FX 5200 128MB (non-Ultra) benchmarks?

[selfbuilt]

Has anyone with one of these cards (FX 5200 non-Ultra) run any game/benchmarks yet?

The cards seem to be starting to become available, but I haven't seen much yet directly comparing them to the Ultra reviews (or to other cards for that matter). One detailed review available here (in French) shows that the 128MB non-Ultra underperforming the Ultra by about 25% on most tests (or put another way, the Ultra is about 33% faster than the non-Ultra). That seems reasonable given the price difference.

Also, any info on specific makers yet? I saw a very self-serving commentary at eVGA about variable quality of RAM chips on these cards. Although I take their "analysis" with a huge grain of salt, it does raise the spectre of the old GeForce4MX memory issues (for example, I've seen GeF4 MX440 64MB cards rated at anywhere from 1.3 to 6.3 GB/sec bandwith, with as little as $5-10 to differentiate them!). I hope the FX 5200 isn't plagued by the same problem (but I suppose it will).

Any good brands standing out so far?

 

[Marshal1stContinental]

got it few minutes ago, here it is, all default

2083

fillrate single: 450
fillrate multi: 691

 

[nemesismk2]

Here's some benchmark scores for my GF4 MX440 (270/400) and Geforce FX5200 (250/400) tested with:-

AMD Athlon XP 2000
SIS745 motherboard
512mb DDR266 memory
Detonator v43.45
SIS AGP v1.15
Windows XP Home edition with SP1

All games tested with max quality settings and with sound enabled for real game results!

(results are=mx440, fx5200)

Quake 3 (demo4):-

800x600x32bit=175,184
1024x768x32bit=140,158
1280x1024x32bit=119,137

Quake 3 2xfsaa:-

640x480x32bit=168,182
800x600x32bit=130,160
1024x768x32bit=86,116

Quake 3 4xfsaa:-

640x480x32bit=104,155
800x600x32bit=73,117
1024x768x32bit=47,74

Serious Sam 2 (Elephant Atrium):-

800x600x32bit=133,134
1024x768x32bit=106,113
1280x1024x32bit=88,99

NOTE:The FX5200's performance will improve when the next SS2 Patch is released which fully supports the Geforce FX.

UT2003 flyby (Antalus):-

800x600x32bit=86,92
1024x768x32bit=58,63
1280x1024x32bit=39,43

 

[selfbuilt]

Thanks for the rapid updates guys ... good to see it consitently outperform the MX440, but I guess I was hoping for a little more on the newer games. As you say, though, newer patches and drivers should help there.

Incidentally, just as way of a comparison, I remember Mloot's PCI MX440 vs Radeon 9100 comparison here showed a pretty significant improvement of the 9100 over the MX440 (both PCI, though). Anyway, I'm still hopeful the FX5200 will turn out to be a decent entry level option.

 

[Mloot]

Hey, I have an update to my PCI comparison that I am currrently working on. I'm doing another comparison w/440 PCI, 9100 PCI, and FX 5200 PCI. I should have it posted here later tonight, or tomorrow morning.

 

[nemesismk2]

Originally posted by: selfbuilt
Thanks for the rapid updates guys ... good to see it consitently outperform the MX440, but I guess I was hoping for a little more on the newer games. As you say, though, newer patches and drivers should help there.

Incidentally, just as way of a comparison, I remember Mloot's PCI MX440 vs Radeon 9100 comparison here showed a pretty significant improvement of the 9100 over the MX440 (both PCI, though). Anyway, I'm still hopeful the FX5200 will turn out to be a decent entry level option.

Remember all of my results are with sound enabled (it's also onboard sound as well), most review sites have results with sound disabled. So my results are what you will get on a PC with sound enabled, loads of background stuff going on etc.

 

[selfbuilt]

Here's a link to Mloot's comparison of MX440, FX5200, 9100, just for completeness.

The PNY FX5200 sample seriously underperformed the 9100 in his tests, often scoring not much better than the MX440. The fact that is was using slower RAM probably has a lot to do with it. Looks like the eVGA comments I linked to at the start of this thread have some validity ... anyone thinking of getting a FX5200 had better carefully scrutinize the memory (just like anyone getting a MX440 has always had to do). Nemesismk2's sample clearly outperformed his MX440, as I would have expected ... obviously a better brand (although it could also be PCI FX5200 cards are crippled in general).

Judging from the one detailed review I found (the Hardware.fr link at the top of the thread), a good sample of FX5200 should just about hold its own overall with the Radeon 9000Pro (and upcoming 9200Pro, which will be the same thing). There, the 9000 Pro outperformed the FX5200 considerably in Quake3, Serious Sam, and Aquanox, but not by much in UT2K3 at stock settings. But with AA/AF turned on its a different story, with the FX5200 considerably outperforming in Quake3 and UT2K3. So for that comparison, I guess it depends on whether or not you plan to play with AA/AF enabled. If yes, get a *good* FX5200, if no, stick with the 9000 Pro.

