nordichardware: GTX 590 in February

[kdubbs]

I feel like this has been discussed before, but what is to prevent nVidia from sticking two full-on, cherry picked, overclocked 580 GPUs on a board, loading it up w/6 GB of memory and throwing this card in a separate, external enclosure, complete w/ its own 600W power supply. This thing would likely crush everything else. They could thumb their noses at the 300W pcie spec, claim absolute supremacy, charge $1200 bucks for it and call it a day? Obviously, they wouldn’t sell too many of these, but the halo effect would be there and the PR people could have a field day. Assuming there wasn’t some latency issue with the external connection (which presumably would just be a PCI-E extender) this would have seem to have tons of advantages. Don’t need to upgrade your PSU, don’t have to come up w/ exotic cooling solutions, just plug it in and get world-beating performance. AMD could do the same w/ a 1900-shader Cypress (assuming such a thing exists). The ViDock works with laptops, why not make a desktop version?

 

[Grooveriding]

I feel like this has been discussed before, but what is to prevent nVidia from sticking two full-on, cherry picked, overclocked 580 GPUs on a board, loading it up w/6 GB of memory and throwing this card in a separate, external enclosure, complete w/ its own 600W power supply. This thing would likely crush everything else. They could thumb their noses at the 300W pcie spec, claim absolute supremacy, charge $1200 bucks for it and call it a day? Obviously, they wouldn’t sell too many of these, but the halo effect would be there and the PR people could have a field day. Assuming there wasn’t some latency issue with the external connection (which presumably would just be a PCI-E extender) this would have seem to have tons of advantages. Don’t need to upgrade your PSU, don’t have to come up w/ exotic cooling solutions, just plug it in and get world-beating performance. AMD could do the same w/ a 1900-shader Cypress (assuming such a thing exists). The ViDock works with laptops, why not make a desktop version?

Sanity, I bet.

 

[Borealis7]

there have been attempts at external graphics cards, but none really caught on.
but it is possible, although i wouldnt run such a thing unless it was snowing inside the room

 

[Phynaz]

charge $1200 bucks for it and call it a day?

They would have to sell it for $5,000 to cover the engineering costs.

 

[AtenRa]

Dual-GPU GTX 590 and new Titanium GTX 550 Ti in new R266.77

http://www.geeks3d.com/20110217/dual-gpu-gtx-590-and-new-titanium-gtx-550-ti-in-new-r266-77/

 

[happy medium]

Dual-GPU GTX 590 and new Titanium GTX 550 Ti in new R266.77

http://www.geeks3d.com/20110217/dual-gpu-gtx-590-and-new-titanium-gtx-550-ti-in-new-r266-77/

Those articles don't matter , Nvidia has to officially release the card first, then they will become avalable, if you believe some people in this thread.
Just like they officially released the gtx580. There was actually a guy in this thread willing to bet 100$ that the 6970 would be faster then the gtx580. Well he was WRONG then too.

I said about a month ago, expect a gtx590 (or some dual Fermil card) around the 23rd of Feb.

Good news about the gtx550, another card some said would not exist.

Looks like I'll be 4 for 5 soon.
http://forums.anandtech.com/showthread.php?t=2126529

 

[Phynaz]

Looks like I'll be 4 for 5 soon.

I'm betting you probably don't want to go there.

 

[tincart]

Looks like I'll be 4 for 5 soon.

Any news on when those dual 460 and dual 480 boards are coming out?

 

[SolMiester]

Why do people keep bring up the max wattage spec's of PCIe 2.0!! It is not our job to worry if it comes under or over, we arent manufacturing it and not responsible for PCIe standards.
If they can build it, yay!

 

[Lonyo]

Those articles don't matter , Nvidia has to officially release the card first, then they will become avalable, if you believe some people in this thread.

