Tweaker Turns GeForce GTX 690 Into a Quadro K5000

[N.Broekhuijsen]

A user from the EEV forums has figured out how to turn a GTX 690 into a Quadro K5000 card, or at least make it appear so.

Tweaker Turns GeForce GTX 690 Into a Quadro K5000 : Read more

 

[anononon]

I am very curious to see how well this performs.

If it does well, thats great! I remember years ago you could solder a couple points to turn a 200 series (I think) into a quadro card as well. That worked great then, why not now?

 

[tipoo]

He (or tomshardware?) should try running things that use Quadro acceleration on it, Mudbox, Maya, etc. Then compare the test back to the actual K5000 and see if it runs stably for long renders, then you can tell if it's just the name that changed or if it really behaves like a K5000. Pretty sweet if it did, I missed the days of turning consumer cards into workstation beasts.

 

[renz496]

ell us, if you had a GTX Titan or GTX 690, would you consider giving this trick a shot?

for people that really intend to use properly it i don't think anyone would dare doing so. unless they can get it free but still able to get replacement on top of that after breaking one. btw why need to flash Titan into Quadro (if ever existed)? the performance for the card has been unlock so if they really want quadro like performance and precision maybe they just need to tinker with quadro driver to work with Titan

 

[nforce4max]

It has been known for years how to mod cards on the software side without the need of taking chances with other roms. At least the hardware mod info got out but all one has to do is save the rom then manually change the device id ect while leaving the timings intact. For this mod to work one has to know how to flash on a multi gpu system and the pci-e switch will show up in nvflash.

 

[JDFan]

I don't see it being useful for actual application as the Quadro's have ECC memory for a reason and that is why they cost so much more and anyone using an application that needs the ECC memory is not going to risk the accuracy of the data just to save a few bucks !

 

[-Jackson]

[citation][nom]JDFan[/nom]I don't see it being useful for actual application as the Quadro's have ECC memory for a reason and that is why they cost so much more and anyone using an application that needs the ECC memory is not going to risk the accuracy of the data just to save a few bucks ![/citation]
Well actually most of the reason why these sort of cards cost more is mainly because of manufacturer support.

 

[glob]

I don't see it being useful for actual application as the Quadro's have ECC memory for a reason and that is why they cost so much more and anyone using an application that needs the ECC memory is not going to risk the accuracy of the data just to save a few bucks !

Well actually most of the reason why these sort of cards cost more is mainly because of manufacturer support.

No, the reason why they cost (much) more is because companies and freelancers see it worthy to spend through the nose for the added performance gained by the drivers. ECC is a fraction of the added cost and the manufacturer support is irrelevant for most people. The productivity gained from faster rendering jobs is not.

 

[falchard]

I think the most recent NVidia cards would not be able to effectively do this trick. NVidia made concessions with this generation in GPGPU functionality for the consumer cards, so I don't see it being softmoded into a Quattro as something more than switching what appears in messages.

 

[Inferno1217]

No I wouldn't do that to mine. I like it the way it is.

 

[alidan]

[citation][nom]glob[/nom]No, the reason why they cost (much) more is because companies and freelancers see it worthy to spend through the nose for the added performance gained by the drivers. ECC is a fraction of the added cost and the manufacturer support is irrelevant for most people. The productivity gained from faster rendering jobs is not.[/citation]

its not that it renders faster, its that its guaranteed to work, as in the programs wont crash.
a single crash can cost a crapton of money, and its better to have no crash than many that add up to costing more than a single pro card.

 

[blazorthon]

[citation][nom]falchard[/nom]I think the most recent NVidia cards would not be able to effectively do this trick. NVidia made concessions with this generation in GPGPU functionality for the consumer cards, so I don't see it being softmoded into a Quattro as something more than switching what appears in messages.[/citation]

The Quadro K5000 uses nearly identical hardware to the GTX 690 and has the exact same concessions made in dual-precision compute performance for GPGPU processing. However, they both perform decently in single precision (at least much better than Fermi).

 

[blazorthon]

[citation][nom]JDFan[/nom]I don't see it being useful for actual application as the Quadro's have ECC memory for a reason and that is why they cost so much more and anyone using an application that needs the ECC memory is not going to risk the accuracy of the data just to save a few bucks ![/citation]

At least some form of ECC is used even in regular graphics cards. It's the cause of why overclocking graphics memory oftentimes reaches a point where performance is hurt instead of improved IIRC.

