VOGONS


Does PCI-E 4x have enough bandwidth?

Topic actions

First post, by weldum

User metadata
Rank Member
Rank
Member

Hi, I'm working in one computer, I want to run some relatively new games on it but i'm getting very low FPS

MB: Asus P5VD1-X
Chipset: VIA PT880 Ultra
CPU: P4 631 3.0 w/Hyperthreading
1.5GB DDR DualChannel
160GB IDE Seagate
Windows XP SP3

So far I've tested with 3 video cards that I have available:

GeForce FX 5200 128MB AGP8x
GeForce 9600GSO 768MB PCI-E 1.0 16x @ 4x
Radeon 6450 1GB PCI-E 2.0 x16 @ 1.0 4x

First thing i've noticed, in older games, such as Half-Life, Incoming, Call Of Duty, Quake 3 and such, the 3 cards perform excellent.

Next, I've tried some "more demanding" games: GTA San Andreas, Call Of Duty 2 and Modern Warfare, UT2004 and UT3, Counter Strike: Source and CS Global Offensive, Left4Dead 2

The FX gets very low fps but still runs most of these games, the Radeon gets really low FPS (in a better spec pc gets higher than 60FPS at 720p with almost everything maxed out) and the 9600gso barely gets better performance, but still low.

I've noticed this a lot in CSGO, when the Radeon overclocked to 725/900 gets almost the same performance as the 9600GSO, but both gets nearly 15 FPS, no graphical adjustment makes it get better speed.

I'know that this machine is very slow but I do think it should be capable of more.

Is there any trick or performance boost possible for this? Thanks in advice.

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 1 of 27, by emosun

User metadata
Rank Oldbie
Rank
Oldbie

What what VERSIONS of the games are you running.

Ideally you want to be using the original disk versions of the games from the time. The worst scenario would be if you are trying to use newer steam versions. Do not worry about the 4x bus speed. Technically the bus speed is less important than the actual cards speed. Meaning that even the fastest 8x 4670 agp would have a hard time competing with a 1x bus speed pci-e gpu that was simply a faster/newer gpu.

Also with 1.5gb of ram you aren't running in dual channel so square up the ram to an equal amount.

As for the newer cards getting worse fps. you most likely didn't uninstall the drivers correctly using driver cleaner. On these old systems you cannot mix and match drivers willy nilly. You have to use driver cleaner and remove the drivers from the previous gpu completely otherwise the machine gets crippled performance.

Reply 2 of 27, by weldum

User metadata
Rank Member
Rank
Member
emosun wrote:
What what VERSIONS of the games are you running. […]
Show full quote

What what VERSIONS of the games are you running.

Ideally you want to be using the original disk versions of the games from the time. The worst scenario would be if you are trying to use newer steam versions. Do not worry about the 4x bus speed. Technically the bus speed is less important than the actual cards speed. Meaning that even the fastest 8x 4670 agp would have a hard time competing with a 1x bus speed pci-e gpu that was simply a faster/newer gpu.

Also with 1.5gb of ram you aren't running in dual channel so square up the ram to an equal amount.

As for the newer cards getting worse fps. you most likely didn't uninstall the drivers correctly using driver cleaner. On these old systems you cannot mix and match drivers willy nilly. You have to use driver cleaner and remove the drivers from the previous gpu completely otherwise the machine gets crippled performance.

Versions: These are the old retail version sold in CD/DVD, CSGO is an old version (one of the first) of the steam version.

Ram: I do have 1.5GB of ram, 2 modules of 256MB and 2 of 512MB, and they're interleaved. CPU-Z, Aida64 and SiSoft Sandra all of 3 say Dual Channel, and even in POST it says.

Drivers: I have 3 HDD, 160GB each, and the 3 have the same files and the same fresh windows installation, the only change is the video driver, to make tests on the fly. So if I put the FX5200, I simply change the HDD and boot.

Is there anything else that can be?

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 3 of 27, by emosun

User metadata
Rank Oldbie
Rank
Oldbie

Thats odd then.

