vindicated when Fermi hits the market.

[Woodelf]

Nvidia is confident fermi will rockrsoxxers!!!. ahhoogah!!.

http://www.pcworld.com/article/184908/amds_lead_with_directx_11_is_insignificant_nvidia_ says.html

[uibo]

Usual shareholders talk - "It's going to be HUGE." :)

"This 60-day lag between these events of when our competition has DX11 and when we're coming to market will absolutely seem insignificant in the big picture," Hara said.
Wasn't the HD5870 released September 23, 2009 ?

[XMAN52373]

Usual shareholders talk - "It's going to be HUGE." :)


Wasn't the HD5870 released September 23, 2009 ?

Thats the day NDA lifted, it wasn't available for about a week or two after that tho cept for the few etailers who got lucky with early shipments.

[Revs]

How is DX11 giving ATI a lead at this stage anyhow? There are hardly any DX11 games out to take advantage of it.... yet.

[Sazar]

How is DX11 giving ATI a lead at this stage anyhow? There are hardly any DX11 games out to take advantage of it.... yet.

DX11 support is ahead of DX10 support at the same time.

http://en.wikipedia.org/wiki/List_of_games_with_DirectX_11_support

Dirt and Stalker are both what I would consider tier 1 titles and most of the other titles announced are also good franchises. Couple that with engines being developed with DX11 support built in, vice some tacked on crap like we saw with some titles under DX10, and you have better overall support at the same stage. At least IMHO :)

[Xion X2]

"This 60-day lag between these events of when our competition has DX11 and when we're coming to market will absolutely seem insignificant in the big picture," Hara said.

Gotta love revisionist history. 5870 has been out since September.

Let's see here... 30 (Oct), 60 (Nov), 90 (Dec), 120 (Jan), 150 (Feb)...

Given the earliest possible release date that I've heard for them (end of February,) that would put them at roughly a 150 day gap, not 60.

What an asshat.

[shadow001]

I'm playing Dirt 2 in DX11 mode and can say that that while the relative improvements over using DX10 in high quality are relatively minor,they are noticable just the same,and there is aliens vs predator being released soon,and battlefield bad company coming out in march.


But the main reason to get the upgrade is definitely the triple display option while gaming....I can't live without that anymore,even if Fermi is a faster gaming card overall,as it's basically enhancing a lot of games i already own,and the latest Catalyst 9.12 hotfix set further increases the number of supported titles.

[josiahsuarez]

But the main reason to get the upgrade is definitely the triple display option while gaming....I can't live without that anymore,even if Fermi is a faster gaming card overall,as it's basically enhancing a lot of games i already own,and the latest Catalyst 9.12 hotfix set further increases the number of supported titles.

I'd guess this will be coming with the Fermi cards. Quadro has supported this for a long time (NV calls it Quadro Plex)

http://www.nvidia.com/object/qplex_display_configurations.html

[Sazar]

I'd guess this will be coming with the Fermi cards. Quadro has supported this for a long time (NV calls it Quadro Plex)

http://www.nvidia.com/object/qplex_display_configurations.html

The Quadro Plex implementation is archaic compared to AMD's solution. If Fermi has a more elegant implementation, this will be cool.

[Xion X2]

The Quadro Plex implementation is archaic compared to AMD's solution. If Fermi has a more elegant implementation, this will be cool.

And it also needs to be designed around gaming instead of workstation use.

[shadow001]

Well,all i know is that with the HD5970's in quad crossfire,i'm largely CPU limited in pretty much every game and benchmark out there,and i'm using an i7 920 running at 4 Ghz,so it doesn't get much faster than that,while the system remains stable for every day use.


Add the DX11 support and the triple monitor gaming,and i can't really see how Nvidia can really top it,and from what we know of it so far,it would be for those that create/run programs with Cuda where the main interest is with Fermi in the end.


Here's my latest 3Dmark vantage score at the extreme settings,and the cards basically ate the bnchmark alive,as there was many parts running over 100 FPS....So much for being extreme when it's running this fast...:D


http://service.futuremark.com/compare?3dmv=1704729

[Viral]

Nv can top it by having something competitive while being single GPU. There are still benefits to having a single GPU system besides most (myself included) not really noticing the difference.

