2nd 4870?

[Techman_15]

I have a Diamond 4870 1mg card. It runs good, but I have a 30in monitor, and I can't run all my games at the full resolution (2560x1600). Am looking for a 2nd 4870 1mg. Does it matter if i get a different brand? Or should I go for a Nvidia 285?

edit - I am running an i7 920 processer on an ASUS P6t Deluxe, so Crossfire (or SLI) should work fine.

 

[EXT64]

Assuming you have a crossfire board, 2 4870's will beat 1 285 almost always. You don't need to get the same brand, but try to get the same specs (clocks, ram) as it will reduce the higher one to the lower one's specs.

 

[gkay09]

^+1...As you already have a 4870, get another one ins crossfire and like EXT said get the same specs...
The 4870 xfire will beat the 285 overall...

 

[quake1fan]

I have a Diamond 4870 1mg card. It runs good, but I have a 30in monitor, and I can't run all my games at the full resolution (2560x1600). Am looking for a 2nd 4870 1mg. Does it matter if i get a different brand? Or should I go for a Nvidia 285?

edit - I am running an i7 920 processer on an ASUS P6t Deluxe, so Crossfire (or SLI) should work fine.

Probably. It depends. A second 4870 can certainly give you better performance as long as you aren't currently using more than 1GB of video RAM in your games. If you are maxing AA and AF and using max texture settings, then it's possible that you are trying to use more than 1GB of video RAM. In that case, no, I don't think a second 4870 won't make much difference.

Otherwise, yes a second 4870 for a 30" screen will make a nice improvement in games, but of course not at the desktop.

 

[boulard83]

+1.

CF HD4870 1gb can do very well at your Res. 2560x1600 need BIG Horse.

 

[HVDynamo]

At that Resolution a second 4870 will make a big difference, but like others said, make sure to match the memory amount and other specs of the card ( ie core speed/memory speed and whatnot) otherwise the card with the smaller/slower specs will force the other card to operate at the same capacity.

 

[Euphoria_MK]

I got myself a second hd4870 1GB and it makes difference for sure.
It doesnt matter what brand hd4870 1gb you purchase as long as they are running at same clock speed.

There are only couple of games that use up to 1GB of Video Ram and you will definitely see a performance increase at that resolution.

Cheers

 

[Techman_15]

THX for the input. Going to newegg now.

 

[radiowars]

I got myself a second hd4870 1GB and it makes difference for sure.
It doesnt matter what brand hd4870 1gb you purchase as long as they are running at same clock speed.

There are only couple of games that use up to 1GB of Video Ram and you will definitely see a performance increase at that resolution.

Cheers

Clock speeds don't matter. As long as you have the same amount of Vram, then everything should be fine. The faster card will match speeds of the slower card.

 

[daedalus685]

I have never seen data on it, and only have my own experience on testing it.

But.. I'd like soemone to prove me wrong i they can, I'm curious.

From my testing it seems that the clock speed increase of one card only DOES increase the crossfire performance. That is to say crossfire scales the workload between the cards based on clock speed. It does not simply truncate the extra clocks as it does with frame buffer. I have found the returns to diminish quite quickly as the gap increases.. perhaps it was just a coincidnece trying to find such small increases. I vaguely remember reading about this years ago.. not sure how it works exactly now..

Would love to see a tech document about it if someone has one from a good source that says one way or another.

 

[boulard83]

Do your OWN testing with you HD4890 !

 

[daedalus685]

lol, I have been , but the difference is soooo small i cant really say it is due to coincidence or if crossfire actualyl works that way .

Just wondering if a real tech paper si floatign around on the subject as I have not foudn one.. Though I do believe currently, that to a point, the crossfire does not actualyl scale either card up or down, and will use the extra clocks.. just not all that efficiently.. as the gains are well below the OC fractions one would expect..

 

[Euphoria_MK]

Clock speeds don't matter. As long as you have the same amount of Vram, then everything should be fine. The faster card will match speeds of the slower card.