Are all the motherboard drivers installed? Are you using asus's motherboard drivers and not the generic windows xp drivers? Does the board have the last bios update on it?

The machine should be able to handle games like san andreas just fine.

Try this , pull out some of the ram. Try the two 512mb sticks and two 256mb sticks alone in the machine and see if it affects anything. Maybe the mixed set , or one of the sets is bad or just conflicting somehow.

Also what's the cpu usage like?

Reply 4 of 27, by weldum

User metadata
Rank Member
Rank
Member
emosun wrote:

Thats odd then.

Are all the motherboard drivers installed? Are you using asus's motherboard drivers and not the generic windows xp drivers? Does the board have the last bios update on it?

I've tested with the Asus drivers, the latest via drivers and the generic windows drivers in both XP and win7, the bios is updated.

emosun wrote:

The machine should be able to handle games like san andreas just fine.

It does run flawlessly, even on the FX5200 (through lowering res), the thing is with latter games like Unreal 3 or CSGO, strangely Left4Dead 2 works almost perfect on the Radeon and the 9600GSO.

emosun wrote:

Try this , pull out some of the ram. Try the two 512mb sticks and two 256mb sticks alone in the machine and see if it affects anything. Maybe the mixed set , or one of the sets is bad or just conflicting somehow.

I've tried that and at best it doesn't make any difference. I've noticed that with 3 or 1 module installed the performance drops a little bit, like 3 to 5 FPS less. Also, in these cases, POST says Single Channel

emosun wrote:

Also what's the cpu usage like?

It's between 1% and 5%, and when idle it goes down to 0%

Also, while in Windows 7 it gets lower fps overall, it repeats the same pattern as in XP. Very strange.

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 6 of 27, by weldum

User metadata
Rank Member
Rank
Member

The thing is thah i don't have money and usually try to fix the shortcommings with what i have

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 7 of 27, by ODwilly

User metadata
Rank l33t
Rank
l33t

+1 on maxing out the ram and cpu. Pentium D's are cheap as chips and DDR is worthless right now. I know in CS:GO at least you should get decent framerates with that HD6450 and a Pentium D of any flavor.

Main pc: Asus ROG 17. R9 5900HX, RTX 3070m, 16gb ddr4 3200, 1tb NVME.
Retro PC: Soyo P4S Dragon, 3gb ddr 266, 120gb Maxtor, Geforce Fx 5950 Ultra, SB Live! 5.1

Reply 8 of 27, by weldum

User metadata
Rank Member
Rank
Member

well, my economic situation doesn't allow me to get more RAM and a better processor, so an upgrade cannot be purchased.

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 10 of 27, by weldum

User metadata
Rank Member
Rank
Member
agent_x007 wrote:

Don't buy beer this weekend, and between 2 weeks to a month, you will have a Pentium D 945 at your mail box : LINK

I wish it were as easy as that, i'm currently unemployed, and got my car fixed, also don't have credit card or bank account (i'm 21 years old) so never bought in ebay.

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 11 of 27, by emosun

User metadata
Rank Oldbie
Rank
Oldbie
weldum wrote:

I wish it were as easy as that, i'm currently unemployed, and got my car fixed, also don't have credit card or bank account (i'm 21 years old) so never bought in ebay.

I can sympathize , I too understand the "just make what i already have work" situation.

weldum wrote:

the thing is with latter games like Unreal 3 or CSGO, strangely Left4Dead 2 works almost perfect on the Radeon and the 9600GSO.

Ok , so what exact games are not working with the newer gpu's.

don't bother testing with the fx5200 anymore just try to get one of the newer gpu's working , ideally the 9600 should be the best match.

weldum wrote:

It's between 1% and 5%, and when idle it goes down to 0%

I meant while gaming , not while idle.

put the 9600 in. test all your games and find which one run poorly , then see what their cpu usage is like in those games.

weldum wrote:

I have 3 HDD, 160GB each

test each drive and find out which is the fastest with some testing software like crystal disk. then , only use that drive from now on.

we have to eliminate variables so testing with one drive and one gpu would help you a lot.

