Memory Clock vs. Memory Size

t33lo

Distinguished
Jul 15, 2005
405
0
18,780
I am looking at two different cards which are roughly the same price.

BFG Tech 9800GTX+

-------------------------
Core Clock: 750Mhz
Video Memory: 1GB
Memory Clock: 2000MHz

EVGA 9800GTX+
------------------------
Core Clock: 738Mhz
Video Memory: 512MB
Memory Clock: 2200Mhz

At relatively the same price, which one is the better buy? Is the 200MHz increase in the Clock better than the 512MB in actual size? I want to think that the extra 512MB would be better for say a few years from now since texture sizes in games are increasing, but for right now, comparing the two side by side, which is better?I also plan to do a bit of overclocking, not sure if that matters. Thank you!
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780



Thanks for your input. I guess my bigger question though is what is better, having the 1GB @ 2000MHZ vs. 512MB @2200MHZ and why? Thanks!
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780



So the extra memory size of 512MB is mainly for higher resolutions? I have a 1080p monitor (1920x1080) so I won't ever be going up to that resolution for a long time. In that case, go for the 512MB?
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780


Thx Zerk, I think I will go with the EVGA 512MB one since its a tad cheaper and has gotten better reviews anyways. I'm kind of a technical guy and like to learn the in's and out's of everything, so do you or does anyone know any technical reasons as to why the 512MB with the extra 200MHz Memory Clock is better than the 1GB with 200Mhz less in the memory clock? And what would be a reason as to why the 1GB would be better where and when? Thank you for all your help!
 

4745454b

Titan
Moderator
Its not quite that simple as the BFG card has a higher core clock as well. Personally I'd get the one with more ram, as long as the bundle is as nice. Clock speeds can be changed, the amount of Vram can't. (I wouldn't worry about how long the warranty is, who wants a 9800GTX four years from now?)
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780



The core clock is a tiny bit higher on the BFG because I believe BFG specializes in pre-overclocked cards. Since it's only 12MHz, I believe the EVGA one would be OC just as easily. It's funny because I am still running my 7800GTX 256MB from almost exactly 4 years ago at 520/1330. I'm not sure how long I will keep this next card for, but I would think at least 2 years.

How I see it is that the core clocks are a wash, the 1GB memory will be better for say a few years from now (when games demand the extra 512MB), but in the meantime, the 512MB EVGA will hold its own and peform even better than the 1GB BFG because it has 200MHz extra in Memory Clock. So I guess it comes down to short term vs. long term? Is this a fair assessment or am I missing something? Thank you for your help!
 

Zerk

Distinguished
May 6, 2008
297
0
18,790



Yea! Good question.

For instance a card that has 1GB of RAM and a 256bit bus will only be able to address 512mb consistently fast... the more RAM you use on the card, the more latency increases so it will be sufficiently slower in memory intensive apps than a card that has 1GB and a 512bit bus.

1Gb - 512Bit Bus
512Mb - 512\256bit Bus
256Mb - 256\128bit Bus
128Mb - 128\64bit Bus

Don't get.

A - 1Gb - 64\128\256bit bus
A - 512Mb - 64\128bit bus
OR A - 256Mb - 64bit bus

Hope that helps you out. Basically the higher the memory bus interface is the better as it will give more bandwidth but more memory give higher latency witch is bad. At the moment a card with 512Mb of ram and a 256Bit memory interface is the best to go for. If you must get a card with 1Gb of ram make sure it has a 512bit memory bus or the extra 512Mb will be almost useless

If you run at resolution higher then 1920x1200 then a 1gig card is better for Crysis or a Very demanding game.


All in all.

Less ram and higher clock rates are better for lower res,Because it will render faster.
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780


Thank you, I think you answered my question. Is that why the Memory Clock on the BFG 1GB is rated lower at 2000MHZ instead of the 2200MHz?

I hope this isn't all in vain, but right now I am running...

