Frequently Asked Questions About NVIDIA PC (non-EFI) Graphics Cards

[h9826790]

Pixlas mod is way to heavy
I just need +47 watt for k5000
1st card max 190W with 2x6pin
2nd card max 122W 1x6pin

Any cable magic ?

ps:SLI is just idea for BC with drv-mod

You assume the K5000 will automatically draw full 75W from the slot but only remaining 47W from the 6pin. Then why not assume it will automatically draw 75W from the 6pin but only remaining 47W from the slot?

Anyway, if you are not running dual Furmark on both cards at the same time. Most likely you can...

1) DUAL mini 6pin -> single 8pin; and

2) single 8pin FEMALE -> 3x6pin (if can't find such cable, then start with dual 6pin, and then further spilt one 6pin to dual 6pin).

So, what will actually happen?

A) In best situation. BOTH GPU draw full 75W from the slot. Which means only need to draw the remaining 162W from the mini 6pins. With the 8pin work as the bridge. The power draw are pretty much shared between mini 6pins no matter which card and which connector draw power. So, each mini 6pin only need to deliver 81W (max) in real world (hopefully). Even though we assume the mini 6pin is only rated up to 75W (like the normal 6pin does), but we found that each mini 6pin in real world can usually deliver up to ~120W before the shutdown protection kick in. So, 81W is way below that. This setup should work.

B) Worst case scenario. Both GPU always draw all 75W from all 6pins, and only the remaining 87W from the slot. In this case, total 225W from the mini 6pins is required. So, 112.5W per mini 6pin. Which still a bit below the 120W real world limit. So, should work as well.


My personal prediction.Should be somewhere between A and B. That means, in really world high stress condition. Each mini 6pin should require to deliver around 100W. Still quite a bit from the shutdown protection. But also way above that 75W limit.

Anyway, for your info. This is from my own 1080Ti. Which also powered by this "bridge method". 6+8 pin setup, max possible power draw via the mini 6pins are also 225W (same as your situation).

With the bridge, the power draw from the mini 6pins are very balanced. Which can effectively avoid shutdown protection. And except Furmark, the GPU usually won't draw that much.

But if you are not 100% sure that your card's real world power draw (TDP is for thermal calculation, not necessary 100% reflecting the card's max real world power draw. You may

1) install ONLY the GTX680, run Furmark (or whatever you will really use in real world), copy down the MAX power draw of BOTH mini 6pins. (e.g. 70W and 60W)

2) install ONLY the K5000, run Furmark (or your most demanding real world applications), copy down the MAX power draw from the mini 6pin. (e.g. 55W)

3) Add all 3 recorded numbers together (70+60+55 = 185)

4) divide that number by 2 (185 / 2 = 92.5)

This will be the number you can expect each mini 6pin have to deliver (when both cards are installed, and under very high stress).

IMO, as long as that number is not above 100W, it will be OK for long term use. HOWEVER, that's just my own standard. At the end, mini 6pins are still 6pins, very possible only rated up to 75W, anything beyond that you should consider unsafe.

I can tell you that I can run dual 7950 with just the mini 6pins.

https://forums.macrumors.com/thread...mac-with-2-d700s.1732849/page-5#post-21722712

This should be more stressful than your setup. So, if I can do that for years, I can't see why your cMP cannot. But just are you willing to do that.

 

[MacGamver]

It's almost laughably better on Windows, just like every other game I've tried. Say what you will about Windows vs macOS, it's very clear that Microsoft takes PC gaming very, very seriously.

Edit: I should also point out that I was shocked when I retired my cMP and switched to a Hackintosh about 4 or 5 years ago, even though the clock speeds weren't that much higher the PC was just way faster.

Not to derail to a gaming discussion, but... When we can no longer upgrade MacOS on our cMPs, how much more life as a windows box you think they have? the XEON 3565 chip is pretty old now, and I'm not sure how long it will even run windows. I'm sticking with Win7Pro still because I absolutely hate W10.

 

[Speedstar]

You assume the K5000 will automatically draw full 75W from the slot but .....

