Radeon 8500 vs. Radeon 8500LE - difference?

[eldren]

I'm hoping to get a Radeon 8500(LE?) soon, mostly just because it's cheap. However, apart from the different core and RAM frequencies (275 for the 8500, 250 for the 8500LE), I can find no differences between the two cards.

So, the question is, which do I want - the 8500 or the 8500LE?
And what's the difference between LE and non-LE chipsets?

Update: The guy sold the card to a friend. Question still stands, though, for interest sake.

[CrapshoT]

good question i would like to know the answer of this as well as i have the 8500 LE its speeds are 249/249

[SilentDeath]

afaik, the r8500 series all has same core, including the 9000-9200.
with the 8500 though, companys started selling them with lots of differnt speeds, crucail being the worst, sticking it at 200/166.. i think my voodoo5500 was clocked higher!

the 9000-9200's were bought out because they were altered to make them slightly cheaper to make, so they also have lower stock speeds than the 8500 (iirc). im unsure if they will clock the same....

all 8500 cores should have about the same maxuimum clock if u try to o/c them.. the mem speed depends on the mem chips u get on yours...

[Bindibadgi]

8500 built by ati's are 275/275 3.6ns ram
8500 le's (3rd party manuf.) 250/250, 4ns ram or early models had 3.6ns ram and reference design which you could easily flash with a normal 8500 bios.
8500 le le - 250/230 4/4.5ns ram iirc

[CrapshoT]

sorry for hogging the thread but how do you tell which ram is on my card and how would i flash the bios with a normal bios?

[streetster]

look at the ram chips on it.. there should be like 40 at the end of the numbers/letters on it, which means 4.0ns, (well in most cases), ie 36 would mean 3.6ns...

go check

mark

[Austin]

As pretty much covered the 8500LE is simply a slower clocked 8500. Clock speeds should be 8500=275/275 (that's 275/550 when you factor in DDR) and 8500LE = 250/250. However there was a lot of variance, some manu's neglected other things too such as Dual RAMDACs for dual display. Generally OEM were clocked slower than the retail versions. As a rough guide most of the cards would get the core at least 25mhz higher, up close to 300mhz was not that uncommon. The RAM was very different though, most cards which used slower clocked RAM also used cheaper RAM and as such you'd get a poor o/c. So basically if you buy an 8500 don't worry if it's an LE but do worry about the clock speeds it uses.

The Rad9000 and 9200 are way slower than the 8500LE, the 8500 technology was castrated to make it more profitable per unit (and lower selling point too). The PRO versions used higher clocks and were only about 10% slower than most 8500LE. The 9100 was supposed to be a return to the superior 8500 but all it used was the technology, clocks were often far lower than most 8500LE let alone 8500 and worst of all they varied across the board once more (9000-9200 had been pretty std). There is something else to worry abot other than pure clock speed. Although it was very uncommon, Powercolor (known to be sub par from time to time) used 64bitDDR on their 9000PRO, that's like using std SDR and makes the actual effective clock speeds 275/138 instead of the 275/275 that almost every other 9000PRO used.

To put perf in perspective a Rad8500 @ 275/275 is pretty close to stock GF4TI4200 perf levels, around that of a GF3TI500 and faster than GF-FX5200ultra or FX5600. The Rad9000-9200 are slower and inferior to FX5200, not good at all. Rad9000/9200PRO and Rad9100 (decently clocked) are roughly inbetween so around FX5200 and GF3TI200 speed. Since 4200 are immensely better than the current crop of entry level cards and can be had for £75 it seems pointless killing perf by saving the £20ish.

[Lexeus]

are you saying all Radeon 9000Pro Powercolr cards use only 64bit ?
As i have a 275/275 Retail 64mb Radeon 9000Pro Powercolor Evil Commando Card. I bought it the week they came out. It doesn't say anything about 64bit or 128bit on the box, so are you saying it definitely only has 64bit DDR ?

[eldren]

Once again, Austin, your reply amazes me. You should go write reviews for some computing magazine or something, your replies are so thorough. Told me just what I wanted to know, thanks.

[Austin]

Originally posted by eldren
Once again, Austin, your reply amazes me. You should go write reviews for some computing magazine or something, your replies are so thorough. Told me just what I wanted to know, thanks.

Wow ... thanks. Well I am learning all the time just like everyone else but if anybody out there does want to pay me I'll learn much faster! Anyone?

Lexeus it's not 100% certain that ALL Powercolor 9000PRO's use 64bitDDR but it is certain that some of them do. Whether there's a diff between OEM and retail or 128MB vs 64MB I simply don't know. I'm also not saying ALL Powercolor cards are bad, many people have had great experiences with them but I think they are most likely just lucky (I'm afraid). Anyway you should benchmark and you'll soon see whether you're turning in perf comparative to other 9000PRO (around 9100 and 8500LE speeds) or something far below (I expect at least 30% slower). Do rem that most 3Dmark scores are derived from o/c'ed systems and cards so bear that in mind, post your spec and I'll find out a stock system if I can. The reason they used 64bitDDR is clear ... they can still shout out DDR and also fast clock speeds (the thing most commonly checked for) ... it also helps them to offer a cheaper price AND yet make more profit per card.

[Lexeus]

Well my system specs are all in my sig.
I ran 3dmark2003 on it a few weeks ago.
Can you see this page ?
http://service.futuremark.com/compare?2k3=1224393

If not, I got:

3DMark Score 1304 3DMarks

CPU Score 437.0 CPUMarks

that aint great is it ? I think they are reliable, if I remember correctly.

what you think ?

[streetster]

unless you have a dx9 gfx card, you are going to get complete crappy results on 3dmark2003.. go run 2001se as you have a dx8 card much more satifying result.

mark

[Austin]

Yes, 2001SE is a much better benchmark, certainly for the DX8 cards. From what I can tell from the 2003 things seem about right. A quick search (same speed CPU & similar gfx) reveals ...

9100 @ 250/200 (similar to 9000 @ 275/275):
TOTAL=1166
FPS=63.5, 8.3, 8.4, n/a.
CPU=34.8, 5.3

9000PRO @ 275/275:
TOTAL=1367
FPS=75, 9.5, 10, n/a
CPU=34.4, 5.4

YOURS (you have Barton and 333FSB though):
TOTAL=1304
FPS=79, 8, 9, n/a
CPU=53, 7

[Lexeus]

So you think its not running any slower than normal ?
I would run 3dmrak2001 but I need dx9 installed and it only accepts dx8
Thanks Austin

[Austin]

3Dmark2001SE only needs DX8 installed and preferably with a DX8 gfx card (like Rad9000PRO). It's 3Dmark03 (2003) that reqs DX9 and preferably a DX9 gfx card. For reference in 3Dmark2001SE an FX5200 (stock 250-400) with XP2500+ scores 5600 while an XP2000+ with GF4TI4200 (250-444) scores 10000.

[5cupa]

can the radeon 8500 still cut it with todays games with its 64mb? as second hand they are nice and cheap