Check your CPU>GPU GPU<CPU bandwidth etc.

[jamesf1 0]

your ASUS P5K contains P35 chipset which supports only PCIe 1.1 so thats why it reports as not 2.0 :)

Doh, of course...

and definitely not AGP :)

I was reading back over that just now and went "whoops". It was late, and I'd not had my meds ;) I have the PCI version... as the original PCI-E version was plagued with latency problems that the PCI version didn't have.

[=S= Den 10]

So, installed Vista64, got another resultats in PCIeSpeedTest.

device0

CPU-GPU 4657

GPU-CPU 1750

device1

CPU-GPU 4556

GPU-CPU 5303

But why is device1 so different in GPU-CPU resultat?

[kklownboy 43]

very useful these but any programs that allow us to optimise the findings brought these programs?

Bios updates, MB drivers, extra card in a extra pci-e slot, or removing a card ( snd, lan, ect) from a pci-e slot, turning off functions and controllers in bios you dont use.

[Hermes 10]

You all know that ATI never get that high Bandwith like NV!?

Its the Bandwirh of the Cardbus, its an physical Limitation. So what are you all doing here?

If you own a ATI Card, most of the Bandwith are ok.

At first you must clear what physical Bandwithlimitation has your Card/Chip and than test it.

Sorry for my bad english!

[moosenoodles 0]

No Burst

[=Spetsnaz= 0]

i can't seem to download the program it directs me to another page

[Dwarden 1124]

You all know that ATI never get that high Bandwith like NV!?

Its the Bandwirh of the Cardbus, its an physical Limitation. So what are you all doing here?

If you own a ATI Card, most of the Bandwith are ok.

At first you must clear what physical Bandwithlimitation has your Card/Chip and than test it.

Sorry for my bad english!

sorry i dont get what You mean ...

i get usually way higher values with ATI than NVIDIA cards ...

it also vary a lot with different drivers and OSes and chipset quality makes huge difference too ...

[Hermes 10]

sorry i dont get what You mean ...

i get usually way higher values with ATI than NVIDIA cards ...

it also vary a lot with different drivers and OSes and chipset quality makes huge difference too ...

the bandwith of NV highend gfx cards is higher than the bandwith of the highend ATI gfx cards.

in some cases the pysical bandwith of a lowendcard is much higher than a faster card.

Bantwith is one factor of the speed but only one.

My 4870X2 Bandwith are 749 MB/s up an 3260 MB/s down.

PCIe Test was 5036 and 6779.

I run Crysis @1920*1200 @ultrahigh 4*AA 16*AF with Avarage 36 FPS

I score 3Dmark Vantage with 16256 GPU Score.

So what does the Bandwithtest says in my case?

If you dont know the physical limitations of your Card, the test cant say the result ist good or bad and the tool doesnt know how a modern gfx card work.

Bandwith is only one thing of the gfx card speed!

[kuss 10]

P5w dh Deluxe

E6600@3.8

ATI 4890

8GigRam

Vista x64

http://img36.imageshack.us/i/d3xk.jpg/

http://img37.imageshack.us/i/gpuzf.jpg/

http://img23.imageshack.us/i/testnky.jpg/

what can you tell me lol im lost i think is it ok

Edited June 16, 2009 by kuss

[PhatVybz 10]

i7 940 at 2.93Ghz

Ati radeon 4890 X2 crossfire enabled

asus gaming rampage 2 extreme x58

Triple channel ddr 6gb

the PCIespeedtest results :

Device 0 :CPU -> GPU = 4.378GB,s on 536870912 bytes

GPU -> CPU = 1,136GB,s on 131072 bytes

Device 1 :CPU ->GPU =4,427GB,s on 536870912 bytes

GPU ->CPU=3.239GB,s on 4194304 bytes

D3Dbantwitch kegetys test:

VM to SM download :359.89 mb/s

SM to VM Upload :3596.26mb/s

DPC latency: is not stable going from 25us up to 93us that is my absolute maximom in a minute

betwee, 1 minute and 5 it changes to 132us

after 5 minutes sometimes it increase to 250 us

There you go

[=S= Den 10]

kuss

Change mainboard for use PCI Express 2.0 :)

Your 4890 can make much better resultats.

Edited June 16, 2009 by =S= Den

[kuss 10]

kuss

Change mainboard for use PCI Express 2.0 :)

Your 4890 can make much better resultats.

ok cool thx but what you think i should get?Foxconn BlackOps Intel X48 (Socket 775) PCI-Express DDR3 Motherboard?

Edited June 16, 2009 by kuss

[waxman 5]

----------------------------------

Direct3D 9 Bandwidth test v1.0

by Kegetys - www.kegetys.net

----------------------------------

Using GetRenderTargetData for downloads, UpdateTexture for uploads.

VM = Video memory

SM = System memory

4.00MB buffer

Adapter 0 (\\.\DISPLAY1): NVIDIA GeForce 9800 GTX/9800 GTX+

VM to SM (Download): 2279.46 MB/s

SM to VM (Upload): 2387.94 MB/s

Adapter 1 (\\.\DISPLAY2): NVIDIA GeForce 9800 GTX/9800 GTX+

VM to SM (Download): 2112.44 MB/s

SM to VM (Upload): 2408.13 MB/s

DPC

~ 18 - 22us with a peak od 50us ever 10s or so

Not to bad though a bit short on the SM - VM though i am running 2 monitors wonder if that affects it much.

