ATi 4870 X2 (R700) thread

[Hauk]

Originally posted by: dasracht

Originally posted by: Janooo
A couple of new pictures here and here.

I'm impressed. ATI cards have looked really low-quality to me, cheesy plastic looking (if that makes sense). This card looks great.

It does makes sense. The red translucent coolers look cheap IMO. This card looks tough!!
I hope this is what they've chosen for final product!

 

[Munky]

Originally posted by: Sentry2

2x1GB is bad news. 1x2GB is good news.

It's still better than 2x512MB....

That doesn't prove or disprove anything as far as shared memory goes. How do you think the memory system works in a multi-socket Opteron system? Each cpu has its own memory, and yet each can access the other cpu's memory via HT, it's not like all data has do be duplicated between the two.

 

[StevenNevets]

I don't have time to read all this and even if I did I'm too dumb.

Would a X2 6000 and PCI-x16 hold this card back a lot or is it still a good purchase?

 

[dasracht]

Has there been any new word or rumor on the release date? I keep hearing both August and 4th week of July, and toms had that news article the other day saying July (this one). Anything new?

 

[ronnn]

Originally posted by: SteelSix

Originally posted by: Janooo
A couple of new pictures here and here.

Sweet! Very industrial looking..

That does look industrial! Personally I was hoping for butterflies.

 

[ronnn]

Originally posted by: SteelSix

Originally posted by: AmberClad

Originally posted by: dadach

Originally posted by: taltamir
yea... i wonder if the promise of solved microstutter was also a lie.

can someone of the moderators please do something about this troll?

Wtf? A noob calling a long-time member a troll? I've seen Talt be just as critical of NV, so take your agenda elsewhere.

:thumbsup:

Trolls usually don't have a camp, just a distorted agenda for attention. :laugh:

 

[Munky]

Originally posted by: StevenNevets
I don't have time to read all this and even if I did I'm too dumb.

Would a X2 6000 and PCI-x16 hold this card back a lot or is it still a good purchase?

Yeah, I would say you'd be heavily cpu limited. In your case I'd just go with a single 4870.

 

[BFG10K]

Janooo:

So what is the reason for the micro stutter?

An irregular spread of frames. The cause appears to be the timing relationship between the first GPU's frame and the second one's; the second comes much sooner in relation to first.

 

[bryanW1995]

Originally posted by: StevenNevets
I don't have time to read all this and even if I did I'm too dumb.

Would a X2 6000 and PCI-x16 hold this card back a lot or is it still a good purchase?

maybe. of course, if you read the entire thread, then you'd know for sure...

 

[tuteja1986]

Micro shutter 101 : http://www.hardforum.com/showthread.php?t=1317582

only real solution i see is to get rid of ARP :! which will never happen :!

 

[betasub]

Originally posted by: AmberClad
Wtf? A noob calling a long-time member a troll? I've seen Talt be just as critical of NV, so take your agenda elsewhere.

Troll != fanboi

Taltamir, who are you accusing of making promises and lying? Or is it just the usual gossip, rumour and heightened expectations?

 

[toslat]

Originally posted by: tuteja1986
Micro shutter 101 : http://www.hardforum.com/showthread.php?t=1317582

only real solution i see is to get rid of ARP :! which will never happen :!

Please correct me if am wrong, but seems micro shutter basically refers to variable frame 'period' (or frame rate) and is exaggerated in multiGPU setups if the frame ready offsets are poorly aligned (e.g. not centered in a 2 GPU case). e.g. using 30FPS

Frame.............Time (ms) when frame is ready
....................Ideal.....................with Micro Shutter (10ms offset)
... 1 ............... 0.0............................0.0
... 2 ............. 33.3 ......................... 10.0
... 3 ............. 66.6 ......................... 66.6
... 4 ............. 99.9 ......................... 76.6
.. 5 ............ 133.3 ....................... 133.3

I don't think you need to get rid of AFR (dont know what you meant by ARP) to fix this as the offset problem can be ameliorated by introducing a delay/buffer in the path of the GPUs. The delay would have to be adaptive in a practical scenario as the correct value depends on FPS and all you really want to get rid of is the bias

 

[CP5670]

Originally posted by: Janooo

Originally posted by: SteelSix
Microstuttering FAQ

"Do all Multi GPU rigs experience microstuttering?
It looks like it, yes. The extent to which might be different on one system than on another, they might even be different every time you start your machine or run the game engine.