My personal prediction.Should be somewhere between A and B. That means, in really world high stress condition. ......
.....
IMO, as long as that number is not above 100W, it will be OK for long term use. HOWEVER, that's just my own standard. At the end, mini 6pins are still 6pins, very possible only rated up to 75W, anything beyond that you should consider unsafe.

I can tell you that I can run dual 7950 with just the mini 6pins.

https://forums.macrumors.com/thread...mac-with-2-d700s.1732849/page-5#post-21722712

This should be more stressful than your setup. So, if I can do that for years, I can't see why your cMP cannot. But just are you willing to do that.

Extremlly good explained, so I need cables I Can 100% trust not to burn above 75W and under 100W
Got it, well If You were me would you use such a rig to render Vray animations on GPU for hours or days, ? Because above standard and the Fur-banchmark makes me scary,...

(I have not a budget for experiments ,I yesterday boughtK5000, but your math is absolutely convincing!)

What is the tool You use for
Measuring ??

Scenario B You said:"87W from the slot" does this not grill the PCI slot?
Is 75W for slot maximum,
How can I further avoid to roast my
McP ?



PS: has anyone a clue, it has been three years ago , most people flashed their gtx680s,
But I did not find anything on
Flashing PNY K5000, (doodled it for hours netkas blogs nowhere)I even did not find how large the flash rom is on PNY card but got a ~270kb
Mac rom which is bigger then 220kb PC rom...
______________________
In the End it's more save to install 1050ti one slot as extra card but be banned with HS...;((


I collected money for McP to swap disks and Apple hardware Easily but after SIP it seams to tune or upgrade dell hardware is way easier...
Once I dreamed to buy a OOB computer which makes my all wishes come true, no viruses,superb UI,
Mac I go with You,
(until Steve sold Pixar)
But Apple made all I hated in self made 486 configs come alive
again ,fight for every little self added peace of hardware, remember those days with 486 Linux and ATI even OpenSTEP drivers from ATi, but then I didn't want to spend all my time with upgrading hardware but with working with computer to produce ART ....

Does anyone remember QuarzComposer On Tiger (PowerPC AlitVec) could accelerate core images effects on running QuickTime in realtime in Full 576p .

Good implementation of Metal-Api in future at this company???

 

[h9826790]

Extremlly good explained, so I need cables I Can 100% trust not to burn above 75W and under 100W
Got it, well If You were me would you use such a rig to render Vray animations on GPU for hours or days, ? Because above standard and the Fur-banchmark makes me scary,...

(I have not a budget for experiments ,I yesterday boughtK5000, but your math is absolutely convincing!)

What is the tool You use for
Measuring ??

PS: has anyone a clue, it has been three years ago , most people flashed their gtx680s,
But I did not find anything on
Flashing PNY K5000, (doodled it for hours netkas blogs nowhere)I even did not find how large the flash rom is on PNY card but got a ~270kb
Mac rom which is bigger then 220kb PC rom...
______________________
In the and it's more save to install 1050ti one slot as extra card but be banned with HS...;((

I often let my dual 7950 or 1080Ti work under stress for hours per day (but not days continuously). So, I am quite sure the cMP can support that.

But if I were you, I will simply get a stronger GPU. But I understand that you don't have the budget right now. So, not an option at this moment.

You can install hardware monitor / iStat to monitor the power draw. You only need to care about the PCIe Booster 12V line's reading.

If you want to know more about "how much you really need to over draw from the mini 6pin", please do the following.

1) install only the GTX680, let it to Vray animations, record down the power draw (or current) on BOTH the PCIe Booster 12V lines.

2) install only the K500, and do the same thing again.

Better with screen capture, and post it here. Then I can go through the calculation with you.

 

[Theophany]

Not to derail to a gaming discussion, but... When we can no longer upgrade MacOS on our cMPs, how much more life as a windows box you think they have? the XEON 3565 chip is pretty old now, and I'm not sure how long it will even run windows. I'm sticking with Win7Pro still because I absolutely hate W10.

I installed a second SSD and installed Windows 10 a couple of months ago and haven't booted back into MacOS since I installed. Honestly, it runs just great. Download Intel XTU and you can bump up a W3690 to ~4Ghz with no issue, the only thing is you can't adjust voltages so anything above that will give you instability.