[Sazar]

Nv can top it by having something competitive while being single GPU. There are still benefits to having a single GPU system besides most (myself included) not really noticing the difference.

Yep. I really want to see what comes out because, if the cards are competitive, we will see price decreases and it's win win for the consumers.

[shadow001]

Nv can top it by having something competitive while being single GPU. There are still benefits to having a single GPU system besides most (myself included) not really noticing the difference.


True,multi-GPU setups need regular driver updates to get the best performance out of them as new games are released,and yes they'll always use more power than single GPU cards,but having this much graphics power is also hard to resist when it makes every single game or benchmark a joke to run,and i have to concentrate on getting faster CPU's for the next couple of years to uncork all the speed out of the cards,or games get way more demanding on the GPU side of things.

[shadow001]

Yep. I really want to see what comes out because, if the cards are competitive, we will see price decreases and it's win win for the consumers.



That's the main advantage of all really...Lower prices.

[Xion X2]

Shadow, you should put your CPU on water so you can go higher than 4GHz. Those chips are made to run faster than that. I've seen several guys who are at 4.4-4.5GHz which would help those 5970s.

I can go to 4GHz even on this E7400 that I'm running in my renderbox and typing this on. And that's on a 30$ air cooler.

[shadow001]

Shadow, you should put your CPU on water so you can go higher than 4GHz. Those chips are made to run faster than that. I've seen several guys who are at 4.4-4.5GHz which would help those 5970s.

I can go to 4GHz even on this E7400 that I'm running in my renderbox and typing this on. And that's on a 30$ air cooler.


It would help but think about it,a 400~500Mhz increase represents roughly 10% more performance than what i'm getting at 4Ghz,assuming whatever software that's running scales linearly with CPU clock speed,which many programs don't,and i can tell you those video cards need a hell of a lot more than that before being anywhere near making them sweat.


Put it this way,there's enough hardware resources available between those 4 GPU's in terms of shaders,texturing and fillrate to be equal to 8 RV770 GPU's,wich are used in the HD4870 cards,and a single HD4870 is still a very decent gaming card overall,now imagine that 8 times over in hardware resources....It's crazy.


What i need is nothing less than a 6 core/12 thread gulftown CPU clocking at 4 Ghz or better,and with improvements made to that processor to overall efficiency,i'd be looking at something potentially 50% faster than what i'm using right now CPU wise,so instead of spending 500~600$ on a good water cooling setup to get a relatively small boost with my current processor,save that cash for that Gulftown CPU instead,which it seems will run on current X58 boards just fine with a bios update....The only drawback is that it seems it'll only be released in 3~4 months from now.

[Xion X2]

It would help but think about it,a 400~500Mhz increase represents roughly 10% more performance than what i'm getting at 4Ghz,assuming whatever software that's running scales linearly with CPU clock speed,which many programs don't,and i can tell you those video cards need a hell of a lot more than that before being anywhere near making them sweat.


Put it this way,there's enough hardware resources available between those 4 GPU's in terms of shaders,texturing and fillrate to be equal to 8 RV770 GPU's,wich are used in the HD4870 cards,and a single HD4870 is still a very decent gaming card overall,now imagine that 8 times over in hardware resources....It's crazy.


What i need is nothing less than a 6 core/12 thread gulftown CPU clocking at 4 Ghz or better,and with improvements made to that processor to overall efficiency,i'd be looking at something potentially 50% faster than what i'm using right now CPU wise,so instead of spending 500~600$ on a good water cooling setup to get a relatively small boost with my current processor,save that cash for that Gulftown CPU instead,which it seems will run on current X58 boards just fine with a bios update....The only drawback is that it seems it'll only be released in 3~4 months from now.

More cores aren't going to help you now, I don't think. The app that you're running must be written for it, and there are still many apps around, especially games, that only scale across 2-4 cores. Some on just one core.

What you need right now is speed.. faster speed that can keep up with all the data that needs to be fed to all 4 GPUs. Or better efficiency, as you mentioned. OC'ng your CPU another 500MHz may net you only 10% gains or it may net you a lot more. Difficult to tell. It would depend on the stress that the GPUs are under, and at 5760x or whatever resolution you plan to run for Eyefinity, you're going to hit some GPU-limited scenarios even on CrossfireX. It's just too high a res not to. I'll bet you that games like Crysis or Clear Sky (or Aliens vs. Predator when it hits shelves in February) will push it. And when that happens, you'll need to feed data to the GPUs as fast as possible.

