Any Fermi Users In Here?

[Pudgie]

Nvidia doesn't like to play fair w/ the lanes on a PCI-E 16x slot w/ 1 of their chipsets onboard & especially w/ a EVGA 780i FTW chipset set for optimizing a Nvidia tri-SLI setup.

Heck I can't run an Asus Xonar 7.1 PCI-E 1x sound card on any of the other 2 PCI-E 16x slots due to their not adhering to industry PCI-E design standards--all proprietary to Nvidia SLI design. Vista reports that it can't setup device due to lack of resources.

I really don't think that an ATI vid card will run right--if at all on this mobo. I haven't heard or read of anybody reporting that an ATI Radeon VC running OK on a EVGA 780i FTW mobo to date so........................... .........

If this mobo only had 1 PCI-E 16x slot onboard & SLI wasn't an issue I bet this issue would be non-gratus.

Thus the interest in the GTX 470.

 

[Pudgie]

Well......................... .after all the discussion, thought, nervousness & suffering.................... ........................

I succeeded in convincing the General House Manager (by striking a contract to provide a granddaughter a graduation gift in return for purchasing rights) so I have jumped off & got a EVGA GTX 470 SC vid card along w/ the Asus Xonar DS 7.1 PCI sound card.

I couldn't resist the risk so I done it. Installed 197.75 drivers (latest WHQL's) & set it up.

1st thing out the box I put it to Freestone Video Card tester software just to see.

The card came in at a score of 697 vs the best score I could get w/ the GTX 260 OC'd to 285 speeds was 547 (at stock speeds scores at 512). That is a huge spread in 3D performance (that's w/ the vc's pushed to 99% GPU load w/o artifacts). The card's temps ran up to 93*C w/ fan speed control set in auto during the test-a little too warm for me so I set the fan speed control to manual & set fan speed at 80%.

Played AH approx 35 mins to tax the card under game conditions--max temp on card was 75*C so this is on par w/ my GTX 260 in that regard. The game ran absolutely smooth w/ graphic detail turned up full & 2048 shadows on while vsynched at 75 Hz (75 FPS) at 1280 x 1024 x 32 res (the same old CRT mentioned eariler).
GPU utilization never went past 54%--in fact it flat lined most of the time at 54% (from EVGA Precision 1.9.3-latest version). This part blows my GTX 260 a new one. I kinda knew this should be the result from the Freestone test score results so I am pleased to see this also during gameplay. So far I'm stoked!!!

 

Now all I need is to get the Automatic Fan Speed Control enabled instead of the Dynamic Fan Speed Control (currently enabled in vBIOS) & I'll be all warm & fuzzy inside.

Too bad that Nvidia had to suffer as the power of these cards is something to behold. BTW this 470 runs 3D on .987v power vs my 260 at 1.25v so there is some power savings for me to glean.

So far it's looking like a good ride.................only time & a lot of AH will tell.

 

[Pudgie]

Since I've found MSI Afterburner 1.5.1 & set up user-defined auto fan speed control to follow what I had set up in my 260's BIOS, I'm all set now. 470 SC doesn't run any hotter than my 260 did but at > 2x the performance.

We'll see how it holds up over time but right now I am a happy camper!

See y'all in the skies!!

 

[Pudgie]

So far the ride is SWEET!!!!!!!!!!!!!!!!!

While messing around on the EVGA BBS I ran across some posts where some folks were actually undervolting their Fermis to conserve power & reduce the heat even more. The rationale behind this is that if you're not going to OC these cards & also that a card series voltage that is set in vBIOS is determined by random testing of several parts to find a middle ground that should be workable on mass-produced series & not on an individual part basis so you might be able to reduce voltage thus reduce heat w/o sacrificing performance.

So I went out on Guru3D & got MSI Afterburner 1.6.0beta (which can unlock the voltage adjust on the GTX 470's) & went to testing.

At the moment I have dropped the 3D voltage from .987v to .950v without issues. When I stress tested card w/ new voltage settings in Freestone the score actually went up 1 point (from 697 to 698) which I thought that was interesting.

Just thought I'd share this w/ y'all. Yes I like to play w/ my toys in more ways than 1.


 

[Chalenge]

Now see I didnt think the .03V would actually make much of a heat difference. I changed some internal fans and got my temps down quite a bit but I havent played with the voltages. If what Skuzzy is saying about the chips is true then it wont make a great deal of difference and I want to keep the evga warranties alive or I would have changed the heatsinks out on day one.

[Pudgie]

Hi Chalenge,

Yeah I can understand what you're saying, but since 1.) the vid card makers along w/ Nvidia have sanctioned this kind of usage w/ their vid cards as long as the users are using these 3rd party programs (that are being created by the same programmer BTW who has extensive knowledge of Nvidia's own binary strings) that maintain Nvidia's safety settings & will only allow changes between these points. RivaTuner, EVGA Precision, MSI Afterburner, EVGA Voltage Tuner, EVGA E-LEET & probably a few others are essencially the same software w/ certain levels of access of certain features thru the driver reading the vid card's BIOS file, only differing due to contractural & NDA-level agreements between the vendor, said programmer & Nvidia. The only piece of software that could bring about a breach of warranty trust would be the BIOS editor software (NvFlash, NiBiTor, Fermi BIOS Editor) because they can allow access for the user to alter the safety settings within the vid card's BIOS file which puts the vendor & Nvidia out of exclusive control of the product. So if you're using any of the EVGA-provided "tweaking" software or MSI Afterburner software on an EVGA vid card (EVGA has given the approval of the usage of this app on their vid cards) you will not void their warranties. 2.) Since this GTX 470 SC is just a vanilla 470 tested to run at SC speeds (set in BIOS) w/ voltages adjusted to ensure set speeds from a queried batch sample of chips (not every single chip is tested--can't afford that) it is also very real that this particular Fermi chip doesn't necessarily NEED .987v to perform these rated speeds & so is no more of a risk to reduce the 3D voltage to reduce power consumption thus heat to maintain a clock speed/performance level as it is to raise the 3D voltage to increase power consumption thus heat to increase a clock speed/performance level. In more accounts than less it is most likely a FAR safer proposition to reduce voltage that it is to increase it AS LONG AS THE CHANGES ARE IN SMALL INCREMENTS--which I am doing.

