Poll, does your server idle at 32 or 50fps ?

[darkpeace]

Poll, does your server idle at 32 or 50fps ?
Can't really make a Poll, so just post



eg: If you are in lobby as admin or voted admin, and CPU load on server is 0-5% does '#MONITOR 1' report ?
A) 32fps - 33fps
B) 48fps - 50fps
C) More than 50fps
D) God No !!! - Less than 28fps in loby

(for Xeons/Opterons use the CPU assigned that server if multiple CPUs are hosting multiple servers, as it ain't using 2 Opteron CPUs to get 50fps contantly as far as I can see)


Personally for me it was 50fps, then without and addons being put to server it is suddenly 32fps

I mean WTF ?

Best I've seen was 55-56fps on a SuSe Linux Server that was idle (doesn't matter what CPU it had, it was not at 100% load in the loby)
And to re-iterate, OFPR_Server.exe runs on around 6 threads, but if you use HyperTheading set affinity to CPU0 as the threads compete with each other, they do not co-operate and increase speed. Not doing so actually causes an effect similar to lag (it is not comms related, put appears to be at first glance)

Ditto for Dual Opterons / Xeons, 1 server per CPU, not 1 server hosted over 4 CPUs with cache coherency causing more performance problems than benefits



Cheers to anyone that posts useful info (eg: how to lock at 50fps maximum in loby, or offset in hex of the 50fps limiter in OFPR_Server.exe, etc)

Would be nice to have Dual Operton or Athlon 64FX 2.6Ghz Winchester core server, with an unlocked FPS limiter, and idle the beast at 255fps, but prob only get 45fps-60fps in MFCTI

[darkpeace]

WTF ?
Considering I posted this 2 days ago, was expecting at least 1 reply

Oh well, I know 2 servers in Australia (Can't say which 2 though) that idle at 32fps, and everyone claims it is ment to idle at 50fps.

Figured someone out there may have made a simple reghack, after finding out what causes it.

[benu]

Ours idles at 50fps.

[darkpeace]

Yeah all I need now is:
-OFPR_Server.exe version
-OS Version / type
-Direct X installed on server
-CPU / bandwidth, the basics from the end of flashpoint.cfg
-Any other important detail

But nice to know it idles at 50fps as it should

I have seen some SuSe Linux server idle at 55fps before (and sustain it too)

Just want the best of the best as a server, give it time, one day it will happen (just software holding me back, not hardware)

[RN Malboeuf]

you'll find all the linux servers should be at 50
most win2k should be at 32
XP shows us 50
most have known this for years its why the linux guys said their servers were beter, but linux does not run cleaner then win2k, we have tested more servers then any one else
and regardless of FPS the P4 win2k servers are the best and thats why you'll see more of them along with XP


and as far as im concened the #monitor is not very acurate since i run netstat live (analog X) and OFP's out put is way off of what its really doing or not doing

I think its more of an OS issue then a settings issue

[benu]

As Malboeuf said, we are running linux. Has been Redhat on the first server, Debian on the second, Suse and now gentoo on the third. Kernel version is 2.6.7. I did install a 32bit version first but will try to move to 64bit soon. Man, am i happy to be away from Suse.

OFP is v1.96.

CPU is an AMD64 3000+ with 2gb of ram.

Bandwidth is 100mbit NIC to switched network, 6.4gbit internet connection from the cc.

DirectX 0.0

But i don't think this really matters as i think we had 50fps with the other two serves too, so it can't really be hardware related. Yesterday it even sometimes showed 52fps btw.

[RN Malboeuf]

sexy, you going to throw MFCTI1.16 on that so we can test, this will help i my august purchace of a P4 3.4oc4.0 compared to that machine, i know the P4 will be sweet, but the 64s may/should be better

[darkpeace]

To isolate what part(s) of a system OFPR_Server.exe hammers when running MFCTI. (does it hammer plain MIPS, FPU/SEE or is memory performance of the utmost importance ?)

The following link contains all the SANDRA 2004 SP1 CPU Benchmarks:
Drystone (MIPS/ALU) & Whetstone (FPU/SSE) & Whetstone SSE2

http://www.users.on.net/~darkpe....RKS.xls

I will post again with the memory benchmarks, but everyone knows the Athlon 64 + Dual DDR (various 3rd party chipsets), is a fair margin ahead of the Pentium 4 + Dual DDR (i875p, i865pe & select P4 SiS chipsets)



The downside is, even though the Athlon 64 performs better in MFCTI (as far as I know, but the P4EE+2MB L3 cache *might* beat it out) the cost is too high for many.

Thus mainboards using the decent SiS chipsets listed at;
http://www.sis.com/products/index.htm#chipset
; are very good value combined with a Pentium 4.

Of course any server should be tested with;
http://memtest86.com
; and overclocked to perfection (without affecting stability, just make sure all components are solid) before being handed over to a server-farm, as the support (and hardware testing ability) of most of them leave much to be desired.


Regards,
Tabris.DarkPeace


eg: Not the fastest around but a Pentium 4 overcloced with a 1000 FSB (+25%) with Dual DDR 500 (PC4000) (+25%) performs very well with its 8gb/sec peak of memory bandwidth using a i875p chipset based mainboard. This way the RAM runs in sync with the Northbridge, and reduces latency.

A 1200 FSB (+50%) is very nice, but would require either Dual DDR 600 (PC4800) which needs a special RAM Booster unit (as seen at www.tomshardware.com) to attain such high memory speeds -OR- would require Dual DDR2 -OR- a RamBus chipset such as the SiS R659, to feed the 9.6gb/sec FSB (1200mhz) of the 50% overclocked CPU.

I am currently unsure if Athlon 64 chipsets (and the currrent AMD-64 die) can handle such an overclock.