Edit: It looks as if Gulftown may be a significant upgrade after all. DailyTech is reporting that Intel states a 22% efficiency increase:

Intel is reporting at least a 22 percent performance increase clock for clock over their 45nm process
http://www.dailytech.com/Gulftown+is+the+Flagship+of+32nm+Westmere+Line/article14227.htm

So you make a good point about waiting for Gulftown. Still, regardless of what CPU you have, I would strongly consider H20 to get maximum performance out of your system. It wouldn't take 5-600$ for a good water setup just for your CPU. They sell nice kits for only 250$.

This one here would do a great job for you. Has one of the best CPU blocks on the market (Apogee XT) and a nice 220 Swiftech radiator for $259:

http://jab-tech.com/Swiftech-H20-22-Ultima-XT-Liquid-Cooling-Kit-pr-4324.html

If you have any hardware skills at all and would feel comfortable enough installing one, I think that it would be a wise investment for that extreme setup of yours. As long as you are careful, they are quite easy to install and maintain.

If interested, read up on the forums over at XtremeSystems. They have hands-down the best liquid cooling forum on the web:
http://www.xtremesystems.org/forums/forumdisplay.php?f=70

[shadow001]

More cores aren't going to help you now, I don't think. The app that you're running must be written for it, and there are still many apps around, especially games, that only scale across 2-4 cores. Some on just one core.

What you need right now is speed.. faster speed that can keep up with all the data that needs to be fed to all 4 GPUs. Or better efficiency, as you mentioned. OC'ng your CPU another 500MHz may net you only 10% gains or it may net you a lot more. Difficult to tell. It would depend on the stress that the GPUs are under, and at 5760x or whatever resolution you plan to run for Eyefinity, you're going to hit some GPU-limited scenarios even on CrossfireX. It's just too high a res not to. I'll bet you that games like Crysis or Clear Sky (or Aliens vs. Predator when it hits shelves in February) will push it. And when that happens, you'll need to feed data to the GPUs as fast as possible.

Edit: It looks as if Gulftown may be a significant upgrade after all. DailyTech is reporting that Intel states a 22% efficiency increase:


http://www.dailytech.com/Gulftown+is+the+Flagship+of+32nm+Westmere+Line/article14227.htm

So you make a good point about waiting for Gulftown. Still, regardless of what CPU you have, I would strongly consider H20 to get maximum performance out of your system. It wouldn't take 5-600$ for a good water setup just for your CPU. They sell nice kits for only 250$.

This one here would do a great job for you. Has one of the best CPU blocks on the market (Apogee XT) and a nice 220 Swiftech radiator for $259:

http://jab-tech.com/Swiftech-H20-22-Ultima-XT-Liquid-Cooling-Kit-pr-4324.html

If you have any hardware skills at all and would feel comfortable enough installing one, I think that it would be a wise investment for that extreme setup of yours. As long as you are careful, they are quite easy to install and maintain.

If interested, read up on the forums over at XtremeSystems. They have hands-down the best liquid cooling forum on the web:
http://www.xtremesystems.org/forums/forumdisplay.php?f=70



Well to put it into perspective for you,i did a 3Dmark vantage run at the extreme settings,both with the cards running at stock clocks and running at 850/1200 Mhz,so that's a 125Mhz increase on each GPU and a 200Mhz(800Mhz effective) for the GDDR5,and the overall score only went up 900 points.


Those are fairly significant increases in clocks,which gives a nice kick upwards to things like pixel shading power,texturing rate,fill rate and memory bandwith,yet the overall score increased less than 5% over running the cards at stock values,with many parts of the benchmark running at the same FPS at either clock settings,and where it did increase,did so to small single digit amounts,so the cards are being held back most of the time by something else,regardless of the actual clocks they operate at.



But in the end you're right,it's all about the software and leveraging all those threads effectively,but that's still some time away,as most games are only using 2 cores right now,save for supreme commander or GTA 4 games,so current higher end CPU's are sorely being underused as it is,and it'll take a long time before there are actual games really leveraging current 4 core/8 thread or worse yet,future 6 core/12 thread CPU's anytime soon,but video cards keep increasing in performance at a lot faster rate compared to CPU's.



