MP 1,1-5,1 What do I lose if I go High Sierra and 1080 Ti?

[sanfrancisofont1984]

Typing on Mojave right now and been researching Vega cards for past few days. From https://forums.macrumors.com/threads/gpu-compatibility-list-for-cmp.2174600/ the 1080 Ti performance is on par with a Radeon VII. I happen to have a 1080 Ti from my "previous life".

At least in theory the 1080 Ti power consumption is more modest. It uses 6 + 8 pins while all (consumer) Vega cards I've seen so far use 8 + 8 (even on reference-like Vega 56 cards!).

I know I am going to lose official dark mode and fancier screenshot tool. And I would have to use an even lower version of Xcode (don't care right now). What else am I going to lose right now?

Or do I actually gain something because of CUDA? E.g., GPU accelerated Cycles renderer in Blender?

 

[mode11]

Well, after Big Sur comes out a month or two, High Sierra will no longer be supported. It'll just be BS, Catalina and Mojave getting security updates. With an Nvidia card you'll be stuck on HS, so if there are any features in later operating systems that are desirable you'll be out of luck. This includes certain integrations with the latest version of iOS. Both the 1080Ti and Vega 64 / Radeon 7 will need a 'Pixla's mod' to power them, so no difference in convenience there. A non-oc Vega56 should be fine powered off the motherboard's mini six-pin connectors though.

CUDA is obviously a non-starter on macOS, so not much software will support it going forward. AMD GPUs support the Metal alternative well. The main advantages of the 1080Ti will only really apply if you regularly boot into Windows. Doubt the power savings of the Nvidia card would be meaningful given the already juicy nature of the MP itself. The 1080Ti would fetch a good price, so could be 'swapped' for a second hand Vega of some sort without much cost.

 

[sanfrancisofont1984]

Thanks. The post I linked says getting additional power from SATA might work for 1080 Ti. I guess I just have to try it out myself. The plan B is staying at RX 580.

Good points about keeping up with macOS releases. My hope is that next-gen graphics cards come out around the same time BS is out. So by that time Vega cards are even cheaper and I could then upgrade to Mojave.

 

[mode11]

I don't have a 1080 Ti myself, but it is a 250W card vs. the RX580's 185W. The Vega 56 is 210W and the Vega 64 is 295W (all figures from GPU Boss). My impression is that the Vega 56 is on the limit for powering from internal connectors (and even factory OCing will push it over). Perhaps the 1080 Ti is doable if you can get an extra 40W from SATA power, but I've always thought this class of card needs a Pixla's mod. Could be wrong though. Of course, the above are all peak power and regular use will be lower, but you don't want random shut downs, or to put excessive / long-term strain on logic board power components.

The RX 5700 can apparently be powered from just the motherboard six-pin connectors, though is still pretty new / pricey (about £370 in the UK). The Vega 56 could be a good buy - they're consistently about £220 used on eBay, for a boxed Sapphire Pulse in good condition.

 

[flyproductions]

What else am I going to lose right now?

The only real "goodie" coming to my mind (aside of not getting updates any more some time soon of cause) is hardware accelleration for video. That's it so far!

I'm happily using HS together with a 1070 (for dual-DVIs) right now and have some (non ti) 1080 too. Both cards are very efficient. The 1070 is idling around 10-12 watts. And even the (factory-overclocked!) 1080 gets nicely along with what the mainbord connectors can offer: at 190 watts in peaks. Ok, wouldn't try this wit a ti.

I have some other reasons to stay at HS, whichof one is a bunch of 32-bit stuff i have no real replacements for. At the moment i see no real reason to update.

 

[mode11]

H264 and H265 4K video is accelerated for encode and decode with Polaris (or higher) AMD cards in Catalina, if you're running the OpenCore bootloader (which is well worth it in lots of ways, including boot screen support and a boot picker).