People have pointed out ENDLESS times how many rumours and preview cards and samples and news articles there have been on dual GPU Fermi cards. How many dual GPU Fermi cards have been for sale? None. At all. Ever. Despite all the rumours. Anyone who blindly believes that it's coming because there are rumours is, quite honestly, something I would get a warning for saying.

Unless of course you are happy with your dual GPU card using dual GTX480/470/460 (pick your rumour of choice) GPUs.

Now that there's a more "official" leak of sorts, it becomes somewhat more verified, but all those many people who have held off and won't believe it until they see something official might be considering the endless other rumours which never came true about dual GPU cards (including even non-Fermi cards from both NV and also ATI/AMD).


Trying to act superior because the dual GTX590 has become more likely with this reference is stupid, same as assuming it will come based on all the rumours up to this point is stupid.

For Gen 1 Fermi:
December 2009: http://www.xtremesystems.org/forums/showthread.php?t=240835
January 2010: http://news.softpedia.com/news/NVID...n-April-Mainstream-Cards-in-June-132488.shtml
March 2010: http://www.dvhardware.net/article41563.html
May 2010: http://www.maximumpc.com/article/news/rumor_nvidia_dub_dualgpu_fermi_card_gtx_490
June 2010: http://www.hardwarecanucks.com/news/video/galaxy-impress-computex-dualgpu-fermi/ (ACTUAL CARD SHOWN)
July 2010: http://www.tomshardware.co.uk/forum/293064-15-dual-fermi-gf104-chips
October 2010: http://www.geeks3d.com/20101026/zotac-geforce-gtx-460-dual-gpu-pcb-gtx-495/ (ACTUAL CARD SHOWN)

Gen 2 Fermi:
November 2010: http://www.pclaunches.com/graphic_c...rmi_architecture_based_video_card_to_2011.php
January 2011: http://www.zdnet.com/blog/computers...x-595-dual-gpu-fermi-based-graphics-card/4781 (ACTUAL CARD SHOWN)

Forgive any cynicism.

 

[Idontcare]

Why do people keep bring up the max wattage spec's of PCIe 2.0!! It is not our job to worry if it comes under or over, we arent manufacturing it and not responsible for PCIe standards.
If they can build it, yay!

It matters in that it places a somewhat artificial barrier or cap on the performance expectations we can have for retail stock-clocked products.

Knowing this boundary condition exists enables us to speak to the ramifications, artificial or practical, of the boundary condition.

No different than how conversations go down in the memory & storage forum whenever hard-drives get close to the upper-limit of the max SATA transfer rate of the time. Until SATA III came out we regularly discussed the relevance of gen N+1 SSD's that would/could have transfer rates in excess of 300MB/s.

My own ignorance speaking here, but what is the reasoning behind the spec calling for a power-limit in the first place? I understand the 75W limit for pulling power through the PCIe interface but why does the spec care how many power connectors you have on your GPU and how much power the sucker uses from the PSU?

 

[Phynaz]

My own ignorance speaking here, but what is the reasoning behind the spec calling for a power-limit in the first place? I understand the 75W limit for pulling power through the PCIe interface but why does the spec care how many power connectors you have on your GPU and how much power the sucker uses from the PSU?

You've read my mind. The old VooDoo cards with the external power supplies must have been out of spec also. Why weren't they an issue then?

Off to study...

 

[notty22]

Looks like evga pulled that thread and the drivers, maybe let the cat out of the bag a little early ! Always like a nice mystery

 

[happy medium]

Any news on when those dual 460 and dual 480 boards are coming out?

You didn't hear, those prototypes were cancelled for the gtx590 and dual Evga gtx560 card. Why rush them, Nvidia has the fastest card, the gtx580, unless you count the 900$ 5970 on newegg.

 

[badb0y]

You didn't hear, those prototypes were cancelled for the gtx590 and dual Evga gtx560 card. Why rush them, Nvidia has the fastest card, the gtx580, unless you count the 900$ 5970 on newegg.

Didn't you hear? AMD to release Antilles by Q1 2011. Why waste time making outdated 5970s? ;P .