 

[downhill911]

Can't wait to start....rendering. LOL

 

[jkflipflop98]

. . . you've been able to do this with Rivatuner since the GeForce2 came out. You just load in a Quadro bootstrapper and you're set. It's the same silicon.

 

[demonhorde665]

[citation][nom]tipoo[/nom]He (or tomshardware?) should try running things that use Quadro acceleration on it, Mudbox, Maya, etc. Then compare the test back to the actual K5000 and see if it runs stably for long renders, then you can tell if it's just the name that changed or if it really behaves like a K5000. Pretty sweet if it did, I missed the days of turning consumer cards into workstation beasts.[/citation]
Renders don NOT run on the gpu , The gpu only powers the veiw ports 3ds max (and any other 3d program worth its salt) has ALWAYS rendered off the cpu. the main reason being that the program(s) suport technologies that can't be rendered in real time such as ray tracing, or technologies that are not present in Direct 3d or open glide yet.

 

[thecolorblue]

[citation][nom]-Jackson[/nom]Well actually most of the reason why these sort of cards cost more is mainly because of manufacturer support.[/citation]

That's what they say... interesting how "manufacturer support" is available as a separate OPTIONAL contract for loads of computer hardware... but when it comes to "workstation cards" the "support contract" is built into the card... even if the user will never ever call support.

for that reason alone, your explanation is nothing more than a repeat of marketing BS

 

[Achoo22]

It's interesting that nobody points out the irony of not being able to just modify and distribute the drivers (which function as a LKM and must therefore be bound to the GPL) instead of modifying the hardware to change the PCIID.

 

[blazorthon]

[citation][nom]thecolorblue[/nom]That's what they say... interesting how "manufacturer support" is available as a separate OPTIONAL contract for loads of computer hardware... but when it comes to "workstation cards" the "support contract" is built into the card... even if the user will never ever call support.for that reason alone, your explanation is nothing more than a repeat of marketing BS[/citation]

I'm pretty sure that this article's comments is a good example of how making assumptions based on a single fact doesn't go well. I'm sure that the prices of the workstation/super computer cards factor in support, the additional hardware expense of higher RAM capacity of superior reliability and more, more stringently tested drivers, and more as well as the fact that they're sold in a market that can afford the high prices even if the cards shouldn't be worth it by performance (relative to the consumer cards) alone.

 

[InTheCity]

[citation][nom]thecolorblue[/nom]That's what they say... interesting how "manufacturer support" is available as a separate OPTIONAL contract for loads of computer hardware... but when it comes to "workstation cards" the "support contract" is built into the card... even if the user will never ever call support.for that reason alone, your explanation is nothing more than a repeat of marketing BS[/citation]

Several render engines run off the gpu only, 'Octane' for example, is a GPU only unbiased render engine. While not as powerful as hybrid engines that are cpu/gpu, they are finding their way into production workflows.

 

[rmpumper]

[citation][nom]glob[/nom]No, the reason why they cost (much) more is because companies and freelancers see it worthy to spend through the nose for the added performance gained by the drivers. ECC is a fraction of the added cost and the manufacturer support is irrelevant for most people. The productivity gained from faster rendering jobs is not.[/citation]

Quadro and GPUs in general have nothing to do with rendering performance (unless you are using vray RT, mr iray or some other standalone GPU renderer). Quadros are only provide better viewport performance.

 

[gofasterstripes]

I have a 470GTX, with ECC memory. I am studying a Design Degree and need to run Autodesk Alias and Solid Edge, maybe Inventor too. I NEED TO KNOW HOW TO DO THIS!

I thought it was impossible for the last few years... Any info available?

 

[christop]

Very cool...

 

[Geoffrey McRae]

Hi All,

I am the author of this mod, I am a little surprised that toms hardware did not post a link to the thread with all the technical details and discussion based on the other cards.

As someone else here has pointed out, the risk of damaging a card to find these straps is rather high, and as such I am currently accepting donations towards another GTX690 so that the straps for GPU1 can be found.

The forum and details can be found here:

http://www.eevblog.com/forum/projects/hacking-nvidia-cards-into-their-professional-counterparts/

 

[N.Broekhuijsen]

[citation][nom]Geoffrey McRae[/nom]Hi All,I am the author of this mod, I am a little surprised that toms hardware did not post a link to the thread with all the technical details and discussion based on the other cards.As someone else here has pointed out, the risk of damaging a card to find these straps is rather high, and as such I am currently accepting donations towards another GTX690 so that the straps for GPU1 can be found.The forum and details can be found here:http://www.eevblog.com/forum/proje [...] nterparts/[/citation]
Geoff, I've updated the article with some relevant information.