With a 980Ti I can run pretty much anything on Steam at 3840x1600 res. Adobe CC apps tend to be less garbage on Windows. Admittedly, Windows does feel a little less polished than MacOS but the tradeoff is that it is far more performant and this old 4,1 feels pretty damn good.

I'd never have done this if Apple hadn't sat on the webdrivers. It feels like a real kick in the gut to be told we'd get Mojave if we had a compatible card when what they really meant is 'you'll get Mojave if you use this specific AMD card with crummy performance relative to alternatives.'

 

[Asgorath]

And just for reference, here's a comparison between that CPU (at stock clocks) and the last two generations of Intel CPUs.

https://www.cpubenchmark.net/compar...l-i7-8700K-vs-Intel-i9-9900K/1275vs3098vs3334

If you're running Windows, at some point it's going to be worth just building a PC (which should be possible for well under $1000 if you already have a graphics card). The cMPs are really, really old at this point and even mid-range Core i5 CPUs are significant upgrades.

https://www.cpubenchmark.net/compare/Intel-Xeon-W3690-vs-Intel-i5-9600K/1275vs3337

 

[pastrychef]

And just for reference, here's a comparison between that CPU (at stock clocks) and the last two generations of Intel CPUs.

https://www.cpubenchmark.net/compar...l-i7-8700K-vs-Intel-i9-9900K/1275vs3098vs3334

If you're running Windows, at some point it's going to be worth just building a PC (which should be possible for well under $1000 if you already have a graphics card). The cMPs are really, really old at this point and even mid-range Core i5 CPUs are significant upgrades.

https://www.cpubenchmark.net/compare/Intel-Xeon-W3690-vs-Intel-i5-9600K/1275vs3337

Yup. A 9900K outperforms the old MacPro5,1 by a lot. My 9900K@5GHz.

 

[Speedstar]

And just for reference, here's a comparison between that CPU (at stock clocks) and the last two generations of Intel CPUs.

https://www.cpubenchmark.net/compar...l-i7-8700K-vs-Intel-i9-9900K/1275vs3098vs3334

If you're running Windows, at some point it's going to be worth just building a PC (which should be possible for well under $1000 if you already have a graphics card). The cMPs are really, really old at this point and even mid-range Core i5 CPUs are significant upgrades.

https://www.cpubenchmark.net/compare/Intel-Xeon-W3690-vs-Intel-i5-9600K/1275vs3337

Strange Comperison for me as Vintage McP 5.1 Dual X5650 CPU User,I have a Score

https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+X5650+@+2.67GHz&id=1304&cpuCount=2

about ~19000 with 2,6
Dual xeons and 120Mbit Internet connection , but a modern Smart phone open some websites in desktop mode sometimes faster on G3/UMTS ;(
[doublepost=1543668726][/doublepost]

Yup. A 9900K outperforms the old MacPro5,1 by a lot. My 9900K@5GHz.
View attachment 807618

http://browser.geekbench.com/geekbench2/2663170

 

[Asgorath]

Strange Comperison for me as Vintage McP 5.1 Dual X5650 CPU User,I have a Score

https://www.cpubenchmark.net/cpu.php?cpu=Intel+Xeon+X5650+@+2.67GHz&id=1304&cpuCount=2

about ~19000 with 2,6
Dual xeons and 120Mbit Internet connection , but a modern Smart phone open some websites in desktop mode sometimes faster on G3/UMTS ;(
[doublepost=1543668726][/doublepost]
http://browser.geekbench.com/geekbench2/2663170

Where are you seeing 19000 in PassMark? The link you gave says 11349 for your dual-CPU system, which means a single 9900K is nearly twice as fast as your dual-CPU config. That means it's nearly 4 times as fast as one of your CPUs (even if it's only 3 times as fast, that's a staggering difference).