The comparison to the 9100 may not be entirely fair, since it is of course a renamed 8500LE and is faster than the 9000Pro most of the time. But as this card is being phased out, and the 9000/9200 Pro is all people will have in the future for comparison.

Still, for the time being, I can get a 9100 for the same price as a 9000Pro around here, and it looks like that will be my best best at stock speeds. Still tempted to get the FX5200 to play with and O/C though ... anyone have what they figure is a good brand?

 

[DPGX]

The thing that impressed me about the PNY is that its equipped with 3.6ns Hynix memory... I have the AGP Model so that may have something to do with it as well, but technically the Ram is rated to 550mhz which I have yet to try but is kind of a nice feature.

 

[nemesismk2]

<DELETED>

 

[nemesismk2]

Originally posted by: selfbuilt
Here's a link to Mloot's comparison of MX440, FX5200, 9100, just for completeness.

The PNY FX5200 sample seriously underperformed the 9100 in his tests, often scoring not much better than the MX440. The fact that is was using slower RAM probably has a lot to do with it. Looks like the eVGA comments I linked to at the start of this thread have some validity ... anyone thinking of getting a FX5200 had better carefully scrutinize the memory (just like anyone getting a MX440 has always had to do). Nemesismk2's sample clearly outperformed his MX440, as I would have expected ... obviously a better brand (although it could also be PCI FX5200 cards are crippled in general).

Judging from the one detailed review I found (the Hardware.fr link at the top of the thread), a good sample of FX5200 should just about hold its own overall with the Radeon 9000Pro (and upcoming 9200Pro, which will be the same thing). There, the 9000 Pro outperformed the FX5200 considerably in Quake3, Serious Sam, and Aquanox, but not by much in UT2K3 at stock settings. But with AA/AF turned on its a different story, with the FX5200 considerably outperforming in Quake3 and UT2K3. So for that comparison, I guess it depends on whether or not you plan to play with AA/AF enabled. If yes, get a *good* FX5200, if no, stick with the 9000 Pro.

The comparison to the 9100 may not be entirely fair, since it is of course a renamed 8500LE and is faster than the 9000Pro most of the time. But as this card is being phased out, and the 9000/9200 Pro is all people will have in the future for comparison.

Still, for the time being, I can get a 9100 for the same price as a 9000Pro around here, and it looks like that will be my best best at stock speeds. Still tempted to get the FX5200 to play with and O/C though ... anyone have what they figure is a good brand?

I disagree with your comments about the radeon 9000 pro, I have a Sapphire Radeon 9000 Pro and my XFX Geforce FX 5200 outperforms it most of the time (usually not by a huge amount). However your right about the poor performance of the Radeon 9000 Pro with fsaa enabled and the ut2003 performance isn't so great either.

The Radeon 9000 Pro isn't too bad but it's not as good as a FX5200, I liked the VIVO features of the version I had though.

 

[nemesismk2]

Here's some benchmark scores for my Sapphire Radeon 9000 Pro (275/550) and XFX Geforce FX5200 (250/400) tested with:-

AMD Athlon XP 2000
SIS745 motherboard
512mb DDR266 memory
Detonator v43.45
SIS AGP v1.15
Windows XP Home edition with SP1

All games tested with max quality settings and with sound enabled for real game results!

(results are=r9000pro, fx5200)

3dmark 2001se (1024x768x32bit)=6872,6679
3dmark03 (1024x768x32bit)=985,1390

Quake 3 (demo4):-

800x600x32bit=174,184
1024x768x32bit=155,158
1280x1024x32bit=134,137

Quake 3 2xfsaa:-

640x480x32bit=160,182
800x600x32bit=118,160
1024x768x32bit=75,116

Quake 3 4xfsaa:-

640x480x32bit=102,155
800x600x32bit=67,117
1024x768x32bit=44,74

Serious Sam 2 (Elephant Atrium):-

800x600x32bit=122,134
1024x768x32bit=116,113
1280x1024x32bit=110,99

NOTE:The FX5200's performance will improve when the next SS2 Patch is released which fully supports the Geforce FX.

UT2003 flyby (Antalus):-

800x600x32bit=86,92
1024x768x32bit=57,63
1280x1024x32bit=37,43

I'll be updating all of my results when I've finished upgrading my testing computer, it's having a SIS746FX motherboard, 512mb DDR333 memory and a Athlon XP 2500 (Barton)cpu with 333mhz fsb.

 

[selfbuilt]

Hi Nemesismk2 ... I was simply commenting on the results of the Hardware.fr analysis, which seemed to give the 9000 pro a clear lead on non-AA/AF modes on most of the games they tested (Quake3, SS, Aquanox, but not UT2K3).

But from your scores, it seems the XFX FX5200 holds its own pretty well under those conditions on Serious Sam and Quake3 ... which is good to hear. And of course, there's no argument that the AA/AF implementation is faster on the FX5200!