You contradict yourself. So clock speed do matter since the slower card would dictate the overall speed, thus is not worthed for him to spend money on a OCed version of HD4870 1GB if he already has one that runs at stock 750MHZ

 

[boulard83]

Euphoria_MK have it right.

Both SLI and CF will clock the 2 cards to the LOWEST one spec And will use the amount of memory from the LOWEST one too.

A 512MB stock HD4870 CFed with an OCed 1GB HD4870 will run at the STOCK clock and will use 512mb. You can OC the slowest one to reach the other one. But youll never be able the add memory to the card

 

[daedalus685]

Obviously that is how it works for memory. But, at least for crossfire, the clocks don't "change."

If card A is faster than card B, A will not show up as runnign at slower clocks in GPUz etc. Card A will finish its workload first and sit there a bit, but Card A will not start doign its workload at the speeds of Card B. In fact I beleive, though have no particualrly conclusive proof as the performance variances are within random error and I have yet to collect enough data, that crossfire actually will give the faster card a different workload such that the system does, however small it may be, perform faster if one card is faster than the other..

I think it is a misconseption about crossfire as it is generally assumed it works the same as SLI, which it does not.

 

[boulard83]

HUmmm ill have to read more about CF ... i not sure about this.

 

[daedalus685]

Let me jsut say thuogh... becasue of the far less than perfect scaling of crossfire/sli (in this case crossfire) it is almsot impossible to detect the difference when one card is even a good bit faster than the slower one... Hense why I can't prove it ..

I find it interesting though.. Guess ill have to send ATI some letters and ask one of thier R&D guys how it works.. this point is totally moot in terms of actual gains.. there is no point increasing one card and not the other, nor a point in spendign mroe for an OC eersion (unless it is for a better cooler).. you see no real world difference.. I'm just blabering on about curiosities..

 

[james49]

well i have two 4870s and cant get catylest program to show the crossfire option has any one else seen this? iv removed all drivers with driver cleaner reinstalled to a differnt versing uninstalled those updated agine and still cant find the crossfire option in the caytalest manager box. phenom II quad core 940, asus m4a79 deluxe, 4870 crossfire ready, 4 gb of ram, 500 gb sata hd.

 

[daedalus685]

That is unfortunately pretty common james. It is due to either a ad driver install or the bios on the cards.

Try 8.10 drivers.

If you want to use newer drivers you may have to flash the cards with a different bios...

 

[boulard83]

5. What happens when graphics cards with varying memory speeds and engine clocks are paired together?

A. In this scenario, overall performance will be adjusted to match the memory speeds and engine clocks of the lower clocked card. To obtain ideal performance, graphics cards should be paired correctly.

DIRECTLY from ATI site ......

 

[james49]

last time i checkedit was the other way around, system uses the higher memory, and clock speed, i;ll check agine, but nonetheless, iv paired 2 4870's with exactly the same specs. and caytalest is still screwed up, ati needs to fix this problem with a good set of driveres that work with the system ur running.

 

[Euphoria_MK]

last time i checkedit was the other way around, system uses the higher memory, and clock speed, i;ll check agine, but nonetheless, iv paired 2 4870's with exactly the same specs. and caytalest is still screwed up, ati needs to fix this problem with a good set of driveres that work with the system ur running.

Hi James,

Use the latest Cat 9.4 drivers. There are few reasons Crossfire will not be reported in the CCC Options:

1) Cards are not connected properly to power supply
2) PSU does not provide enough power (You'll need min of 650W just to be on the safe side, it all depends on how may devices you have connected in your pc)
3) Crossfire bridge is not properly connected
4) Your motherboard does not support crossfire

If none of the things above are true than power down your PC take the 2nd card out, reset your BIOS to default and boot the PC. When is done booting, power down again. wait for couple of minutes and plug in your 2nd hd4870. Connect all hte cables correctly and power up again.
Go into CCC and see if crossfire option is available.

Cheers

 

[james49]

ok ill do that, i have750w psu, i know the cards are good i tryed em each alone. i know the pcie 16 slots are good i used em both, know that the mobo supports crossfire as i have the m4a79 delux, phenom II x4 940, 4gb of ram. thanks friend ill let u know.