If I notice microstuttering, can I minimize/eliminate it?
Yes. By running the game at a setting where your graphics cards are able to output more than the monitors refresh rate (that is, the maximum FPS the monitors are capible of; the pixels on your screen can only change so fast) microstuttering is eliminated completely. Most monitors have a refresh rate of 60 or 70Hz, meaning you would need 70 or 80 FPS to eliminate microstuttering.."

Thank you.
What's the relationship between min fps and micro stutter?

In general, bigger framerate swings result in a more pronounced microstutter. This is a frequent issue in itself on multi GPU setups, so it just amplifies the effect even more.

Also, I think that "minimize/eliminate it" part only applies if vsync is enabled (without triple buffering, which is apparently incompatible with AFR). The refresh rate wouldn't affect this otherwise.

I don't think you need to get rid of AFR (dont know what you meant by ARP) to fix this as the offset problem can be ameliorated by introducing a delay/buffer in the path of the GPUs. The delay would have to be adaptive in a practical scenario as the correct value depends on FPS and all you really want to get rid of is the bias

The problem with this is it would introduce additional input lag, something that is already an issue due to the variable frame delays.

 

[sxr7171]

Originally posted by: SteelSix

Originally posted by: Janooo

Originally posted by: CP5670
I don't see how that can be solved as long as it operates on Crossfire. That problem is inherent to AFR.

So what is the reason for the micro stutter? Honest question, I don't know myself.

Microstuttering FAQ

"Do all Multi GPU rigs experience microstuttering?
It looks like it, yes. The extent to which might be different on one system than on another, they might even be different every time you start your machine or run the game engine.

If I notice microstuttering, can I minimize/eliminate it?
Yes. By running the game at a setting where your graphics cards are able to output more than the monitors refresh rate (that is, the maximum FPS the monitors are capible of; the pixels on your screen can only change so fast) microstuttering is eliminated completely. Most monitors have a refresh rate of 60 or 70Hz, meaning you would need 70 or 80 FPS to eliminate microstuttering.."

There is now way that in Crysis 1920x1200 on very high that a CF or SLI setup will take me from below 60FPS to above that. I guess I'll ride these price drops on single GTX solutions, unless ATI can show a 4870 x2 without microstuttering.

 

[toslat]

Originally posted by: CP5670
The problem with this is it would introduce additional input lag, something that is already an issue due to the variable frame delays.

I dont see it introducing additional input lag, as only one of the GPUs will actually be delayed at any time, and the other GPU output is still delivered as scheduled. Probably because I said in the path of both GPUs, but that was more towards being able to adjust the delay on either GPU and not both simultaneously (bias is one-way). Looking at the example I posted previously, only the even frame GPU would have been delayed by 23.3ms and it would result in perfect 30FPS

 

[Hauk]

Originally posted by: sxr7171

Originally posted by: SteelSix

Originally posted by: Janooo

Originally posted by: CP5670
I don't see how that can be solved as long as it operates on Crossfire. That problem is inherent to AFR.

So what is the reason for the micro stutter? Honest question, I don't know myself.

Microstuttering FAQ

"Do all Multi GPU rigs experience microstuttering?
It looks like it, yes. The extent to which might be different on one system than on another, they might even be different every time you start your machine or run the game engine.

If I notice microstuttering, can I minimize/eliminate it?
Yes. By running the game at a setting where your graphics cards are able to output more than the monitors refresh rate (that is, the maximum FPS the monitors are capible of; the pixels on your screen can only change so fast) microstuttering is eliminated completely. Most monitors have a refresh rate of 60 or 70Hz, meaning you would need 70 or 80 FPS to eliminate microstuttering.."

There is now way that in Crysis 1920x1200 on very high that a CF or SLI setup will take me from below 60FPS to above that. I guess I'll ride these price drops on single GTX solutions, unless ATI can show a 4870 x2 without microstuttering.

1920x1200 on very high? No offense to you, many people make similar statements; but I wish Crysis never existed. Current high end configs allow for max detail settings at that resolution on almost all games out there, except for Crysis of course.

I hope there's a day when references to that bloated piece of crap are laughed at..

 

[Janooo]

If multi-GPU solution doubles min fps does it mean that the game play is smoother? Would micro stuttering still be an issue here?

 

[Elcs]

Has anyone tried to remove the plastic covering over the heatsink like some people have done with the 8800 series?