 

[Theophany]

And just for reference, here's a comparison between that CPU (at stock clocks) and the last two generations of Intel CPUs.

https://www.cpubenchmark.net/compar...l-i7-8700K-vs-Intel-i9-9900K/1275vs3098vs3334

If you're running Windows, at some point it's going to be worth just building a PC (which should be possible for well under $1000 if you already have a graphics card). The cMPs are really, really old at this point and even mid-range Core i5 CPUs are significant upgrades.

https://www.cpubenchmark.net/compare/Intel-Xeon-W3690-vs-Intel-i5-9600K/1275vs3337

Yep, I'm planning on a Ryzen 7 build in the New Year. They're priced very competitively and I'm sure it'll feel like a monstrous performance upgrade.

 

[Hugebricks]

I have a question, to which the answer is probably just "update" but I'll ask anyway: I have 10.12.5 installed currently with the 378.05.05.15f01 driver, and I just added in a 1070 TI. Runs with no issue so far, other than macOS saying the card is an "NVIDIA Graphics Device 8191 MB". Is that to be expected with the older driver, or maybe it's happening because it's not a regular 1070?

 

[flowrider]

I don't know why or how, but I read a couple of reports stating that the 108 Web Driver works with High Sierra, Security Update 2, Build (17G3025). I didn't really believe it, but this morning I thought I'd give it a try. I cloned my Boot Disk, and updated to the new build. AND - GOSH DARN - IT WORKED! I was blown away. I don't know of this is true of the 105 driver or not, but for the 108 driver a BIG YES!!!!

And The old 108 Nvidia web Driver (387.10.10.15.15.108) is still working with Security Update 003 Build No. (17G4015). And Nvidia Driver manager lists it as up to date.

Lou

 

[angelsevov]

And The old 108 Nvidia web Driver (387.10.10.15.15.108) is still working with Security Update 003 Build No. (17G4015). And Nvidia Driver manager lists it as up to date.

Lou

for me too
version (387.10.10.15.15.108) still works after all security update installations

 

[bsbeamer]

And The old 108 Nvidia web Driver (387.10.10.15.15.108) is still working with Security Update 003 Build No. (17G4015). And Nvidia Driver manager lists it as up to date.

Lou

Something tells me NVIDIA knew there were going to be Mojave issues and dropped this version. How they got around the update requirement might be why there were install issues for some.

Still on this driver and should be on security update for HS next week.

 

[flehman]

Anyone try installing today's High Sierra security update with the 387.10.10.10.40.108 drivers?

 

[Speedstar]

Where are you seeing 19000 in PassMark? The link you gave says 11349 for your dual-CPU system, which means a single 9900K is nearly twice as fast as your dual-CPU config. That means it's nearly 4 times as fast as one of your CPUs (even if it's only 3 times as fast, that's a staggering difference).

I assumed it's a score of one of this Dual CPUs my foult ...
[doublepost=1544090562][/doublepost]

Yep, I'm planning on a Ryzen 7 build in the New Year. They're priced very competitively and I'm sure it'll feel like a monstrous performance upgrade.

in Mojave beeing tyied to AMD's VEGA is not a solid option

 

[Theophany]

I assumed it's a score of one of this Dual CPUs my foult ...
[doublepost=1544090562][/doublepost]
in Mojave beeing tyied to AMD's VEGA is not a solid option

I won't be Hackintoshing

 

[rjtiedeman]

Anyone try installing today's High Sierra security update with the 387.10.10.10.40.108 drivers?

The update is NOT COMPATIBLE with my NVIDIA GeForce GTX 680. Fortunately it's A Mac Edition and is supported under MAC OS. No CUDA for me. I suspect I will have to delete the NVIDIA web drivers and CUDA drivers now and just move on.

 

[bsbeamer]

The update is NOT COMPATIBLE with my NVIDIA GeForce GTX 680. Fortunately it's A Mac Edition and is supported under MAC OS. No CUDA for me. I suspect I will have to delete the NVIDIA web drivers and CUDA drivers now and just move on.

This is typically common after an update install since it changes the OS version. Wait a few days to see if updated NVIDIA web drivers become available, or switch to the 387.10.10.15.15.108 driver. The 387.10.10.15.15.108 driver does not appear to "break" with OS updates.