Although I haven't seen a FX5200 in my area yet, XFX is a popular maker around here so I'll keep my eyes open for that board. Maybe I'll see about doing a few comparisons myself if I can find one

 

[nemesismk2]

*** The FX5200 seems to run pretty good with a slower cpu, even though the cpu is
running at only 800mhz all of the results are playable! ***

Here's some benchmark scores for a XFX Geforce FX5200 (250/400) tested with:-

AMD Athlon XP 2000 1.67ghz (12.5x133mhz) & underclocked to 800mhz (6x133mhz)
SIS745 motherboard
512mb DDR266 memory
Detonator v43.45
SIS AGP v1.15
Windows XP Home edition with SP1

All games tested with max quality settings and with sound enabled for real game results!

(results are=800mhz & 1.67ghz)

3dmark 2001se (1024x768x32bit)=5270,6679

Quake 3 (demo4):-

800x600x32bit=129,184
1024x768x32bit=125,158
1280x1024x32bit=118,137

Quake 3 2xfsaa:-

640x480x32bit=128,182
800x600x32bit=125,160
1024x768x32bit=106,116

Quake 3 4xfsaa:-

640x480x32bit=123,155
800x600x32bit=105,117
1024x768x32bit=74,74

Serious Sam 2 (Elephant Atrium):-

800x600x32bit=91,134
1024x768x32bit=83,113
1280x1024x32bit=79,99

NOTE:The FX5200's performance will improve when the next SS2 Patch is released which fully supports the Geforce FX.

UT2003 flyby (Antalus):-

800x600x32bit=71,92
1024x768x32bit=60,63
1280x1024x32bit=43,43

*** This will be the last of the results because my PC will be getting upgraded today! ***

 

[selfbuilt]

I'm not surprised the FX5200 performs well on those tests on an underclocked processor, since they are all somewhat relatively processor independent. I don't know how that relates to older processors and systems, but thanks for the update ... the more data the better.

Unfortunately, I haven't been able to find a good deal on FX5200 locally yet (I live in Canada, where most US online sites like newegg won't ship). The best I've seen is about $115 US for a no-name brand, which seems a little steep since I can get a Sapphire Radeon 9100 128MB for only $85 US locally. But I'm keeping my eyes open for the XFX ...

 

[selfbuilt]

Just an update ... in the end, I couldn't resist going for a $65US Radeon 8500LE 64MB (retail, built by ATI, so should have decent clock speeds) as the replacement for my son's computer. Until the FX5200 come down considerably in price to something closer to the 9100s, I can't justify it.

 

[nemesismk2]

BTW the max I could overclock my XFX FX5200 (without replacing the heatsink with a fan) is 300mhz core and 500mhz memory.

 

[nemesismk2]

Here are some NEW benchmark scores for my GF4 MX440 (270/400) and Geforce FX5200 (250/400):-

AMD Athlon XP 2500 Barton (11x166mhz)
SIS746FX motherboard
512mb DDR333 memory
Detonator v44.03
SIS AGP v1.15
Windows XP Home edition with SP1

All games tested at 1024x768x32bit with max quality settings and with sound enabled for real game results!

(results are fx5200 & mx440)

3dmark 2001 se (1024x768) = 7656,6367
3dmark 2003 (1024x768) = 1423,261

Quake 3 (demo4):-

800x600x32bit = 212,198
1024x768x32bit = 159,145
1280x1024x32bit= 132,121

Quake 3 2xfsaa:-

640x480x32bit = 212,184
800x600x32bit = 169,132
1024x768x32bit= 112,86

Quake 3 4xfsaa:-

640x480x32bit = 135,96
800x600x32bit = 90,66
1024x768x32bit= 55,43

Quake 3 anisotropic filtering at 1024x768x32bit:-

2x=138,125
4x=126,N/A
8x=117,N/A

Return to castle wolfenstein (IXBT):-

800x600x32bit = 125,158
1024x768x32bit = 82,121
1280x1024x32bit= 66,101

NOTE:Yes these scores are correct, the FX5200 gets it's ass kicked by the little GF4 MX440 big time. Hey Nvidia I think you forgot to obtimise this game for the Geforce FX?

Serious Sam 2 (Elephant Atrium):-

800x600x32bit = 160,152
1024x768x32bit = 120,108
1280x1024x32bit= 99,87

NOTE:The FX5200 stutters when the benchmark starts which results in a lower score, Nvidia should of fixed this problem in the last driver release.

UT2003 flyby LQ (Antalus):-

800x600x32bit = 124,124
1024x768x32bit = 83,84
1280x1024x32bit= 55,56

UT2003 botmatch LQ (Antalus):-

800x600x32bit = 74,76
1024x768x32bit = 61,64
1280x1024x32bit= 43,46

NOTE:Both of these video cards should be used in a slower system because they are both really holding back the Athlon XP 2500. The GF4 Ti4200 which will be replacing the FX5200 in this system will deliver much better performance!