance is that you are now transferring more data across the bus, Also your cache can now cache only about half the data.
These downsides can be partly avoided with clever coding but it's non-trivial work.
|
Khaleb Flosk
|
Posted - 2008.10.03 14:10:00 -
[33]
Quote: 64bit programming only results in a speed boost in certain cases. Video decoding for instance when you can pack extra data in the 64bits and perform twice the work with one operation. The main thing which reduces performance is that you are now transferring more data across the bus, Also your cache can now cache only about half the data.
there is pretty much everything wrong in this statement that could be wrong.
cheers
|
Thebro Nobrunder
Schrodinger's Renegades
|
Posted - 2008.10.03 14:28:00 -
[34]
Edited by: Thebro Nobrunder on 03/10/2008 14:29:19
Originally by: Khaleb Flosk
there is pretty much everything wrong in this statement that could be wrong.
cheers
I find your argument surprisingly lacking in content :P
I've been writing 64bit code since the mid 90's. While my statements are indeed a vast generalization... they are basically true.
|
Khaleb Flosk
|
Posted - 2008.10.03 14:45:00 -
[35]
Originally by: Thebro Nobrunder Edited by: Thebro Nobrunder on 03/10/2008 14:29:19
Originally by: Khaleb Flosk
there is pretty much everything wrong in this statement that could be wrong.
cheers
I find your argument surprisingly lacking in content :P
I've been writing 64bit code since the mid 90's. While my statements are indeed a vast generalization... they are basically true.
sorry...still at work.
twice the data == wrong;
twice the work == wrong;
thing which reduces performance is that you are now transferring more data across the bus == wrong .. cause the main performance trade off is the "more calculation" of the fattened code...you get performance gains by staying in the cache levels with less access time...next level would get you to your harddrive;
your cache can now cache only about half the data == wrong;
didnŠt want to flame...but statements like yours spread this fugly 64bit rumors across the internet.
cheers.
|
Thebro Nobrunder
Schrodinger's Renegades
|
Posted - 2008.10.03 15:04:00 -
[36]
Originally by: Khaleb Flosk
twice the data == wrong;
twice the work == wrong;
thing which reduces performance is that you are now transferring more data across the bus == wrong .. cause the main performance trade off is the "more calculation" of the fattened code...you get performance gains by staying in the cache levels with less access time...next level would get you to your harddrive;
your cache can now cache only about half the data == wrong;
didnŠt want to flame...but statements like yours spread this fugly 64bit rumors across the internet.
cheers.
Well, the simple act a dereferencing a pointer now has to pull 8 bytes of data from main memory instead of only 4.
For certain operations twice the work is quite simple. For example loading two 32bit words of data into a 64bit value and applying logical transforms on them effectively applies the operation to both 32bit works simultaneously.
As for cache... for 64bit code the data alignments are larger. This can be avoided but you have to intentionally do so. Therefore while your cache line is (for instance) 64 bytes... it can now cache only 8 64bit pointers instead of 16 32bit pointers.
Granted newer processors are being optimized better for 64bit operations but 32bit vs 64bit is still a consideration which must be considered carefully.
|
Khaleb Flosk
|
Posted - 2008.10.03 15:16:00 -
[37]
yeah...for a pro itŠs easy to find a vindication for every sentence youŠve typed..(exact or not cause numbers pointers != data) .. but if you write like a noob youŠve got to think like a noob .. otherwise nobody will understand what youŠve wanted to say .. and that gets us to this point ;)
cheers
|
Thebro Nobrunder
Schrodinger's Renegades
|
Posted - 2008.10.03 15:43:00 -
[38]
Originally by: Khaleb Flosk yeah...for a pro itŠs easy to find a vindication for every sentence youŠve typed..(exact or not cause numbers pointers != data) .. but if you write like a noob youŠve got to think like a noob .. otherwise nobody will understand what youŠve wanted to say .. and that gets us to this point ;)
cheers
Well, I've given you evidence on why 64bit code can be slower than 32bit code. What is your validation to the contrary?
|
Khaleb Flosk
|
Posted - 2008.10.03 15:55:00 -
[39]
Quote: What is your validation to the contrary?
[ ]"never pi*s against the wind. it will get your feet wet and stinky"(validated) [ ]"if you find a ho*ker with herpes, move along"(validated) [x]"pardon me, what was the question?
cheers
|
Generium
|
Posted - 2008.10.03 17:47:00 -
[40]
Originally by: Thebro Nobrunder 64bit programming only results in a speed boost in certain cases. Video decoding for instance when you can pack extra data in the 64bits and perform twice the work with one operation.
The main thing which reduces performance is that you are now transferring more data across the bus, Also your cache can now cache only about half the data.
These downsides can be partly avoided with clever coding but it's non-trivial work.
Since you have more registers available (3 times the count available) I would say 64 bit would result in increased performance, given ofcourse the software was optimized properly.
|
|
Thebro Nobrunder
Schrodinger's Renegades
|
Posted - 2008.10.03 18:16:00 -
[41]
Originally by: Generium
Since you have more registers available (3 times the count available) I would say 64 bit would result in increased performance, given ofcourse the software was optimized properly.
As long as your code fits entirely in cache... then that would be a big gain. besides... I thought x86 only added 8 registers in the more to 64bit? I could be wrong on that however.
|
Generium
|
Posted - 2008.10.04 16:37:00 -
[42]
Originally by: Thebro Nobrunder
Originally by: Generium
Since you have more registers available (3 times the count available) I would say 64 bit would result in increased performance, given ofcourse the software was optimized properly.
As long as your code fits entirely in cache... then that would be a big gain. besides... I thought x86 only added 8 registers in the more to 64bit? I could be wrong on that however.
Yes they only added 8 general purpose registers, so the total is now up to 16. But since you weren't able to use all 8 registers before, the amount of registers actually available to you is trippled. Further more they increased the size of all registers.
Besides they've added 8 128-bit XMM registers (SSE) so the total of XMM is now also up to 16.
Besides with larger registers 64 bit mathematical operations should have increased speed.
Further more with 64 bit processors came the ability to produce position independent code, since instructions can reference relative to the instruction pointer. |
gallminin
Gallente Silver Lining Corp The Black Omega syndicate
|
Posted - 2008.10.05 22:33:00 -
[43]
i have 6GB of ram and a computer that is more then capable to run the client in 64-bit mode.why dont you just have both?
---Asus M2N SLI Deluxe wireles edition.- XFX 8800GT 512MB DDR3 - 6GB G.Skill - AMD 6400+ am2 @3.2 with a zalman cooler--- |
|
|
|
Pages: 1 [2] :: one page |
First page | Previous page | Next page | Last page |