 

[sanfrancisofont1984]

I don't have a 1080 Ti myself, but it is a 250W card vs. the RX580's 185W. The Vega 56 is 210W and the Vega 64 is 295W (all figures from GPU Boss). My impression is that the Vega 56 is on the limit for powering from internal connectors (and even factory OCing will push it over). Perhaps the 1080 Ti is doable if you can get an extra 40W from SATA power, but I've always thought this class of card needs a Pixla's mod. Could be wrong though. Of course, the above are all peak power and regular use will be lower, but you don't want random shut downs, or to put excessive / long-term strain on logic board power components.

My impression from reading some other threads in this forum is that the peak of Vega 56 could be much higher, e.g., https://forums.macrumors.com/threads/using-vega-56-in-mp-5-1-questions.2247269/post-28706841

OTOH, there are some posts elsewhere claiming that adding a SATA to 6 pin is enough for 1080 Ti / Titan.

The 2010 MacPro and the GTX TITAN – Mazze Aderhold

www.mazze-aderhold.com

The Pesky Power issue with pascal 1080ti and titan

So, as much as I'd like to ship off a few dozen of these babies we've been holding off due to lingering apprehension about power. I sent some cards to Barefeats.com for testing. He got FANTASTIC...

www.macvidcards.com

( I do not endorse MVC just to be clear. )

I will try a slightly safer approach as in https://forums.macrumors.com/threads/evga-powerlink-with-amd-gpus.2166917/

 

[sanfrancisofont1984]

The only real "goodie" coming to my mind (aside of not getting updates any more some time soon of cause) is hardware accelleration for video. That's it so far!

Just curious, are you using WhateverGreen? I think AMD cards need it to activate hardware acceleration for video.

Activate AMD hardware acceleration

[For existing OpenCore users, you can always download my latest OpenCore package at post #1314 for update] Full HWAccel achieved now! Everything is covered in the OpenCore thread. The step by step tutorial to enable AMD hardware acceleration is now at here, but I strongly recommend you go...

forums.macrumors.com

 

[mode11]

My impression from reading some other threads in this forum is that the peak of Vega 56 could be much higher, e.g., https://forums.macrumors.com/threads/using-vega-56-in-mp-5-1-questions.2247269/post-28706841

OTOH, there are some posts elsewhere claiming that adding a SATA to 6 pin is enough for 1080 Ti / Titan.

The 2010 MacPro and the GTX TITAN – Mazze Aderhold

www.mazze-aderhold.com

The Pesky Power issue with pascal 1080ti and titan

So, as much as I'd like to ship off a few dozen of these babies we've been holding off due to lingering apprehension about power. I sent some cards to Barefeats.com for testing. He got FANTASTIC...

www.macvidcards.com

( I do not endorse MVC just to be clear. )

I will try a slightly safer approach as in https://forums.macrumors.com/threads/evga-powerlink-with-amd-gpus.2166917/

Fair play, it sounds like the 1080 Ti will run with 2x 6-pins + a SATA power cable then.

I did say the Vega 56 is on the limit. But according to your Tom's Hardware link, if you use BIOS 2 (on cards that have it), the power draw is lower (with perhaps a small loss of performance). Don't know if it's possible make sure the card is in the 'Gaming Balanced' profile rather than 'Gaming Turbo', whilst in macOS? Perhaps if set in Windows, it persists? Assuming it can be set reliably, power consumption seems to peak at just over 200W.

 

[mikas]

My impression is that the Vega 56 is on the limit for powering from internal connectors

Yes it's on the limit.

I have been using PowerColor Red Dragon Vega 56 (low bios mode) in 2 Mac Pros for over a half years now. I power the card(s) with 2x mini 6-pin + 2x SATA power. No shutdowns with my usage: 3D work + TwinMotion. TwinMotion is an Unreal Engine based architectural 3D visualisation tool. It will stress the GPU to it's upper limits according to apple system monitor. The GPU usage bar is all saturated in the graph while computing the renders.

With low bios mode it draws most of the time approx 180 watts total. It does spike momentarily at 210 watts, which is something to consider carefully. This (210W) is what you should aim at when designing the power feed and/or power draw balance. You won't know from where your card draws all the power at times. You have to meter it, or learn from others with same kind of card and/or software usage combination.