Also I'm a tinkerer by nature & I can't help myself.  

Just to illustrate again, I have lowered the 3D voltage on this 470 SC again from .950v to .937v & ran card thru Freestone stress test--card ran w/o issue & came in at a score of 971 AND dropped under 70*C in the process (actual temp was 67*C). Nvidia std 3D voltage set for the 470 vanilla is .925v so right now I have a 470 SC that is showing to perform to it's settings on .05v less of the std SC rating & within .012v of the LESSER card's voltage std & actually is showing to increase performance level!  Makes you go hhhhmmmm..................... .

 

Now I'm certainly NOT advocating that any one who reads this thread go out & try any of this on their Fermis (or any vid card for that matter) but since I'm a willing volunteer at least I thought I would post the results for the community to see. When it's all said & done there may be some good stuff that comes out of all this.

At least I hope so.

 

[Pudgie]

Also before I forget,

These Fermis are being cooled by HDT-type HSF's (heat pipe direct touch--similar to the Xigmatek CPU HSF's). Anyone who's familiar w/ these type of HSF's will know the issues concerning them. I am using a Xigmatek HDT-S1283 HSF on my C2D E8600 right now. When the TIM is installed properly concerning this type of HSF they perform VERY, VERY well. When it isn't well they don't do so well.

I will probably reapply the TIM on my 470 SC at some point in time just to ensure good thermal contact is there between the GPU & HSF.

Now this is 1 area of concern that a Fermi user does need to consider when confronted w/ perceived temp issues. Can you see a mass-produced line of cards having this type of HSF attached knowing the special issues concerning the design of the contact area of the HSF?

You can read up on this for yourselves if you like.

 

[Pudgie]

Ok, I have now took her apart & have reapplied the TIM on the heat sink using OCZ Freeze TIM. After I used the OCZ cleaning kit to clean off the TIM & remove the old film off the GPU & HSF, I reapplied the Freeze in the same manner as I did w/ the Xigmatek HDT CPU HSF.

Using the line method I put a 1/2 rice grain line of Freeze down the center of each heat pipe approx 3/4 of the width of the GPU (so the spread will cover the other 1/4) after checking the HSF surface to ensure that there is no gap between the heat pipe surface & the plate surface that is between the individual HP's (so that the heat pipes will actually "touch" the GPU's heat spreader). Put her back together & ran it for about 1 1/2 hrs.

I ended up w/ a 4*C temp drop across the board from idle to full load (in ambient room temp of 82*F idle temp 55*C > 51*C, full load of 76*C > 72*C using MSI auto fan control set to ramp from 40% to 100% fan speed across 20*C slope w/ start temp set at 45*C). This should get better when I get the air conditioner running in the house to lower the room ambient temps (yes I'm 1 of them "old guys" that has the box in a mid-tower desk slot that does recirculate some "warm" air back thru the case due to a side fan--CM 690 Nvidia case) to get some air to move to the 780i's HSF (dead area between the vid card & CPU HSF) so it can cool better. In the end the Freeze did improve the situation somewhat so I'm satisfied. Fan speed at 51*C is running between 59%-61%. Since I'm off work for the weekend I'll check all this again when ambient temp drops in early morning (around 3-4 o'clock AM  ).


I also ended up going back to 197.41 drivers from the 197.75's due to drivers started to downclock the card while I was flying regardless of graphics load (was still getting between 40-60 FPS-but would stutter. 197.41's fixed that  ). Yes I run w/ the drivers in Adaptive Mode & the card undervolted to .937v. I even put the voltage back to stock & reset driver to high quality-still downclocked w/ 197.75's under load.

Ok enuff typing--time for flying!

 

[Pudgie]

Have once again dropped 3D voltage on this 470 SC from .937v to .925v while maintaining SC clocks & ran card thru Freestone & got score of 779--that's vanilla 470 voltage.

I think I'll stop at this level for now. This is starting to look like it would have a better move to get a 470 vanilla & OC it up to SC or higher speeds. It also looks to be that the 197.41 drivers are better suited for this 470 as well than the 197.75's were.

 




PS--In post reply #20 I made a typo on the Freestone score--that should have been 671 instead of 971. My bad.

[Ghastly]

After the fiasco that Fermi has become for desktop cards, I honestly have to say that this is about the last thing I ever expected to see...

http://www.zdnet.com/blog/hardware/nvidia-fermi-comes-to-notebooks/8397?tag=nl.e539

<S>

[Skuzzy]

It is about marketing Ghastly.  I doubt any OEM would be stupid enough to offer this as an option in a laptop.  Power/heat issues aside, NVidia's yeilds are so bad there is no way they could meet the demand from an OEM.

Besides, OEM's are still reeling from all the bad GPU's they had to recall for NVidia to replace from the last design fiasco.