Reply 12 of 27, by Robin4

User metadata
Rank l33t
Rank
l33t

I would avoid such AGP / PCI-e motherboards.. You dont know how the PCI-e 1.0 slot is electronically connected to the motherboard.. The first irritations of such combo boards are mostly bottlenecked on the pci-e bus motherboards.

Better go for N-force 4 boards or later.. Those have 100% faster PCi-e connections

~ At least it can do black and white~

Reply 13 of 27, by agent_x007

User metadata
Rank Oldbie
Rank
Oldbie
Robin4 wrote:

I would avoid such AGP / PCI-e motherboards.. You dont know how the PCI-e 1.0 slot is electronically connected to the motherboard.. The first irritations of such combo boards are mostly bottlenecked on the pci-e bus motherboards.

Better go for N-force 4 boards or later.. Those have 100% faster PCi-e connections

Don't worry. I tested similar board (PT880 Pro), with up to GTX 580 and HD 5870.
Anything slower than GTS 450/GTX 460 or Radeon 5750/5770 isn't going to be b. necked by PCI-e speed in any meaningfull way.
HD 5870 @ PCi-e 1.0 x4

3DMark Vantage X-Score mini.png
Filename
3DMark Vantage X-Score mini.png
File size
864.33 KiB
Views
1063 views
File license
Fair use/fair dealing exception

HD 5870 @ PCI-e 2.0 x16

Radeon 5870 (test 3) 3DMV Extreme mini.png
Filename
Radeon 5870 (test 3) 3DMV Extreme mini.png
File size
1.1 MiB
Views
1063 views
File license
Fair use/fair dealing exception

If Pentium D isn't possible, I would check if Pentium 4 isn't overheating (stock cooling has easy to break push-pins and if they get damaged CPU may throttle).
Also, DO NOT use Low Profile Intel BOX cooler on this CPU.

157143230295.png

Reply 14 of 27, by weldum

User metadata
Rank Member
Rank
Member
emosun wrote:

Ok , so what exact games are not working with the newer gpu's.

don't bother testing with the fx5200 anymore just try to get one of the newer gpu's working , ideally the 9600 should be the best match.

CSGO (very low fps), Need For Speed Hot Pursuit 2010 (very low fps), I also have to test Unreal tournament 3. Note that I don't want to run these games at highest res/graphical settings, I just want to run them in lowest res and settings at, ideally, 60 fps. Is for Lan Party.

emosun wrote:

I meant while gaming , not while idle.

put the 9600 in. test all your games and find which one run poorly , then see what their cpu usage is like in those games.

Later I will test that, now i'm using that card another machine, sadly my computers are all assembled wrong, I'm thinking in showcasing all of my machines and projects later in the forum.

emosun wrote:

test each drive and find out which is the fastest with some testing software like crystal disk. then , only use that drive from now on.

we have to eliminate variables so testing with one drive and one gpu would help you a lot.

I will do that then, i've got all 3 drives to be IDE, so there are no crappy Sata1 controllers with compatibility issues

Robin4 wrote:

I would avoid such AGP / PCI-e motherboards.. You dont know how the PCI-e 1.0 slot is electronically connected to the motherboard.. The first irritations of such combo boards are mostly bottlenecked on the pci-e bus motherboards.

Better go for N-force 4 boards or later.. Those have 100% faster PCi-e connections

The thing is, when I had money to spend, I ended with 4 working computers, while I was triyng to upgrade just one. Don't want to buy because don't have the money and because of don't be able to get rid of that I don't need/use.

agent_x007 wrote:
Don't worry. I tested similar board (PT880 Pro), with up to GTX 580 and HD 5870. Anything slower than GTS 450/GTX 460 or Radeon […]
Show full quote

Don't worry. I tested similar board (PT880 Pro), with up to GTX 580 and HD 5870.
Anything slower than GTS 450/GTX 460 or Radeon 5750/5770 isn't going to be b. necked by PCI-e speed in any meaningfull way.
HD 5870 @ PCi-e 1.0 x4
If Pentium D isn't possible, I would check if Pentium 4 isn't overheating (stock cooling has easy to break push-pins and if they get damaged CPU may throttle).
Also, DO NOT use Low Profile Intel BOX cooler on this CPU.