You can see numbers and measurements there:
#329 (original install in late january)
#32 (watt monitoring end of march)
and now this post (today with no GPU shutdowns for a half year)

If only 2x mini 6-pins there used, even if balanced with EVGA PowerLink, it just might have shut down in the middle of rendering. That's because the draw would have been over 105 watts per 6-pin connector (210,7W / 2 = 105,35 W). They say the shutdown limit is in the ballparks of that number or maybe max at 110 W.

This all could be card specific behavior, 3D / game engine specific etc, and it probably is too. I strongly recommend the Pixla's mod too like most of the other posters do, though I haven't done it myself. Not yet, because this arrangement works for me now. But I will do the mod eventually. I've got the cables ready, only need the right clips.

So if your choice would be a Vega 56, I personally would advice in this order:

1) Pixlas mod (you can use any other card too with this)

If low power bios selected (there is dip switch in most of the cards, but not all)

2) 2x mini 6-pin to 8-pin and 2x sata power to 6-pin
a) with power link in between (lots of cabling needed)
b) without PowerLink (my setup as of today)

3) 2x mini 6-pin
a) power balancing with powerlink
b) power balancing with cable balancing (use forum search for this, there are good advice)

I wouldn't use option 3 myself if I had free sata power connectors, and at least not without some kind of power balance. Without power balancing shutdowns would become inevitable at times. Or worse still, maybe something will actually burn.

ps. I did study this forum carefully before daring to try this, so thank you to all you early birds !

 

[mode11]

Interesting. Did you explore undervolting and/or slightly underclocking your Vega 56? There would be a slight performance decrease with the latter, but it seems that backing off the voltage and clock speed even a little can have quite a big reduction in power consumption. I think AMD cards in particular seem to come from the factory with more voltage than is necessary. It may be that your Low BIOS is already doing both of these, however. The other thing is that presumably this all has to be done in Windows, then trust the settings are retained.

At the end of the day the Pixla's mod isn't that complicated, so is probably makes sense to just take a deep breath and do that.

 

[h9826790]

Interesting. Did you explore undervolting and/or slightly underclocking your Vega 56? There would be a slight performance decrease with the latter, but it seems that backing off the voltage and clock speed even a little can have quite a big reduction in power consumption. I think AMD cards in particular seem to come from the factory with more voltage than is necessary. It may be that your Low BIOS is already doing both of these, however. The other thing is that presumably this all has to be done in Windows, then trust the settings are retained.

At the end of the day the Pixla's mod isn't that complicated, so is probably makes sense to just take a deep breath and do that.

Both AMD and Nvidia GPU shipped with way higher voltage than they really need.

My Radeon VII can be downvolt from 1093mV to 981mV @1801MHz. (~11% higher than required)

My 1080Ti can be downvolt from 1062mV to 975mV @1949MHz. (~9% higher than required)

Of course, for 1080Ti, I can only downvolt in Windows. But for Radeon VII, I can downvolt it in both Windows and macOS.

 

[mode11]

How is a Radeon downvolted in macOS? Would this work for the Vega 56, or just the Radeon VII? Would a 'Low' BIOS already do this, and even if so, could it be further optimised by doing it manually?

Does a 10% reduction in voltage give a 10% reduction in power, or is the effect greater than that? I know P=IV, but I think I read somewhere that power's actually proportional to voltage squared in this type of situation.

 

[flyproductions]

I think AMD cards need it to activate hardware acceleration for video.

Yes, but NV cards (the ones in question here) are not capable of it at all.

 

[flyproductions]

H264 and H265 4K video is accelerated for encode and decode with Polaris (or higher) AMD cards in Catalina, if you're running the OpenCore bootloader (which is well worth it in lots of ways, including boot screen support and a boot picker).

Didn't this also work for mojave and even without OC just with the right set of kexts loaded?

 

[mode11]

Quite possibly, but OC makes it easy to make sure that the right set of kexts are loaded. There may be other ways of doing it (there's a thread in this forum about activating AMD video acceleration, but I believe older methods had certain side effects).

 

[h9826790]

How is a Radeon downvolted in macOS? Would this work for the Vega 56, or just the Radeon VII? Would a 'Low' BIOS already do this, and even if so, could it be further optimised by doing it manually?

