NVIDIA GF100 & Friends speculation

[DarthShader]

Those on the backside are tantalum capacitors. Most likely rated above 125C.

 

[neliz]

Those on the backside are tantalum capacitors. Most likely rated above 125C.

yep Hi-c CAP.. and no. they don't get that hot.

 

[Kaotik]

"According to MSI the HI-c caps can last as long as 16 years when running at a constant high temperature of 85C in a high loading environment."

 

[DarthShader]

yep Hi-c CAP.. and no. they don't get that hot.

Sure, but when everything else around gets hot...

 

[nikos64]

I see that the VRMs are DrMOS Infineon TDA21211. These are 35A devices, so 5 of them can supply 175A max. Living on the edge me thinks...

 

[fellix]

Those integrated DrMOS controllers are pretty similar to the ones used in the high-end MSI motherboards, and I can recollect many RMA cases due to blown up VRMs. The main reason, as far as I can remember, was because of overloading of an active phase in a case of erratic dynamic phase switching in idle state, i.e. turning off VRM phases to save power in presumably CPU idle mode.
So, I wonder if NV employs some sort of phase switching in GTX590?

 

[neliz]

So.. the 590 was lowered in price before launch, and supposedly clocked UP (from 550) to try and compete with the 6990.

and the issues? evident:
http://ic.tweakimg.net/ext/i/imagelarge/1301569916.jpeg
http://ic.tweakimg.net/ext/i/imagelarge/1301569917.jpeg
http://ic.tweakimg.net/ext/i/imagelarge/1301569918.jpeg

around the globe end-user reports:
http://foro.noticias3d.com/vbulletin/showpost.php?p=3960849&postcount=352
http://hardforum.com/showpost.php?p=1037046423&postcount=170

#FAIL

I don't think NV employs Phase Switching on the 590, because (as far as I know) it will require software support for that particular setup.

 

[Bouncing Zabaglione Bros.]

Also http://hardforum.com/showpost.php?p=1037041946&postcount=141

Looks like NV drivers are downclocking to get safe OCP, right into games performance levels, not just the "power virus" realms of Furmark.

 

[mczak]

Also http://hardforum.com/showpost.php?p=1037041946&postcount=141

Looks like NV drivers are downclocking to get safe OCP, right into games performance levels, not just the "power virus" realms of Furmark.

Techreport tried that out and didn't find any such downclocking. It is always possible though this is based on sample variance (power draw is a bit different, as is voltage).
Still, VRM blowing up (at or near default voltage at least) is surprising. It doesn't really look underspecced (with 5 phases per chip), nor would it make any sense for nvidia to save some pennies there (not for this card). Maybe design error? If I'm not mistaken if you're not careful you can get oscillations, transient power spikes and other nice things with those VRMs.

 

[compres]

...
http://foro.noticias3d.com/vbulletin/showpost.php?p=3960849&postcount=352
....

Let me know if you (or someone else) wants a proper translation of that one from Spanish to English.

 

[Kaotik]

Techreport tried that out and didn't find any such downclocking. It is always possible though this is based on sample variance (power draw is a bit different, as is voltage).
Still, VRM blowing up (at or near default voltage at least) is surprising. It doesn't really look underspecced (with 5 phases per chip), nor would it make any sense for nvidia to save some pennies there (not for this card). Maybe design error? If I'm not mistaken if you're not careful you can get oscillations, transient power spikes and other nice things with those VRMs.

They only tested 3 games, none of which are used by any site for power measurements (meaning pretty much no-one has noticed them to be powerwise heavy for cards)

 

[AnarchX]

GF110 based "GTX 560 Ti"?

NVIDIA_DEV.1080.01 = "NVIDIA GeForce GTX 580"
NVIDIA_DEV.1081.01 = "NVIDIA GeForce GTX 570"
NVIDIA_DEV.1082.01 = "NVIDIA GeForce GTX 560 Ti"
NVIDIA_DEV.1086.01 = "NVIDIA GeForce GTX 570 "
NVIDIA_DEV.1088.01 = "NVIDIA GeForce GTX 590"
[...]
NVIDIA_DEV.1200.01 = "NVIDIA GeForce GTX 560 Ti "

In older beta driver NVIDIA_DEV.1082.01 appeared as D13U.

 

[Man from Atlantis]

optimus for desktop called synergy ?

http://translate.google.com/transla...ditt-geforce-kort-prata-med-sandy-bridge.html

 

[onethreehill]

NVIDIA GeForce GTX 560 to be launched on May 17
http://www.techpowerup.com/144669/May-17-Launch-for-NVIDIA-GeForce-GTX-560.html

 

[onethreehill]

ASUS MARS II and MATRIX GTX 580

http://pctuning.tyden.cz/index.php?option=com_pctlongblock

http://pctuning.tyden.cz/component/...alni-zpravy/21081-asus-geforce-gtx-580-matrix

ASUS MATRIX GTX 580
http://www.xfastest.com/cms/tid-61294/

 

[mczak]

Hmm so full GTX 580x2 AND the ability to tweak frequencies and voltages?
Just how many power connectors is this thing going to have, 3 8pin?

 

[Malo]

Why isn't it called a 590?

 

[DuckThor Evil]

Mars 2 is a 2x card and the Matrix is just a 580.

 

[mczak]

Why isn't it called a 590?

I dunno maybe to indicate it's better than a 590? Everyone knows that the GTX590 is clocked low, blows up when overclocking and is just nowhere near as fast as 2xGTX580.
But of course, you're right the Mars II IS a GTX 590. Hopefully with higher quality voltage regulation circuitry. And possibly with higher clocks (though really I think if they want anywhere close to GTX580 clocks at stock they are going to need that 3rd 8pin connector - but it might just end up as a pretty regular GTX 590 with better overclocking capability, then noone really cares how much juice it draws from 2 connectors when overclocked...).

 

[DuckThor Evil]

Asus has made several limited edition cards that don't have the number in the name, these cards have usually been very expensive. If you just call it GTX590 thunder or something like that, it's hard to squeeze big margins from it.