64MB Radeon 8500LE vs. 128MB

[MikeD2k3]

Is double the memory worth an extra $22? How much memory do you really need? Do textures take up that much space?

Thanks

 

[BFG10K]

Most definitely. Many of today's games absolutely kill 64 MB cards.

 

[railer]

Please name 1 game that kills a 64 mb card. My new GF4 4200 64 mb kills my old 8500 128 megger, in every way, shape, and form, in every game I've compared the two. If you plan on keeping your vid card for the next 3 years, get the 128 megger. If you're like a normal person around here, or if you play at extremely high resolutions, save your $22.

 

[BFG10K]

Please name 1 game that kills a 64 mb card.

Almost every single FPS made in the last ~2 years - RTCW, JK2, SOF2 etc.

My new GF4 4200 64 mb kills my old 8500 128 megger, in every way, shape, and form, in every game I've compared the two.

In high resolution situations combined with anisoptropic filtering the Ti4200 is totally inferior in performance to the 8500.

 

[LS20]

disagree... ive gotten by fine on my 32mb gf2 gts for a long while and havent wished for any more memory on my 64mb 8500le... this thing hasnt met something it coulnt play well

 

[SFang]

In high resolution situations combined with anisoptropic filtering the Ti4200 is totally inferior in performance to the 8500.

Has BFG10K made it clear enough for you guys?

 

[railer]

He hasn't made it clear that he knows anything...but he has made it clear as to what his opinion is. I'm playing RTCW right now....so I suppose I could benchmark that between my radeon and gf4. Not sure if I'm that bored just to prove a point. If I ever inherit a video card company the first thing I'm going to do is start slapping on the memory and jacking up prices. I'll sell a voodoo 2 based card with 2 gigs of 30 pin 70 ns circa 1992 memory for a cool $600. And people will buy it too 'cause "it has more memory so it must be faster even if every benchmark in the history of humanity says that it isn't".
While I'm selling those cards for $600, I'll be gaming on my "totally inferior" $85 GF4 that's 20 times faster but doesn't have the right amount of memory. That's what I'll do....
BTW, BFG is saying that 1st person shooters kill 64 meg cards, not 1st person shooters run under high resoluitons with ansiotropic filtering on. I guess the RTCW that I've been flying through the past few weeks has been a mirage. I've played this on my spare rig with a 32 meg radeon without a hitch.

 

[BFG10K]

I'm playing RTCW right now....so I suppose I could benchmark that between my radeon and gf4.

And you'd do well to do so because you might be surprised at just what you find.

But just in case you're too lazy, here is one of the many people who have already done it. Even at the low reoslution of 1024 x 768 the Radeon is already pulling away from the Ti4200.

Main page

Also on the same graph you can see just how much faster the 128 MB Radeon is over the 64 MB Radeon, a trend that continues in most of today's games. In fact, even the slower clocked 8500 LE is often faster than the higher clocked 64 MB 8500, and the same thing applies to the 128 MB vs 64 MB Ti4200 cards.

BTW, BFG is saying that 1st person shooters kill 64 meg cards, not 1st person shooters run under high resoluitons with ansiotropic filtering on.

Well sure, if you want to play at butt ugly settings then by all means, get the 64 MB card. Also benchmarks don't always tell the whole story as 64 MB cards experience texture swaps frequently, even at medium resolutions. A 64 MB card frequently texture swaps in RTCW, even at a medium resolution of just 1152 x 864 x 32 with no anisotropic filtering, and even with texture compression enabled.

I guess the RTCW that I've been flying through the past few weeks has been a mirage.

Most probably. You'd be surprised just how many people tell me something is "smooth and doesn't have any problems" only to change their story after they run benchmarks are are shown quantifiable evidence to the contrary.

 

[jiffylube1024]

Originally posted by: BFG10K

Well sure, if you want to play at butt ugly settings then by all means, get the 64 MB cards. Also benchmarks don't always tell the whole story as 64 MB cards experience texture swaps frequently, even at medium resolutions. A 64 MB card frequenclty texture swaps in RTCW, even at a medium resolution of just 1152 x 864 x 32 with no anisotropic filtering, and even with texture compression enabled.