Does a 10% reduction in voltage give a 10% reduction in power, or is the effect greater than that? I know P=IV, but I think I read somewhere that power's actually proportional to voltage squared in this type of situation.

How to - downvolt a Radeon VII

Just upgraded from RX580 to Radeon VII (it's just $350 from a local seller, like new condition, with 2 years+ warranty. I really can't skip it :D) Anyway, as usual, all AMD GPU are seriously overvolted from factory (Default setting in Wattman shows 1801MHz @1093mV). Properly downvolt the card...

forums.macrumors.com

For Vega, you can use VGTab.

 

[h9826790]

Didn't this also work for mojave and even without OC just with the right set of kexts loaded?

That's correct.

In fact, from my own test, it seems Big Sur allow AMD GPU to provide HWAccel natively on cMP without zero 3rd party kext / parameter injection / etc.

Of course, "how to install /run Big Sur on cMP" is another matter.

 

[h9826790]

Quite possibly, but OC makes it easy to make sure that the right set of kexts are loaded. There may be other ways of doing it (there's a thread in this forum about activating AMD video acceleration, but I believe older methods had certain side effects).

We can manually edit the AppleGVA.framework to get HWAccel. I can't quite remember if there is any side effect. But HEVC encoding won't work due to lack to PCIe path name patching.

 

[sanfrancisofont1984]

So I installed the 1080 Ti card into my cMP 5,1 and nothing would show up in the screen. Latest High Sierra. The boot ROM is 144.0.0.0.0 (since I upgraded to Mojave already).

Am I supposed to install Web driver before using the card? Or am I supposed to stick with boot ROM that comes before Mojave upgrade?

I didn't quite have the power setup I mentioned. The reason being the SATA cable I got is a bit too tight. 2 mini 6 pin on the board + PowerLink should be enough to boot to desktop, right?

 

[mikas]

Did you explore undervolting and/or slightly underclocking your Vega 56?

No I did not, only low power bios switched on.

I have bought EVGA powerlinks, but not sure if I'm gonna use them after all. If I am, I'm not gonna feed my 2x sata-power through the power link though. My watt measurements suggest to me they would do better left alone to feed the 6-pin plug at the card (32 W). I don't want to draw 60-70 W from 2x sata power.
Reason for that:
I take sata power from optical drive power connector, straight from the motherboard pin, with a spare optical cable+ eBay 2xsata power to 6-pin. I plug out the present cable and plug in the spare one, so I don't have to disassemble the motherboard to route the cables inside the machine. The plug is a 4-pin plug only, and if there are 5V and 12V lines, that leaves only one pin pair for 12V. There are more pairs with PCIe power plugs. I'm not sure about the volts though, maybe somebody else does know about it?

 

[sanfrancisofont1984]

Am I supposed to install Web driver before using the card? Or am I supposed to stick with boot ROM that comes before Mojave upgrade?

So installing Web driver on HS made it work for me. My system seems stable using PowerLink without additional power. I tested with Heaven, Luxmark and CUDA-Z Heavy Load Test Mode.
EDIT: Grammar.

 

[r6mile]

So installing Web driver on HS made it work for me. My system seems stable with PowerLink but without additional power. I tested with Heaven, Luxmark and CUDA-Z Heavy Load Test Mode.

I use a GTX750Ti which also requires Webdrivers. However I've found that a PRAM reset, or even just a normal reboot sometimes, can for some reason default to the OSX drivers which means I get a black screen, so I have to screen share from another Mac to switch back to the Webdrivers. Very annoying!

 

[sanfrancisofont1984]

I use a GTX750Ti which also requires Webdrivers. However I've found that a PRAM reset, or even just a normal reboot sometimes, can for some reason default to the OSX drivers which means I get a black screen, so I have to screen share from another Mac to switch back to the Webdrivers. Very annoying!

Thanks for the note. I should get my screen sharing set up.

 

[sanfrancisofont1984]

Yet another thing that requires Catalina: https://support.apple.com/en-us/HT210220