ATI Radeon HD 4000 Series discussion

[MrHydes]

lol...

[By0T0xN]

Well said Solus.

[Philip_J_Fry]

i wont make any judgment on what it is ... but if cnet's article is right as pointed out by zerazax

http://news.cnet.com/8301-13924_3-9978746-64.html

"When you gang up graphics chips (using the traditional Scalable Link Interface or CrossFire technologies) they roll off pretty fast. ("Roll off" implies that performance doesn't scale up well.) "So when you put two boards in, you don't get twice the performance but you (only) get one and a half. You put four boards in and you (only) get about 1.7, 1.8. What ATI is saying is that with two chips using (their) proprietary inter-bus, they will get 1.8 (the performance) with two chips. If that's true, you can expect to see four of them giving you something around 2.5."

Getting 2.5 times the performance from four boards would be a masterstroke for ATI.

sounds really interesting, although those "1.5 times for 2 cards, 1.8 times for 4 boards" numbers are somewhat arguable

i will definitely wait for benchies then make decision, thankfully i waited to see how GTX 280 played out

[By0T0xN]

Arguable, yes. There are certain games and 3D apps that scale better than others. But the majority is what I believe the writer is refering to. And I am with you on just layin low and waiting for the real world benchies to come out and tell the true story.

[Kingcarcas]

nobody helps you with that avatar of yours! Repulsive + Ugly!

Booo!!!! Ban Hammer! j/k

[XCheater]

I'd go with the 4870... 2 reasons

1.) In cf, adding a 4850 would slow your 4870 down to act almost like a 4850.. load balancing.

2.) In time, if the 4870 gets a bit slow for your requirements, a second will be cheaper then what you bought the first one for. Additionally, adding a 4870x2 may be a very viable option for a tri crossfirex setup. In a years time, they could also be quite cheap.

in another thread in this forum this is not true, since 4870 and 4850 in crossfirex are scale very well indeed unless your cpu is bottlenecking

[w0mbat]

As i said before, r700 isnt just like r680. It still has the plx chip (now w/ pcie 2.0) but it also uses a so called "crossfirex sideport" for communication. Just wait a bit and u´ll see - it scales better than normal cf.

[Apoc@]

well, if your on a tight budget and im assuming you are or you would consider 4870xfire, why dont you buy 1 4850 and see if that suits you and if it dont then you could always add another or a 4870.

It's not the budget, it's more whats realistic for a 1680x1050 reso. I can easy burn 450 euro on a set of HD4870 but do you really need that on sutch a reso? That's my question

I'd go with the 4870... 2 reasons

1.) In cf, adding a 4850 would slow your 4870 down to act almost like a 4850.. load balancing.

2.) In time, if the 4870 gets a bit slow for your requirements, a second will be cheaper then what you bought the first one for. Additionally, adding a 4870x2 may be a very viable option for a tri crossfirex setup. In a years time, they could also be quite cheap.

1.) I know about load balancing, but that's not the problem. I won't be buying 2 differend cards. It's 1 HD4870 or 2 HD4850.

2.) In time.... The setup only needs to last till the next generation Radeons hit the shops ( the HD5xxx if that will be the name ).

[annihilat0r]

If 4870x2 solves microstutter it'll be a great card; if it scales well with all games, of course. CF scaling is less consistent than SLI.

But I still can't find a reason to have more performance than a single 4870 below 1900x1200 resolution. All games are console ports.

[u22]

Hi

OC Tool für 4870 ?

CCC max 790/1100

27004 im 06er
http://service.futuremark.com/compare?3dm06=7240841

[Logos]

was expecting some test results from GAR this morning... concerning how CF performs on P45 (2 x 8 lanes) compared to X48 (2 x 16)...

[Macadamia]

It will use AFR, that's the nature of a dual chip card. Unless they have a big memory space for shared workload exchange that is.

What's interesting here is that with the "sideport", there's actually a chance of syncing frames or negotiating delays, spreading them around evenly so that the experience is great overall. AFR without the worst of AFR.

[annihilat0r]

But with less FPS associated with delaying one of the GPUs.

[Helmore]

But with less FPS associated with delaying one of the GPUs.

Not necessarily....
If you do it right then it will only be a small hit, nothing drastic. Now the GPU starts rendering the next frame as fast as it gets it's frame from the CPU, but you could also make it start rendering at around halfway through the render pass of the other GPU. This will be a bit inconsistent for the first couple of frames that the GPUs have to render, but this will even out over time and the framerate will only be a little slower (like 90% of 'normal') of what it is now.

[Helmore]

"I currently have that Rage3d.com link open in another tab atm . I never said that that sideport was not part of the chip, just that there has been this sideport on the chip since the X1950Pro. This CrossFireX sideport is connected to the hub this time around and it apparently is a 'relatively low bandwidth interconnect'."

Do you have a link for the rv670 block diagram with crossfire side port shown on it? i can't find one. Please share your source diagram... Ditto for x1950xt... no Crossfire side port connected to the memory bus / hub seen there either.

http://www.google.ca/imgres?imgurl=h...&ct=image&cd=1


The physical connectors on the card, are not what we mean when we say Crossfire X side port... In this case it seems to be something worthy enough of note that it is now included on the block diagram for RV770, and does not appear on the Ringbus diagram for r600, nor any block diagram I can find for rv670... Notice in particular the Ring Bus diagram for R600 - this shows the ring stop for PCIe, which is now located attached to the new hub on rv770... The Crossfire X side port is also connected to this hub. So how does that equate to Cross Fire X side port being low bandwidth? (can see relatively low bandwidth as in less then GDDR5 throughput, even PCIE 2.0 x16 is relatively lower bandwidth then that, granted) As well, if it was present on R600, why is it absent from all the diagrams?

http://www.ixbt.com/video3/images/r600/diag_ringbus.png

http://hothardware.com/Articles/ATI_...rrived/?page=2

It's currently looking like you were right after all, to be honest I also hoped you were right as this is better for ATI and competition in general. I hope that this new PLX chip is just a 48 lanes PCIe gen2 chip that splits the PCIe lanes so both chips can interface with the system simultaneously and that this CrossFireX Sideport will provide an extra connection between the two GPUs directly. That would be better as this would provide lower latencies for this sideport and could also mean higher bandwidth. Here is hoping that it will smash all expectations, I just hope it will not be to much of a disappointment.