AMD Athlon 64 3700+ San Diego 2.2GHz 1MB L2 Cache Socket 939 Single-Core Processor
2x512MB Crucial Ballistix PC3200 2-2-2-6
2X512MB Crucial Ballistix Tracer PC4000 2.5-4-4-8
DFI LANPARTY nF4 SLI-DR 939 NVIDIA nForce4 SLI ATX AMD Motherboard
BFG Tech GeForce 7800GTX 256MB 256-bit GDDR3 PCI Express x16 SLI Supported Video Card
Western Digital Raptor WD740GD 74GB 10000 RPM 8MB Cache SATA 1.5Gb/s 3.5" Hard Drive
OCZ PowerStream 600W ATX12V 2.01, ATX, BTX, SATA, P4 and EPS12V Power Supply

I mainly play CS:S for the time being, but do plan on getting into some of the other games that have come out in the past few years as well as games that will be coming out in the coming months. Will the EVGA 512MB be bottle-necked by the CPU, memory, mobo, or all of the above? My plan was to keep this system from the Summer of 05' to maybe a year from now. I also plan on giving the CPU a slight overclock (have done this before and gotten stable at around 2.7GHz) but am not familiar with overclocking two different types of RAM especially 4 sticks at different timings (probably not a good idea). Thanks!
 

4745454b

Titan
Moderator
Since it's only 12MHz, I believe the EVGA one would be OC just as easily.

And the BFG won't overclock its memory? We're only talking 100MHz to get to 200MHz effective. Seems to me you'd be able to overclock both of these cards to the others specs.

For instance a card that has 1GB of RAM and a 256bit bus will only be able to address 512mb consistently fast... the more RAM you use on the card, the more latency increases so it will be sufficiently slower in memory intensive apps than a card that has 1GB and a 512bit bus.

Any proof to this? I assume you also mean DDR3 and not DDR5. Or are you implying that the 4870/4890 are junk cards because they have 256bit memory buses?
 

wh3resmycar

Distinguished


its a resolution high enough to warrant a 1gb card. and 100mhz (x2) mem overclock is not that hard to achieve.

get the 1gb card.
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780


Yeah, I was just saying the Core Clock was a wash. So I guess OC'ing wouldn't be an issue as I could probably get the 1GB Memory Clock to roughly 2200MHz (effective).



any bottlenecking involved? Also, should I just throw in $80 more and get the GTX 260 which is 448-bit as opposed to 256-bit? I also noticed that the GTX 260 has a memory clock of 2106MHz, @ 896MB in size, so it does seem that more memory does slow down the memory clock (or so it appears that way).
 
G

Guest

Guest
no gtx 260 has more stream processors and wider bus, but in your res you wont see any big difference
 

t33lo

Distinguished
Jul 15, 2005
405
0
18,780
no gtx 260 has more stream processors and wider bus, but in your res you wont see any big difference

I'm confused by what you said. It has a wider bus 448 to 256 and more stream processors 192 to 128, albeit, with a lower core clock and memory clock. I wish there was some formula to plug in all the specs to determine the better performance overall. Like, the extra 64 stream processors, how much more is that worth than the diminished core and memory clock speeds?
 

wh3resmycar

Distinguished


why wouldnt he? hell ive upgraded from a 4850 up to a 4870, and now a gtx260-216. and @ a resolution running @ 1440x900 im feeling a difference already.

the higher you go up the resolution, the more gpu extensive the game becomes. and considering there is a world difference between a gtx260 and a gts250(9800gtx+), your statement makes no sense.



if $80 seems not to be a problem, by all means get the gtx260 ( get the 55nm version), be sure to have a big big chassis though. its an awfully big card.

 

4745454b

Titan
Moderator
why wouldnt he? hell ive upgraded from a 4850 up to a 4870, and now a gtx260-216. and @ a resolution running @ 1440x900 im feeling a difference already.