Is it possible to remove the cover and are there any gains to doing so? To be honest, in my theoretical and unscientific opinion, it should be a positive thing especially when couple with a quiet 120mm fan blowing from front to back. Could even unplug the blower fan.

 

[taltamir]

Thanks bryanW1995, AmberClad, SteelSix

Originally posted by: SteelSix
:thumbsup: Let's not get wound up, it's so unimportant. Talt is okay. He's a fair weather fan..

Words to live by.

Who was I calling a liar? well the rumor mill has been abuzz about this card, some sites specifically said that AMD told that that it will have solved the wasted ram issue AND the microstutter issue. So either those sites are lying about AMD telling them that, or AMD lied to them.

Before we "knew" the following things:
- It will feature 1 GB of ram shared between both cores. (no waste)
- It will solve micro stutter.
- It will be 15% faster then 2 units of 4870 in CF
- It will cost 500$ (vs 300$ per 4870).

In a way it all made sense, there were AMD graphs showing the internal construction of the chip, including a special crossfire port that would be used in an x2 card, connecting the two chips directly. The PLX was explained away as a method to communicating with the CPU only, that the GPUs will communicate with each other entirely through the shared ram and the special port on the chip.
The price could be explained away by not wasting ram, by using reusing the same PCB and power circuitry, etc... it actually makes sense, since eachchip should cost less then 200$, they could actually make more money, yet still make it a more attractive choice.
EDIT: Not to mention the 4870 is a 4850 with the ram replaced from GDDR3 to GDDR5... so upgrading 512MB from 3 to 5 is 100$ extra? how much extra is 1.5GB going to be?

Then we suddenly find reports disproving the first claim. It will be 2x1GB, not shared... What is this? I remember that supposed "statement from ATI" said that the GPUs will communicate via shared memory and there will be no memory waste (which contradicted what anand says about the special CF port, although, wouldn't that be needed for quad crossfire?).
Will micro stutter really be solved?
Will it really cost 500$? (remember, they have to up the GDDR5 ram from 512MB to 2GB for this card, that is 4 times the ram, that is in very VERY short supply and is highly expensive!).
Will it really be 15% faster then two individual 4870 cards in CF (in the past, two individual cards always outperformed the sandwich package)...

I will go as far as saying someone has been lying... the question is:
1. Did AMD spread lies?
2. Did a review site make up lies, claimed to have received it's info direct from AMD, and was quoted by every other site on the planet as fact? (news outlets tend to steal their news from other news outlet).
3. Did a misinformed AMD rep accidentally mislead some news site with no direction from above or intent?

I was holding my breath for this card, since it was shaping up to be an even better bang/$ then the 4850, which I currently use because it is THE BEST bang per buck card out there. I was expecting to pass it along to my brother and getting a 4870x2 when it is out. Now I am not so sure this will happen, it seems the card was exaggerated. I might still get it, but right now it is in a definite "maybe".

I can assure you, however, that my intent was not to inflame, cause derision, or thread crap.

 

[nRollo]

Originally posted by: bryanW1995

Originally posted by: chizow

Originally posted by: Compddd
I really hope this thing blows everyone and everything away, just so Chizow will hush up.

And if it doesn't, will you just crawl back under your bridge until ATI rises again? Nothing wrong with hoping I suppose, but based on 4870CF and GTX 280 SLI I don't think we'll see massive gains with the 4870X2 over a single 4870 or even 4850CF.

wow, you're my nomination for "nzone moderator praetor exemplar". how badly does nvidia have to get their ass kicked for you to shut up?

Hmm.

I don't see NVIDIA's ass as kicked.

Top single GPU? Check

Competitive at most price points? Check.

Top Multi GPU? Check.

Value? Check.

With GTX260s at $260 and GTX280s at $460, any excitement ATi had going for it is largely over.

The world won't balk at paying $460 for the best overall GPU, and $260 for the GTX260 vs $309 for the 4870 is a flip a coin kind of decision.


What do you mean "NVIDIA's ass is kicked, badly?"

Where I'm sitting I doubt they'll even lose much market share with current bang for buck.

 

[nRollo]

Originally posted by: taltamir
Thanks bryanW1995, AmberClad, SteelSix

Originally posted by: SteelSix
:thumbsup: Let's not get wound up, it's so unimportant. Talt is okay. He's a fair weather fan..

Words to live by.

