5870, 5850 of 5770 Xfire?

[Pisklink]

Thanks for all the valuable input!!!! I read several arguments about micro stuttering but after reading the comments it's still not clear to me what is exactly.

Before I decide which setup to go for I have to measure my case to see if the 28 cm long 5870 together with power connectors will fit my case (CM690). I think it will hardly fit though.

Two 5770's are interesting to me due to financial reasons because they are together about 60 - 70 euros cheaper than the best available 5870 in my country. But since I don't have the money this month to spend on a new videocard, I think I will wait till the release of Fermi. I guess that the Fermi release will drop prices of the 5870 significantly. If that's true I would fancy a 5870 over 5770's in CF. However the 5770 will get cheaper as well of course. Mmmm, hard choices are to be made!!!!

[Musho]

Thanks for all the valuable input!!!! I read several arguments about micro stuttering but after reading the comments it's still not clear to me what is exactly.

Before I decide which setup to go for I have to measure my case to see if the 28 cm long 5870 together with power connectors will fit my case (CM690). I think it will hardly fit though.

Two 5770's are interesting to me due to financial reasons because they are together about 60 - 70 euros cheaper than the best available 5870 in my country. But since I don't have the money this month to spend on a new videocard, I think I will wait till the release of Fermi. I guess that the Fermi release will drop prices of the 5870 significantly. If that's true I would fancy a 5870 over 5770's in CF. However the 5770 will get cheaper as well of course. Mmmm, hard choices are to be made!!!!

Why not go for the 5850 if you are on a budget? It isn't that much slower clocked at the same speed (and they clock about the same on stock cooling). Reviews show a 2-5% difference at the same clock speed with some odd spike to 9% maximum or so. It isn't enough to justify the large increase in price of going from 5850 to 5870. 5870 is only a viable solution if you simply want the best of the best single GPU (at the moment), or if you don't overclock, in which case the higher clock speeds on the 5870 together with the extra shaders definitely help. But we're at xtremesystems.org, so both should be running their maximum stable clocks, which should be roughly the same

Also, because the 5850 slightly cheaper, but still a single card, it's a viable option to go crossfire 5850 later on if your budget grows

[JaD]

I read several arguments about micro stuttering but after reading the comments it's still not clear to me what is exactly.

Download this little demo and see for yourself:

http://www.computerbase.de/downloads...er_single-gpu/

[janolle]

I had two HD5870 cards for a month and I kept disabling CF all the time because it just didn't feel any faster to the eye. FPS counters showed huge improvements and that's when I started to look for testmethod and found out that frametime graphs could easily be made with FRAP+Excel. There was no coming back from that point, my observations where now backed by real evidence so I sold the other card.

The simple fact is that high FPS number doesn't matter because the eye can see the slow frametimes that every other frame has in CF. There simply is no point to CF for gamers. There are people that only look at FPS numbers and don't believe what their eyes tell them. Benchmarkin is a different world, who cares about stuttering when you get huge scores.

EDIT:
Let me brake down the problem with looking at FPS values. Imagine 100fps. That's 100 frames in one second. The frames can be spread evenly and you'll see a perfectly smooth movement. But if the frames a spread unevenly like in CF it won't be smooth anymore. Five frames can be shown for 0.1 second each (wich is a rate of 10fps) and 95 frames are evenly spread between these. That would be 100fps to fraps or any other benchmarking software but it would look just like 10fps. This is what microstuttering does: it shows slow frames in between fast ones and it looks very slow compared to an even framerate.

[Pisklink]

Why not go for the 5850 if you are on a budget? It isn't that much slower clocked at the same speed (and they clock about the same on stock cooling). Reviews show a 2-5% difference at the same clock speed with some odd spike to 9% maximum or so. It isn't enough to justify the large increase in price of going from 5850 to 5870. 5870 is only a viable solution if you simply want the best of the best single GPU (at the moment), or if you don't overclock, in which case the higher clock speeds on the 5870 together with the extra shaders definitely help. But we're at xtremesystems.org, so both should be running their maximum stable clocks, which should be roughly the same

Also, because the 5850 slightly cheaper, but still a single card, it's a viable option to go crossfire 5850 later on if your budget grows

I have to admit I didn't seriously consider this option because when I review the benchmarks this card isn't very much faster than my current 4870 512Mb at 1680 x 1050 resolution which justifies its price. Of course the 5000 serie has eyefinity, draws less power in idle and supports DX11. The only thing that would concern me is 1Gb DRAM and DX11, I couldn't care less about eyefinity and power draw.