 

[flehman]

It is interesting if installing the High Sierra security update while on the 387.10.10.10.40.108 drivers did not offer the 387.10.10.15.15.108 drivers as an upgrade path for the new build through the NVIDIA control panel. Maybe a different build of drivers will be offered through the control panel after the typical day or three passes...which will mean that 387.10.10.15.15.108 drivers are still sort of off by themselves.

 

[bsbeamer]

It is interesting if installing the High Sierra security update while on the 387.10.10.10.40.108 drivers did not offer the 387.10.10.15.15.108 drivers as an upgrade path for the new build through the NVIDIA control panel. Maybe a different build of drivers will be offered through the control panel after the typical day or three passes...which will mean that 387.10.10.15.15.108 drivers are still sort of off by themselves.

387.10.10.15.15.108 included VOLTA drivers and was pulled by NVIDIA. Rumor has it at Apple's request.
387.10.10.10.40.108 series of drivers do not include VOLTA and "break" with each OS update.

I would expect an update to the 387.10.10.10.40.1XX drivers to be released in the next week or so.

 

[Speedstar]

You assume the K5000 will automatically draw full 75W from the slot but only remaining 47W from the 6pin. Then why not assume it will automatically draw 75W from the 6pin but only remaining 47W from the slot?

...

This should be more stressful than your setup. So, if I can do that for years, I can't see why your cMP cannot. But just are you willing to do that.

It is not McP 5.1 I tried this PC Card in my Mcp 3.1 downloaded FurMark from GPUtestOSX and started same App as You...

As You said It's taking 47-50 Watt from PCIbus and 68W from P-Connector


Mac680gtx takes 52,5W max from PCIbus
and 63,2W from 6pinPower-Connector1
61,5W -Connector2

122W K5000ALL -50PCIbus -( p1_63,2+p2_62,5+p1&2Y_68)/2 = means
96,85 watt per one PCIe 6Pin Connector cable right ?

and now the question is do I roast anything ?
and which Y-PCIe 6Pin splitter I have to use ...

PS:booted without web driver

PSS: If anyone knows how large PNY VBIOS ROM is please tell me.
I read here PC file is 220kb mac 280 uncompressed

 

[Asgorath]

I won't be Hackintoshing

Just as long as you understand how much CPU performance you're giving up by sticking with a cMP, then sure, do whatever works best for you. Low CPU performance can also mean low GPU performance, as your decade-old CPU simply cannot feed work quickly enough to a modern high-end GPU to keep it busy.

 

[h9826790]

It is not McP 5.1 I tried this PC Card in my Mcp 3.1 downloaded FurMark from GPUtestOSX and started same App as You...

As You said It's taking 47-50 Watt from PCIbus and 68W from P-Connector


Mac680gtx takes 52,5W max from PCIbus
and 63,2W from 6pinPower-Connector1
61,5W -Connector2

122W K5000ALL -50PCIbus -( p1_63,2+p2_62,5+p1&2Y_68)/2 = means
96,85 watt per one PCIe 6Pin Connector cable right ?

and now the question is do I roast anything ?
and which Y-PCIe 6Pin splitter I have to use ...

PS:booted without web driver

PSS: If anyone knows how large PNY VBIOS ROM is please tell me.
I read here PC file is 220kb mac 280 uncompressed

Yeah, seems should peak at 97W in worst case. Should be OK to use the "bridge" method.

Any good quality 6pin / 8pin can do the job. No matter which one, only pin 1,2,3 will carry the 12V current. At the "bridge" the worst case is each 12V connection pin need to handle about 5.5A. That shouldn't be a problem.
[doublepost=1544211356][/doublepost]

Just as long as you understand how much CPU performance you're giving up by sticking with a cMP, then sure, do whatever works best for you. Low CPU performance can also mean low GPU performance, as your decade-old CPU simply cannot feed work quickly enough to a modern high-end GPU to keep it busy.

I can confirm this.

MY PNY XLR8 1080Ti on my cMP (with W3690) perform like this.

The exact same card on my Hackintosh (with 8700k) can do this

 

[Speedstar]

Strange McP3.1 has min FPS better then McP5.1 with better card