What makes me wonder why the 2 cards have almost the same performance in this machine (the GF 9600GSO and the R6450)
The cooler is from intel, it was in a E2140 but the heatsink is taller than most other and it has copper core, the temperature is always below 40 degrees celsius, even with Aida64 stability test trottling the processor

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 15 of 27, by emosun

User metadata
Rank Oldbie
Rank
Oldbie
weldum wrote:

CSGO (very low fps), Need For Speed Hot Pursuit 2010 (very low fps)

That machine doesn't meet the minimum cpu requirements for either game , even a pentium D would struggle with both those games. You need to be testing games that the machine can actually play. You're about 5 years too far into the future with those games really.

There was a monumental leap in speed from the pentium 4 to early core 2 cpu's. Meaning even some of the worse core two cpu's can be 3 times the speed of a pentium 4.

I honestly dont even need to to post your cpu usage during those games to tell you that that is what the problem is.

Reply 16 of 27, by weldum

User metadata
Rank Member
Rank
Member
emosun wrote:
That machine doesn't meet the minimum cpu requirements for either game , even a pentium D would struggle with both those games. […]
Show full quote
weldum wrote:

CSGO (very low fps), Need For Speed Hot Pursuit 2010 (very low fps)

That machine doesn't meet the minimum cpu requirements for either game , even a pentium D would struggle with both those games. You need to be testing games that the machine can actually play. You're about 5 years too far into the future with those games really.

There was a monumental leap in speed from the pentium 4 to early core 2 cpu's. Meaning even some of the worse core two cpu's can be 3 times the speed of a pentium 4.

I honestly dont even need to to post your cpu usage during those games to tell you that that is what the problem is.

I know but, why does Left 4 Dead 2 runs at 720p with everything set to max at around 30 to 60 fps but CSGO runs at 10fps? they have almost the same requirements

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 17 of 27, by emosun

User metadata
Rank Oldbie
Rank
Oldbie
weldum wrote:

I know but, why does Left 4 Dead 2 runs at 720p with everything set to max at around 30 to 60 fps but CSGO runs at 10fps? they have almost the same requirements

Your machine meets the minimum requirements for l4d2 , it does not for csgo

idk where you are reading system requirements from but they clearly are for the steam versions or are not accurate at all.

Reply 18 of 27, by weldum

User metadata
Rank Member
Rank
Member
emosun wrote:
weldum wrote:

I know but, why does Left 4 Dead 2 runs at 720p with everything set to max at around 30 to 60 fps but CSGO runs at 10fps? they have almost the same requirements

Your machine meets the minimum requirements for l4d2 , it does not for csgo

idk where you are reading system requirements from but they clearly are for the steam versions or are not accurate at all.

I see, then the only way is to upgrade to a better processor, the best i could find here is Pentium D 915 which is dual core but with a little bit lower clock.
We'll see. Maybe after christmas

DT: R7-5800X3D/R5-3600/R3-1200/P-G5400/FX-6100/i3-3225/P-8400/D-900/K6-2_550
LT: C-N2840/A64-TK57/N2600/N455/N270/C-ULV353/PM-1.7/P4-2.6/P133
TC: Esther-1000/Esther-400/Vortex86-366
Others: Drean C64c/Czerweny Spectrum 48k/Talent MSX DPC200/M512K/MP475

Reply 19 of 27, by emosun

User metadata
Rank Oldbie
Rank
Oldbie
weldum wrote:

I see, then the only way is to upgrade to a better processor, the best i could find here is Pentium D 915 which is dual core but with a little bit lower clock.
We'll see. Maybe after christmas

The pentium D 915 also doesn't meet the minimum requirement for the game. It's 3 TIMES slower than the minimum cpu which is stated as an C2D E6600.