However if you can convince me that the 5850 is really much faster than my current card I would definitely consider it because of my budget, 1Gb DRAM, DX11 support and possibility to fit in my case more easily due to approx. 25 cm length.

Download this little demo and see for yourself:

http://www.computerbase.de/downloads...er_single-gpu/

I will take a look but I'm a bit carefull with installing programs I don't know.

I had two HD5870 cards for a month and I kept disabling CF all the time because it just didn't feel any faster to the eye. FPS counters showed huge improvements and that's when I started to look for testmethod and found out that frametime graphs could easily be made with FRAP+Excel. There was no coming back from that point, my observations where now backed by real evidence so I sold the other card.

The simple fact is that high FPS number doesn't matter because the eye can see the slow frametimes that every other frame has in CF. There simply is no point to CF for gamers. There are people that only look at FPS numbers and don't believe what their eyes tell them. Benchmarkin is a different world, who cares about stuttering when you get huge scores.

EDIT:
Let me brake down the problem with looking at FPS values. Imagine 100fps. That's 100 frames in one second. The frames can be spread evenly and you'll see a perfectly smooth movement. But if the frames a spread unevenly like in CF it won't be smooth anymore. Five frames can be shown for 0.1 second each (wich is a rate of 10fps) and 95 frames are evenly spread between these. That would be 100fps to fraps or any other benchmarking software but it would look just like 10fps. This is what microstuttering does: it shows slow frames in between fast ones and it looks very slow compared to an even framerate.

Thanks for your extended explanation. If I understand you correctly you try to say that the phenomenon microstuttering actually means that the achieved frames are not evenly distributed over 1 second. For example when 90 frames are shown in 0,1 second and 10 frames in 0,9 second the eye won't perceive this as a smooth 100 fps while a benchmark program like fraps does.

[Musho]

Well, you asked us to make a choice between 5850, 5870 and 5770. I never doubted the ability of the 4850 to be able to drive a 1680*1050 resolution just fine. For the best performance/price, go with the 5770CF. With future upgrades in mind, a 5850 might be a better option, as you can decide to run that one in crossfire later down the road. Plus it won't suffer from any crossfire problems if you run it solo, vs the 5770CF. 5850 will also draw less power and produce less heat than a crossfire setup. 5870 simply isn't an option for you, since you seem to be on a budget. It performs up to 5% faster than a 5850 clocked at the same speeds. And they both clock roughly the same. For 100€ extra, that isn't a very good deal. It's only a valid option if you only run stock (Huh? ), or if you simply want the best of the best single GPU from the ATI camp.

Bottomline, if you are really on a budget, just don't upgrade. The 4850 is fine for that resolution, really. If you have the itch for something new, there's 2 options: 5770CF and 5850. Both have their pros and cons.

[demonkevy666]

rats should have waited longer I see showing up on newegg 2Gbs 5870 now >_< darn it.

[Pisklink]

I never doubted the ability of the 4850 to be able to drive a 1680*1050 resolution just fine.

I have a 4870 at the moment and I have to admit that it still fits the purpose. Besides that it is fun to have the newest available hardware, I started to look for a new card because I started to have problems running new games at 1680 x 1050 with full options enabled (e.g. AA and AF) and incompatibility with DX11. I think the problem with enabling full options is that my card only has 512Mb memory.

It performs up to 5% faster than a 5850 clocked at the same speeds. And they both clock roughly the same. For 100€ extra, that isn't a very good deal. It's only a valid option if you only run stock (Huh? ), or if you simply want the best of the best single GPU from the ATI camp.

Do you mean that the 5850 overclocked to 5870 clock speeds only performs 5% slower than a 5870? And is it possible to run a 5850 at 5870 clock speeds, does it have that much headroom?

I also thought that the 5850 runs much hotter in proportion to the 5870 at idle as well as at load?

[Musho]

I have a 4870 at the moment and I have to admit that it still fits the purpose. Besides that it is fun to have the newest available hardware, I started to look for a new card because I started to have problems running new games at 1680 x 1050 with full options enabled (e.g. AA and AF) and incompatibility with DX11. I think the problem with enabling full options is that my card only has 512Mb memory.



Do you mean that the 5850 overclocked to 5870 clock speeds only performs 5% slower than a 5870? And is it possible to run a 5850 at 5870 clock speeds, does it have that much headroom?

I also thought that the 5850 runs much hotter in proportion to the 5870 at idle as well as at load?

The 5850 has a slightly inferior cooler as far as I know, but it doesn't result in a big temperature difference. A few degrees max. The 5850 still clocks about just as well as the 5870. Both seem to max out at around 1000 core/1300 memory speeds on stock cooler if you get a good card.