[cstkl1]

was expecting some test results from GAR this morning... concerning how CF performs on P45 (2 x 8 lanes) compared to X48 (2 x 16)...

E8500@4ghz 5x8
Biostar Tpower I45
Team Xtreem PC2-6400 CL3 4x1gb 1200mhz CL5
Sapphire HD4870 Crossfire
BFG Ageia Physx
Vista Ultimate 64bit SP1
Silverstone Op1000

Ati 8.6 Hotfix
PS i had/still have a ton of issue with p45 and crossfire
example random bsod's
NB voltage for single card is +0.1v
for both cards i need to pump up to +0.2v
rams from 2.14v for single card needed to pump to 2.25v
still not stable.. as dreamscene cannot be run in dual monitor always hangs... on stock
but on the overclock setting.. it lasted atleast 10 minutes before hang
3dmark06 default

3dmark06 1920x1200


3dmark06 1920x1200 8AA 16AS pawah



vs

Forumer Irenic from LYN with
http://forum.lowyat.net/index.php?sh...ost&p=18261421
vantage extreme

4870 cf default
x3350 default
2gb crucial ddr3 default
vista ultimate 64bit

http://i207.photobucket.com/albums/b...ageextreme.jpg

so basically there is a bottleneck... atleast for biostar tpower i45
will post the maximus 2 formula results later

[Logos]

@ cstkl1: hey thanks for that!

[XCheater]

how about some HD4850 or HD4870 in CF using P35 motherboard (16x + 4x)????

[AriciU]

how about some HD4850 or HD4870 in CF using P35 motherboard (16x + 4x)????

That would suck so bad, it's best even not to think of it.

[Helmore]

That would suck so bad, it's best even not to think of it.

And that's mainly because the 4x PCIe links are coming from the southbridge and the connection between the north and southbridge itself is barely (or not) enough to support that.

BTW, 4870 prices are still dropping over here, they are approaching €215 here (incl tax) and that's very tempting. Need to hold out until R700 launches.....

[AriciU]

And that's mainly because the 4x PCIe links are coming from the southbridge and the connection between the north and southbridge itself is barely (or not) enough to support that.

I did a test with my GTX, put it in the "supposedly" x4 slot (cpuz reported the speed as x1 for some reason) and i had like 19-20fps in GRID compared to 50-55 in the x16 slot. Huge difference.

[XCheater]

That would suck so bad, it's best even not to think of it.

And that's mainly because the 4x PCIe links are coming from the southbridge and the connection between the north and southbridge itself is barely (or not) enough to support that.

BTW, 4870 prices are still dropping over here, they are approaching €215 here (incl tax) and that's very tempting. Need to hold out until R700 launches.....

ok thanks
I wonder what motherboard I should buy for CFing since I only have 150 bucks

[AriciU]

I'm either gonna get DFI LP X38-T2R or UT X48-T3RS. Can't make my mind if i should keep my current DDR2 or switch to DDR3. If you could add another 40$ i would suggest you the DFI DK X38-T2R or just ebay for a slightly used X38 mobo.

[MnM]

was expecting some test results from GAR this morning... concerning how CF performs on P45 (2 x 8 lanes) compared to X48 (2 x 16)...

GAR had problems:

"I have since returned this board, it was the worst motherboard compatibility i have ever had in 15 years of building computers, i am not going to buy new products again, wait months for the problems to be solved.............DDR3 is problems for now, intel chipsets suck because memory controller is on the NB and not on the cpu........that is where AMD wins, and i cannot wait for the day that AMD makes a comback, hopefully its in the next year."

"I have gotten myself an ASUS Rampage Formula with some g.skill 4gb ddr2 8500 set, x48 is better for me because of the 4850 crossfire i will be using...."

All in here http://www.xtremesystems.org/forums/...=191848&page=5

cstkl1 - nice benches however those do not show the real problem with P45 chipset motherboards.

You need to run 4xAA/16xAF in games and at high res. 3DMark/Vanatage will not really show the problem.

I decide to go with X48 as opposed to P45. Just like GAR I dont want to be an unpaid BETA tester for Asus. I think WS EVO has been proven as very good motherboard so that will be my next upgrade.

I have seen this ASUS cycle so many times it makes me sick in the stomach now. With each new chipset (like P45 now) they bring out a handful of motherboards - the ones that are in the "cheaper" price brackets by Asus standards. Once those motherboard are out to the public then they use the guineapigs to do their work for them (for Asus) for free and they keep making changes in the bios, sometimes new motherboard revisions. Once the dust settles a bit and many of the issues have been solved, Asus releases the new chipset "heavyweights" at higher prices. Has anybody noticed how those "heavyweight" motherboards are way better than the original motherboard realeased? Yes they are better since they had guinespigs worldwide to beta test their products and improve them FOR FREE!
Well from now one I refuse to be guineapig anymore. WS Evo here I come.

[Logos]

@ MnM: thanks for the link. I missed that post; so it sounds like P45 doesn't always meet expectations when compared to X48. Not the first time I read about users getting huge issues with this chipset. Hard to understand the good reviews on the web though...well let's wait and see, may be a chipset revision or a better bios