Seems to be pretty average in all. i wonder if those of you who are experiancing low results have many processes kicking off i noticed spikes etc. when opening windows and doing dtuff both on DPC (expected) and on the vid BW.

I did noticed quite variable results running the Bandwidth test a few times though.

System:

I7 920 @stock (2.67Ghz)

Asus P6T MB

12GB 6x2 GB Patriot 1600Mhz DDR3 (8GB ram drive holding ARMA currently)

ASUS EN9800GTX+

320GB hitachi 7K1000 (soon to be 2 raid 0 after RMA)

PCI X-Fi Fatality

win Server 2008 x64

[Potatomasher 0]

Adapter 0 (\\.\DISPLAY1): NVIDIA GeForce 9800 GT

VM to SM (Download): 514.38 MB/s

SM to VM (Upload): 675.28 MB/s

That 9800 GT is 512mb PCI-E card so my score is fine i suppose ?

[Dwarden 1124]

the bandwith of NV highend gfx cards is higher than the bandwith of the highend ATI gfx cards.

in some cases the pysical bandwith of a lowendcard is much higher than a faster card.

Bantwith is one factor of the speed but only one.

My 4870X2 Bandwith are 749 MB/s up an 3260 MB/s down.

PCIe Test was 5036 and 6779.

I run Crysis @1920*1200 @ultrahigh 4*AA 16*AF with Avarage 36 FPS

I score 3Dmark Vantage with 16256 GPU Score.

So what does the Bandwithtest says in my case?

If you dont know the physical limitations of your Card, the test cant say the result ist good or bad and the tool doesnt know how a modern gfx card work.

Bandwith is only one thing of the gfx card speed!

erm ?

in pure numbers of cards You may be right

~Pixel Fill Rate , ~Texture Fill Rate, Video Memory Bandwidth (GPU-VRAM, dont mismatch with SMtoVRAM)

HD4890 13.5k MPixels, 34.0K MTexels, 125 GB/s

GTX285 20.7K MPixels, 52.0K MTexels, 159 GB/s

bandwidth to GPU card depends on drivers,chipset, PCIe type and the card VRAM speed plus internal bus solution (memory bus), incase of multiGPU type of interconnect used (PCIe, CF connector, internalinterconnecting components etc), other devices, other drivers, operations going on and so on ...

yet don't forget to put into fray also several propriertary and ffa compression (texture,Z,stencil,color etc.) on which was always excelling ATI and ofcourse that depends what exactly is transfered too ...

btw didn't you mixed it up ? your system to vram memory is what you mark as down right ? :)

in Your case it says you got fast transfers for data from system memory to VRAM so there is no need to worry ... for you!

point of this thread is to help these who got issues on that front (SMtoVRAM) or various stutters (DPC spikes etc)

Edited June 16, 2009 by Dwarden

[flake 10]

another important factor to watch over would be DPC (Deffered Procedure Calls) latencies

download http://www.thesycon.de/dpclat/dpclat.exe

read more http://www.thesycon.de/eng/latency_check.shtml

on ideal idle system state they <20us

and on most of systems depending on number applications / work and various drivers <75us stable

with high spikes only when something intensive is going on (e.g. starting new application or doing CPU or I/o intesive ops)

if your system shows different values or random / looping spikes then Your system gunna have perf issues with realtime data streams

i did this and found a spikey driver that i disabled. has had a marked positive effect on streaming quality, worth a shot.

[Kenjineering 10]

Hey Fellas!

I was told that this thread could answer why I am getting such bad performance no matter what in ARMA2.

System Specs

Core i7 920 @3.5Ghz (Hyperthreading + Turbo Enabled)

6GB RAM (1333Mhz) Kingston

Intel DX58SO "Smack Over" Motherboard

Bios Version: 4158

2x GTX280 1G Sli

Nvidia Drivers: 186.18 + EVGA Sli Enhancer

PhysX: 909.04.28

Windows Vista Home Premium SP2 64bit

1000W PSU

DPC Result

120μs ±20μs.

D3D Bandwidth Test Results

Idle:

VM to SM = 1409.7 Mb/s

SM to VM = 1658.46 Mb/s

After I minimized ARMA2 I got something like thi: (I forgot to record them down)

VM to SM = 1804.6 Mb/s

SM to VM = 1658.46 Mb/s

Questions

What the f?

How does this help?

Edited July 3, 2009 by Kenjineering

[Dwarden 1124]

erm if you ran these tests while ARMA 2 was running then that wasnt good idea :)

[Kenjineering 10]

erm if you ran these tests while ARMA 2 was running then that wasnt good idea :)

Good idea fail.:dance1:

The 2of2 D3D test was run while Arma was minimised. The first was just in normal idle.

My problems are probably related to micro-stuttering or this mystery Nvidia 2xx series performance issue with ARMA2

[Dwarden 1124]

for some reason your SM to VRAM is extremely low , shall be like double