Given that rather bleak scenario in terms of gaming,it doesn't surprise me why Nvidia is making a lot more noise for GP-GPU usage with Fermi,rather than gaming scenarios,as the latter is hoplessly behind the curve with regards to cutting edge video cards in terms of feature usage.

[Xion X2]

^ Yeah, your above scenario is why I think you need a faster CPU. I think it's probably being bogged down by all the graphics power.

Every time that I upgrade, I try to do a balance of CPU/Grafx. 5970 is so strong that it's tough to do that with, but between running at a higher res w/ 3 monitors and OC'ng the CPU as much as possible, I believe it can be done somewhat.

Just my theory.

[shadow001]

^ Yeah, your above scenario is why I think you need a faster CPU. I think it's probably being bogged down by all the graphics power.

Every time that I upgrade, I try to do a balance of CPU/Grafx. 5970 is so strong that it's tough to do that with, but between running at a higher res w/ 3 monitors and OC'ng the CPU as much as possible, I believe it can be done somewhat.

Just my theory.



Well,up to this point,and if you exclude Multi-GPU setups,it's still possible to push settings and resolutions on the latest games high enough,that even with the highest end single GPU video cards,that the main bottleneck is still the video card in many situations,but multi GPU setups give us a preview of what a single GPU video card might perform like within the next 24 months,because that's how fast GPU's are still evolving,pace wise.


They're gaining in speed and features at a much faster pace than CPU's,so as impressive as a pair of these HD5970's are in quad crossfire,a single GPU card will match or beat this performance within 2 years,since there's still the 32nm,28nm,22nm,18nm and even 11nm fabrication processes to go thru before current manufacturing techhologies for computer chips really hit major technological hurdles to overcome,where even silicon hits it's practical limits,and a new material has to be used.


We'll be lucky if in that same 2 year period,CPU's become twice as fast as they are now,and imagine the crazy amount of power 4 GPU's,using the 22nm process will be in 2 years time,as each GPU may be packing over 5 billion transistors a piece,and it won't matter what CPU you're using,it'll never keep up anyhow,no matter how overclocked it is.

[XMAN52373]

Shawdow, Xion is right, with very few exceptions, there isn't many apps, including games that make proper use of Multi core setups. WoW requires some registry/ini/inf files hacks to get it to work, FarCry 2 has support out of the box, any other outside of that I have no idea. You would think that with almost every computer being sold today being atleast dou core setups, atleast game devs would put more effort into making games make use of that fact.

[Redeemed]

Pour nitrous oxide onto your processor while the system is running.

That'll make it fast. :lol:

[shadow001]

Shawdow, Xion is right, with very few exceptions, there isn't many apps, including games that make proper use of Multi core setups. WoW requires some registry/ini/inf files hacks to get it to work, FarCry 2 has support out of the box, any other outside of that I have no idea. You would think that with almost every computer being sold today being atleast dou core setups, atleast game devs would put more effort into making games make use of that fact.



The the professional market,such as photoshop,or 3D studio max,or lightwave or database/webserver software and more,there is support for multithreading for years now,since at the software level,there isn't any difference if the CPU cores are built into the same die,or each core has it's own CPU package,but in most cases,it's software that costs thousands of dollars.


The company i work for as an example uses a program that called Catia,which is used extensively in engineering environments and the last time i checked,that software package costs 100 000$,so that's not exactly chump change,and benefits from multiprocessor systems in a big way.....


Desktop software for the average user though,multhreading is only starting it's first steps these last couple of years,and games do have some,but it's mostly because the vast majority are console ports,which the current X360 and PS3 have multicore CPU's in them,otherwise,i'd doubt we'd have any.



Pour nitrous oxide onto your processor while the system is running.

That'll make it fast.


I'd also go broke too,since it isn't exactly cheap and evaporates pretty fast once in use,as you have to fill the LN2 pot on top of the CPU every 30 secs or so...Not practical to game while doing that att the same time :D

[mailman2]

I see the FX 5800 all over again...

Get ready for FERMI -
http://www.nvnews.net/vbulletin/showthread.php?t=6749