 

[happy medium]

Didn't you hear? AMD to release Antilles by Q1 2011. Why waste time making outdated 5970s? ;P .

Must have missed that rumour. Is that gonna be before or after Dell buys them?

 

[badb0y]

Must have missed that rumour. Is that gonna be before or after Dell buys them?

Nah, Intel pulled out the checkbook so AMD is still a sinking ship .

 

[Phynaz]

My own ignorance speaking here, but what is the reasoning behind the spec calling for a power-limit in the first place? I understand the 75W limit for pulling power through the PCIe interface but why does the spec care how many power connectors you have on your GPU and how much power the sucker uses from the PSU?

Did a little research and found the answer. There appears to be more than a few working groups that all contribute to the spec. One of these groups is Electro-Mechanical. It's this group that lays out the connectors physically and electrically.

Now that I think about it, it makes sense. Those 6 and 8 pin connectors are all the same. We don't have to worry about the way they are wired or if they will carry enough power. Card has a 6 pin connector? Plug in the 6 pin connector and it will (should) work. Could you imagine if every vendor designed their own physical power interface? How about the electrical interface, such as the power-up sequence and timing.

Or to your point, if the power supplied to / consumed by those connectors isn't standardized then it just turns into a crapshoot doesn't it.

Slide deck here.

 

[SHAQ]

It matters in that it places a somewhat artificial barrier or cap on the performance expectations we can have for retail stock-clocked products.

Knowing this boundary condition exists enables us to speak to the ramifications, artificial or practical, of the boundary condition.

No different than how conversations go down in the memory & storage forum whenever hard-drives get close to the upper-limit of the max SATA transfer rate of the time. Until SATA III came out we regularly discussed the relevance of gen N+1 SSD's that would/could have transfer rates in excess of 300MB/s.

My own ignorance speaking here, but what is the reasoning behind the spec calling for a power-limit in the first place? I understand the 75W limit for pulling power through the PCIe interface but why does the spec care how many power connectors you have on your GPU and how much power the sucker uses from the PSU?

I agree. 375 watts makes more sense for a limit anyhow, but I imagine air cooling will limit it at about that spec as well.

 

[Castiel]

And so it begins.. Looking to be fake

 

[Skurge]

And so it begins.. Looking to be fake

You didn't make that yourself did you?

J/k

 

[Arkadrel]

Those articles don't matter , Nvidia has to officially release the card first, then they will become avalable, if you believe some people in this thread.
Just like they officially released the gtx580. There was actually a guy in this thread willing to bet 100$ that the 6970 would be faster then the gtx580. Well he was WRONG then too.

I said about a month ago, expect a gtx590 (or some dual Fermil card) around the 23rd of Feb.

Good news about the gtx550, another card some said would not exist.

Looks like I'll be 4 for 5 soon.
http://forums.anandtech.com/showthread.php?t=2126529

Gonna be fun to see if in 5days time or so, there ll be a 590 out....

*if* a 580 costs around 500$+, will a 590 be 1,000$+ ?

Because to me, a 1,000$+ card just seems abit.. overboard (reguardless of brand).

I remember seeing dual 480, dual 460 rumors... and yet where are they?
Does nvidia still make dual gpu cards?

 

[tincart]

You didn't hear, those prototypes were cancelled for the gtx590 and dual Evga gtx560 card. Why rush them, Nvidia has the fastest card, the gtx580, unless you count the 900$ 5970 on newegg.

Ah I see. Could you link to the official release where they announced the cancellation please?

 

[Lonyo]

Ah I see. Could you link to the official release where they announced the cancellation please?

That would require an official announcement that they were ever going to actually be made.

 

[happy medium]

gtx590 confirmed.

http://vr-zone.com/articles/confirm...tx-590-in-february-with-dual-gf110/10954.html

Edit: sounds like Nvidia is just sitting back cherry picking chips waiting for the 6990.