This is absolutely true about the texture swapping. I am playing NOLF2 on an XP 2100+ with my "backup" GF3 64MB right now (until my Ti4200 128MB arrives in a week), and am getting subtle, but still annoying texture swaps, even at 1024X768 and even somewhat at 800X600. This wasn't near to as bad as on my Radeon 8500 128MB.

As for the issue at hand, the argument is so subjective it's not even funny. Here's the facts, though. A GF4 64MB is faster than an 8500 128MB as long as you run at 800X600 or even 1024 in many/most games. However, once you turn on Anisotropic Filtering and/or AA, the 128MB card starts to pull away, as well as set the resolution to 1280 or 1600.

BFG10K is obviously a detail/eye candy whore, since he is describing situations that are basically unfathomable to me. Since he has a Radeon 9700, he can have his cake and eat it too: he can put on AF and also AA and still run at high resolutions with good framerates.

Since I myself am always a generation behind, and I'm such a stickler for a SMOOTH framerate (probably like you, MikeD2K3), I like to set it on 1024X768 for all games, or even, gasp 800X600 with this GF3. No AA or AF for me, any loss in FPS is bad.

At these settings (1024 or 800), the GF4 is indeed better than the 8500 128MB. At higher res, or with AA/AF, the 8500 starts to pull away. So, essentially, both of you are right, based on how you like to play.

In high resolution situations combined with anisoptropic filtering the Ti4200 is totally inferior in performance to the 8500.

^BFG loves baiting people by making very specific claims like this above, and then tries to pin you in a corner with your own words. "How can my 8500 be inferior to my GF4? It's not possible" you say, then get into an argument (which he loves, apparently). But, his original claim was so specific, that he IS indeed right, and will catch you in the end.

Here's the point: the GF4 is right for you, as it would be for me, since at 1024 or especially 800X600 with no AA or AF, the GF4 is as good as, if not better than the 8500 128MB. For you, the GF4 is faster. For someone who runs everything at 1600X1200, the 8500 is basically faster in all situations.

Again, don't bother arguing with BFG's favourite line "In high resolution situations combined with anisoptropic filtering the Ti4200 is totally inferior in performance to the 8500. " Just say, at realistic settings, with no AA or AF, the 8500 is totally inferior to the Ti4200 . Cheers!

 

[rogue1979]

I have these video cards in the house right now.

Radeon 9000 64MB
Geforce3 Ti200 64MB
Radeon 8500 64MB
Radeon 8500 128MB
Geforce4 Ti4200 128MB
Radeon 9000 Pro 128MB

Like BFG10K, I prefer the Radeon 8500 to the Geforce4 Ti4200 when playing 1600 x 1200 x 32 no FSAA 16-tap anisotropic. However, I notice very little difference in gaming framerates between 64MB and 128MB versions of the same video cards. I think the best buy out there now for modern gaming is a 64MB version Radeon 8500 LE that will run 300/300. You have to spend a whole lot more to get a faster card. I plan on selling off all these cards and keeping three Radeon 8500 64MB cards for myself and both my childrens' rigs. I just don't get any benefit from my 128MB 8500 over the 64MB. As a matter of fact, I have been testing different bios versions of the 8500's and found the original 1009 version overclocks the best. My current 128MB Radeon came with the 7006 bios and ran 312/301, but with the 1009 bios it goes 312/315. I can't use the 1009 bios for the 128MB version because it will only recognize 64MB of memory. So the 64MB cards with the 1009 bios actually run slightly faster because I can overclock the memory 10-15MHz higher.

 

[DieHardware]

The videocard benchmarks BFG10K linked to use;

The NVIDIA driver revision is the latest 28.32 version available through though their website, as is the 7.68/6043 driver release used with the Radeon 8500 cards.

I would like to see the results of those same benchmarks/cards using the latest Detonator 40.XXs and Catalyst 3s. Nvidia's reference drivers were already optimized(28.32s are close to 40.XXs in performance), haven't the ATI drivers improved quite a bit since April 22?