Care to quantify that statement? 1440x900 isn't much different then 1280x1024. All three of those cards should be able to max out that res. If you're already running max with the 4850, why would the 4870/GTX260 make it look better?
 

wh3resmycar

Distinguished


simple.

farcry2, crysis (crysis wars) would let you know what im saying. and i play both online so putting a 4870 previously and my new gtx260 really helped p@wn people a lot (more fluid fps)while giving me the sense of visual satisfaction..

nba2k9 with 16xAF and supersampling filtering is very doable with vysnc-on with the 4870 and the gtx260-216 @ a very steady 60fps without a flinch (with the 2k9 cam). you cant do those on a 4850. you just cant. and anything below 60fps on my nba2k9 game is unacceptable.

devilmaycry4 will fall below 60fps with all the settings cranked + the driver enhancement with a 4850, not gonna happen on 4870/gtx260. (vsync on as always with 3rd person perspective games)

prince of persia / assassins creed as well.


sure a 4850 is a pretty decent card for a 19" panel, but heaven starts with the 4870/gtx260-216.

 

4745454b

Titan
Moderator
Sorry, but I'm having trouble believing you. Here is my link.

http://www.hardocp.com/article.html?art=MTU3MSwzLCxoZW50aHVzaWFzdA==

Here is the 4850 playing Farcry2. Notice that they are playing at 1920x1200 This is massively higher then the 1440x900 res you play at. (2,304,000 vs 1,296,000. 1920x1200 is nearly twice as many pixels.) It has the same settings as the other cards, and it has 16x AF enabled. The only thing missing is AA, which I'm sure you could use if you halved the number of pixels being drawn. The frame rate graph shows no massive spikes or valleys, it is a rather steady ~40FPS, AT NEARLY DOUBLE YOUR RESOLUTION.

Either your computer isn't capable of driving the 4850 as fast as it could, your lying, you never bothered to test the cards properly, or you are simply making yourself feel better about spending the money. 1440x900 is low res gaming, nearly every card can do it. I get the feeling that if anyone else tested the 4850 at 1440x900, we would hear that it can play at that res quite well, even with AA/AF enabled.

 

wh3resmycar

Distinguished


im sorry? lol. btw, my rig is a q6600 @ 3.0ghz + 3gb RAM @ 800mhz+ asus p5k-se. now that would bottleneck a 4850 would it?

ive done extensive testing on my previous cards. hell ive fully documented how a new catalyst performs on my system specifically using the farcry2 bench. here they are. i think ive posted this on every "catalyst 9.x released" thread out here.

the card that i used on these successive test was my old (and loved) hd4870 1gb from Palit:

ati1.jpg


the settings:

92_settings.jpg


catalyst 9.1

91_avg.jpg


______________________________________________



catalyst 9.2

92_avg.jpg


_____________________________________________

catalyst 9.3

93_avg.jpg


____________________________________________


catalyst 9.4

fc2_94.jpg



but then again you cant compare these numbers on what hardocp has, because they are doing a different test (not the farcy2 bench).

my cat 9.5 numbers are similar to the 9.4 results. but i stick with the 9.3 anyway because its a tad faster.

____________________________________________________________________________________________________________

im not yet able to upload my gtx260-216 test but its faster, around 58fps on avg if im not mistaken.




i never said it cant. but theres a big difference between playing it with a 4850 vs a 4870/gtx260-216m, unlike what the other dude is saying. like what ive said the 4850 is a decent card but heaven starts with a 4870/gtx260-216. hell ive been blown away the first time i used my 4850, got blown away even farther with my 4870.

and i have guru3d to attest to that: http://www.guru3d.com/article/ati-radeon-hd-4770-review/9

30fps on 1280x1024 is not bad. but is it great?






 

wh3resmycar

Distinguished
i found my old FC2 test from my old 4850:

here they are: same settings as above.

catalyst 8.12

4850_812.jpg



_________________________________________

and from my new gtx260-216 as well:

gtx260_fc2.jpg



theres a world difference playing it @ 38fps against 58fps.

and if you still have doubts, i'd be more than happy to send you all the benchmarks result files.
 

wh3resmycar

Distinguished


????

you are aware that hardocp is not using the built-in farcry2 benchmark are you?

yeah you are missing something, common sense.