Who was I calling a liar? well the rumor mill has been abuzz about this card, some sites specifically said that AMD told that that it will have solved the wasted ram issue AND the microstutter issue. So either those sites are lying about AMD telling them that, or AMD lied to them.

Before we "knew" the following things:
- It will feature 1 GB of ram shared between both cores. (no waste)
- It will solve micro stutter.
- It will be 15% faster then 2 units of 4870 in CF
- It will cost 500$ (vs 300$ per 4870).

In a way it all made sense, there were AMD graphs showing the internal construction of the chip, including a special crossfire port that would be used in an x2 card, connecting the two chips directly. The PLX was explained away as a method to communicating with the CPU only, that the GPUs will communicate with each other entirely through the shared ram and the special port on the chip.
The price could be explained away by not wasting ram, by using reusing the same PCB and power circuitry, etc... it actually makes sense, since eachchip should cost less then 200$, they could actually make more money, yet still make it a more attractive choice.
EDIT: Not to mention the 4870 is a 4850 with the ram replaced from GDDR3 to GDDR5... so upgrading 512MB from 3 to 5 is 100$ extra? how much extra is 1.5GB going to be?

Then we suddenly find reports disproving the first claim. It will be 2x1GB, not shared... What is this? I remember that supposed "statement from ATI" said that the GPUs will communicate via shared memory and there will be no memory waste (which contradicted what anand says about the special CF port, although, wouldn't that be needed for quad crossfire?).
Will micro stutter really be solved?
Will it really cost 500$? (remember, they have to up the GDDR5 ram from 512MB to 2GB for this card, that is 4 times the ram, that is in very VERY short supply and is highly expensive!).
Will it really be 15% faster then two individual 4870 cards in CF (in the past, two individual cards always outperformed the sandwich package)...

I will go as far as saying someone has been lying... the question is:
1. Did AMD spread lies?
2. Did a review site make up lies, claimed to have received it's info direct from AMD, and was quoted by every other site on the planet as fact? (news outlets tend to steal their news from other news outlet).
3. Did a misinformed AMD rep accidentally mislead some news site with no direction from above or intent?

I was holding my breath for this card, since it was shaping up to be an even better bang/$ then the 4850, which I currently use because it is THE BEST bang per buck card out there. I was expecting to pass it along to my brother and getting a 4870x2 when it is out. Now I am not so sure this will happen, it seems the card was exaggerated. I might still get it, but right now it is in a definite "maybe".

I can assure you, however, that my intent was not to inflame, cause derision, or thread crap.

Anyway you cut it, this card will be the fastest single slot card at launch, probably by a good margin, and that has value itself. Microstutter, RAM use, etc can't change that.

People with single slot motherboards need options for performance as well.

 

[AmberClad]

Originally posted by: nRollo
With GTX260s at $260 and GTX280s at $460, any excitement ATi had going for it is largely over.

If you're talking about that MSI firesale me and a couple of other ATers discussed at length a few days ago, both of them went back up in price.

Maybe some of the other manufacturers will follow suit -- those prices seemed to be pretty attractive.

 

[nRollo]

Originally posted by: AmberClad

Originally posted by: nRollo
With GTX260s at $260 and GTX280s at $460, any excitement ATi had going for it is largely over.

If you're talking about that MSI firesale me and a couple of other ATers discussed at length a few days ago, both of them went back up in price.

Maybe some of the other manufacturers will follow suit -- those prices seemed to be pretty attractive.

Those prices are still in stock/effect at ZZF and newegg.

Competition brought better deals for all.

 

[Hauk]

Originally posted by: Elcs
Has anyone tried to remove the plastic covering over the heatsink like some people have done with the 8800 series?

Is it possible to remove the cover and are there any gains to doing so? To be honest, in my theoretical and unscientific opinion, it should be a positive thing especially when couple with a quiet 120mm fan blowing from front to back. Could even unplug the blower fan.

This was done with the 8800 GT single slot cooler with some success. Helps a little, but fin surface area is king.

 

[AmberClad]

Originally posted by: nRollo
Those prices are still in stock/effect at ZZF and newegg.

Competition brought better deals for all.

Oh, they're at NewEgg. I only thought to check ZZF, and the prices were either jacked back up or OOS.

On a sidenote - Kind of interesting trend I noted a while ago is that the NV-exclusive companies don't seem as willing to offer the crazy-hot deals that Palit and the Taiwanese-motherboard-manufacturers are giving out...