 

[CraigRT]

Originally posted by: BFG10K

Please name 1 game that kills a 64 mb card.

Almost every single FPS made in the last ~2 years - RTCW, JK2, SOF2 etc.

My new GF4 4200 64 mb kills my old 8500 128 megger, in every way, shape, and form, in every game I've compared the two.

In high resolution situations combined with anisoptropic filtering the Ti4200 is totally inferior in performance to the 8500.

well, i would agree 128 could be helpful.
i haven't come across a game my 64MB 8500 doesn't run very nicely.
including heavy amounts of UT2K3.

 

[MikeD2k3]

jiffylube1024 and bfg10k - thanks for your indepth posts. I looked up the GF4 Ti4200 @ Newegg and found the cheapest cards to be $118 for a 64MB version. A Radeon 8500 LE 64MB is $72 and a 128MB is $94. The increased frame rates at lower resolutions probably aren't worth $46 I wouldn't think. I'm trying to get the most bang for buck.

I'm trying for a nice combination of eye candy and fps. Right now I'm running UT2k3 at 512x384 with all eye candy disabled and all details set to absolute minimums. With flyby-antalus I get an average of 22fps. If I could get to enable almost all the features to nearly maxed out at 1024 and get 40+ fps I would be overjoyed. I plan on keeping this card for 3 or so years. If you couldn't tell from my current possession of a TNT2, I tend to hold onto things for a long time to maximize value.

Thanks again!

 

[rogue1979]

Be careful when purchasing the Radeon 8500 LE, some of them have lower speed memory that doesn't overclock very well. You want to find one that comes stock at 250/250.

 

[Slaimus]

Almost all the built by ATI 8500, both LE and regular, 128MB's have 3.3ns BGA memory, guarenteeing OC to 600mhz mem. When newegg had the 64MB 8500LE bulk's, they also had 3.3ns memory.
As for needing 128MB, you can easily find ways around not having 128MB, which can be as simple as turning on texture compression. IMHO if a game needs 100MB of textures on a card, there isnt enough bandwith to use all that efficiently.

 

[BFG10K]

MikeD2k3, get the 128 MB card over the 64 MB one, especially if the price difference is only a few dollars. I cannot stress this enough.

Pretty much any 3D game from Quake3 and onwards will benefit from 128 MB RAM over 64 MB RAM, especially if you like to run the texture detail levels at their maximums. You don't even need high resolutions as medium resolutions with high texture detail levels are enough to kill 64 MB cards. I highly recommend you read this article as it explains the basics very well and it has comprehensive benchmarks too.

In addition, benchmarks don't tell the whole story as 128 MB cards will give you smoother and more consistent gameplay and you'll have less stuttering from texture swaps. From my own testing I've found that 64 MB cards are severely disadvantaged in today's modern games because the high texture detail settings really kill them. RTCW and JK2 (for example) frequently texture swap at even a medium resolution like 1152 x 864 combined with the highest texture detail levels. And if you choose to turn down the texture details it might help but you're simply not experiencing the game as it was meant to be experienced.

The increased frame rates at lower resolutions probably aren't worth $46 I wouldn't think. I'm trying to get the most bang for buck.

The cards with higher amounts of RAM will really start to shine in high resolution and high detail settings, not at low resolutions.

Just say, at realistic settings, with no AA or AF, the 8500 is totally inferior to the Ti4200 . Cheers!

Dude, if you get a Radeon card those settings I described become realistic. At the very least you can turn on 16x anisotropic and suffer only a negligible performance loss. There's nothing wrong with having high standards and I'm only encouraging others to do the same.

would like to see the results of those same benchmarks/cards using the latest Detonator 40.XXs and Catalyst 3s. Nvidia's reference drivers were already optimized(28.32s are close to 40.XXs in performance), haven't the ATI drivers improved quite a bit since April 22?

Agreed but even back then those benchmarks show quite well what I'm trying to say.

 

[MikeD2k3]

bfg10k - thanks for the link, it was very informative. I didn't know the Ti4200s took such a performance hit when you enabled AA. It's a shame the reviewer couldn't have run the benchmarks with the full version of UT2k3 on. Sounds like I should spring for the 128MB card to help futureproof it.

 

[BFG10K]

If you think the performance hit on Ti4200s is big with FSAA, wait until you see how many frames they lose with anisotropic filtering.

This is a stark contrast to 8500s which can have full anisotropic filtering with only a negligible performance hit.

 

[railer]

Nice postings you guys. Even you BFG. MIke D, if you're looking for bang for your buck (as I always do), the 8500's are hard to beat. I've seen refurbed 128 meggers for $69 shipped. My GF4 4200 was still $85 shipped, and to me is just barely worth the price premium, because to me, it is significantly faster. I turned on AA and AF in RTCW and honestly could not tell the diff. Playing at the (to me) absured detail levels BFG plays at, you may well want the 128 meg card. For us normal folks, the GF4 4200 will be faster every time.

Jiffy said it best:

Again, don't bother arguing with BFG's favourite line "In high resolution situations combined with anisoptropic filtering the Ti4200 is totally inferior in performance to the 8500. " Just say, at realistic settings, with no AA or AF, the 8500 is totally inferior to the Ti4200 . Cheers!

 

[rogue1979]

I don't have a 9700 like BFG10K, but I have to agree with him. 1600 x 1200 x 32-bit with no FSAA and 16X anisotropic are very realistic for a Radeon 8500 with a decent cpu. My retail 8500 128MB @ 312/301 is averaging more than 100fps at these settings in Unreal Tournament. 55fps is about the lowest framerate I see in a free for all with 16 other bots, guns blazing. I must also admit that I was wrong about the extra performance boost with 128MB of video memory. After closer examination I found there is a noticable boost. The average fps stays close to the same, but the minimum is higher and gameplay during ultra harsh/intense action is a little smoother. At lower resolutions and settings the Geforce4 Ti4200 is definately faster. But at 1600 x 1200 and 16X anisotropic with details set at high/quality, my Gainward Geforce4 Ti4200 128MB @ 300/540 falls behind the 8500 and does not look as good. I now have three other 64MB Radeon 8500's in the house, and they are roughly 90-95% the speed of the 128MB version, which still puts them ahead of the Geforce4 Ti4200 at the settings mentioned above.

 

[MikeD2k3]

On an interesting sidenote, I got to play the UT2k3 demo with a 128MB Ti4200 over the weekend. I was at a Christmas party and the host had a 2GHZ P4/512MB system. It was definitely a lot smoother than my system (TNT2 32 MB 1GHz Athlon TB/256MB). Oddly, whenever we benchmarked we always seemed to get 60fps--even when comparing the settings I run at currently (512x384, all textures min, all eye-candy off) to 1024x768, all textures demo-max, all eye candy on. Running it maxed out at 1280x1024 the flyby-antalus benchmark dropped to around ~40fps I think. Doesn't the lack of higher fps at lower resolutions/low texturing/no-eye-candy seem a little odd? I didn't have time to try out enabling AA or AF--he had both disabled. However, I didn't really notice a ton of aliasing everywhere. Compared to my system, aliasing was non-existent. I thought he had AA and/or AF enabled it was so clear. Makes me wish I had my updated video card now!

 

[BFG10K]

Oddly, whenever we benchmarked we always seemed to get 60fps

It sounds like your friend's computer had vsync enabled.

 

[clicknext]

Not getting the extra memory would be a big mistake.

 

[GtPrOjEcTX]

Originally posted by: BFG10K
Most probably. You'd be surprised just how many people tell me something is "smooth and doesn't have any problems" only to change their story after they run benchmarks are are shown quantifiable evidence to the contrary.

Wow, you must be in sales. If a person is happy with their setup, why show them some numbers of what they could bench when they are satisfied with what they currently have?

 

[Mustanggt]

You will be ok with the 64 Mb card, save the $22 for a future upgrade, I can run any game just fine on my system with 64 Mb Radeon 8500, 10400 3d mark scores is good to go.