Even so, I don't think that's a big number. I could see if it was thousands...but I understand your point.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
MP 7,1 Mac Pro 7.1 owners! What would 8.1 need to be to get you to leave your 7.1 behind?!
- Thread starter maikerukun
- Start date
- Sort by reaction score
Sneak Preview date was June 2019. Not really a "launch" date.
Pragmatically nobody could even order a Mac Pro until December 2019 . Let alone ,buy Apple didn't fully reveal all the BTO prices until the late Fall. ( starts at and a few other tidbits like a $999 stand, but not the full pricing line up. Many folks didn't even know how much money they 'needed' to spend for months. )
New Mac Pro and Pro Display XDR Now Available to Order
More than two years after Apple promised a new modular high-end desktop machine for its professional users, the new Mac Pro is now available for...
www.macrumors.com
Lots of folks didn't even receive on until January/February 2020.
Mac Pro entry priced increased 100% ( ~$3K -> ~$6K ). There are a significant number of folks who have "overbought". Lots of hype on forums and elsewhere about "worth the cost to future proof " and some folks probably overspent. (many have not, but substantive numbers have and there is no 'unexpected future windfall of revenue to pay for it'). Throw in the pandemic which disrupted lots of businesses and there has been a constant flow of folks trading out of Mac Pros ( e.g. Apple's refurbished listings never go 'dry'. For example, right now Apple's USA refurb listings for Mac Pro spans two pages on Apple website. Mini one page . iMac about 1.4 pages; a bit less if drop the 24" model. 3 pages of iMac Pro. 5 pages of MBP )
A couple 100 isn't relatively much abnormal churn for the Mac Pro. Tax write off bubble is probably not the main driver here.
If folks are selling them at high prices then they probably do not have large depreciation write offs. ( Or are cheating on taxes ... because if sell more than book value that is now income and gets taxed. ) If is a contributing factor in systems being sold off , but not a bubble supply driver.
That carries a huge presumption that folks can only sell off a Mac Pro to get another Apple Mac product. That is probably doesn't hold up. There are folks leaving ecosystem also. Some companies laying folks off ( so don't need extra 'seat' ) . Anyone spending 70+% of these time bootcamped over in Windows running a Nvidia card ... the GPU pricing crisis is largely over now. ( don't sell your used car unless know can buy a decent priced new car). New Threadripper 5000 systems are available from multiple vendors at this point.
The other part is that massive global recession is hyperbole. This isn't 'massive' in many places. Folks can't spend money like drunken sailors anymore, but GDP dropping to negative 1-4% isn't massive. If the Mac Pro was a revenue producing tool then selling it off only makes even less money. Large capital sell offs in the face of a single digit dip more likely indicates overspent in the first place. ( and rising economy covered up the flawed decision making. )
Powered off with the shell off how many Mac Pro users make money in that context ?
Does using the tool to produce work produce revenue higher than the costs of using the tool? Yes or no. That is the metric on why would upgrade to some. Upgrade for upgrade sake does not make profits for most users.
The flawed presumption here is that most folks only bought a Mac Pro for the upper half ( > $1600 ) GPU options. Studio doesn't cover the whole GPU performance range, but the bottom 50% (which was most commonly sold Mac Pro configurations)? It does.
Ultra 20CPU / 64GPU / 128GB / 4TB --> $6,799
MP 2019 16CPU / 192GB / W57000 / 4TB --> $12,599
(MP 2019 16CPU / 32GB + 6 *32GB 3rd party / W57000 / 2TB + 2TB 3rd party --> $10,699 )
If sold that 2019 (7,1) for 5,799 then could get a faster system for $1K. If sell for $6K then for $800. If ROI on $1k comes in 2-3 months then it is basically paid for working tool for next N years.
Pro Vega II ( and Duo ) and W6800 (and Duo) MPX modules is not the most popular sold units for Apple. ( Apple said as much at the Studio intro). Same thing with the high RAM tax Intel CPU options ( 24 and 28 core ) .
Apple marketing never said that the Studio was a full replacement for the Mac Pro BTO options. The performance definitely overlaps. And the M-series laptops overlap where much of the legacy Intel desktops were . The Studio is also fending off the mainstream 16 core PC with a mid-upper range card that is far more common now than 3-4 years ago.
What the Studio is not going to go is get the bulk of folks still clinging to 5,1 systems to move. That is likely just as much as bout costs than it is about performance. ( buy non bleeding edge system now and pour money into it over time. Not really pressed for bleeding edge performance in the short term. ). That is where a significant number of chants about upgrade worth more money than performance comes from. That probably never really was the core of the Mac Pro market that Apple was interested in the first place ( even back in the 3,1, 4,1, 5,1 era ).
There is little solid , open evidence that W7800X are "just around the corner". Apple/AMD haven't even covered the 6x50 GPUs that have been out for a while (not quite a whole Quarter , but also not much of a change either; basically overclocked tweak to design as a gap filler. ) .
Even if Apple works with AMD to do a W7800 card it would likely the mainstream release of W7800 by 4-6 months. The plain 7800 isn't coming real soon. If have a 3-4 timeline that would be 'soon' , but generally not.
Similar 7800 Duo may or may not happen. Nvidia 4090 is faster but also repordly jump way up on power consumption to get there. There are several reports of TDP creep for the upcoming AMD solution . No where near as high as Nvidia , but higher. There is no 6900 duo. If the 7800 edges up into the 6900 TDP zone don't know if going to get two of those on a single MPX module.
Apple is not likely going to get into a "more power!!!!" race with Nvidia ( or AMD). Everything Apple has openly stated is about how that is not the future for them. Also extremely likely not to get into a "more power" war with the Mac Pro 7,1 because it already pulling max normal wall socket power. There is a limit on MPX power provision. Going up into the 600-700W range isn't going to happen. Nor is out of PCI-e slot spec too tall or too long cards.
Which generation Threadripper. 3000 or 5000 series? 6000 ?
0
Don't hold your breath.
Highly likely not coming to any Apple Silicon system that has Thunderbolt build into the SoC. About 2/3 of MPX is a provisioning solution for Thunderbolt. That 'issue' is basically gone with Apple SoCs.
Last edited:
You didn't answer the question lol.
S**t. I read that as ‘post production society collapses’
If the 7800 is at all competitive with an Apple Silicon Mac Pro - Apple won't ship support. They'll want to maximize their "X times faster!" graphs. Tossing a W7800X into the mix is going to make those graphs a lot less dramatic. Worst case - if the 7800 is faster in a single or duo configuration, Apple won't want reviewers talking about that in new Mac Pro vs old Mac Pro comparisons.
If Apple is doing a 4-tile/chip solution then that is pretty likely backwards according to recent rumors.
First, that the Mac Pro starts off with an Ultra (2 tile ) and only gets to Extreme for the upper 'half' configurations. Second that the M2 GPU core count bump isn't that large. The Ultra only competed with a 6800/6900 in " rolling downhill in a hurricane" contexts. Second, 7800 could pretty close to double up on the 6800 performance (much different memory subsystem with few of the NUMA aspects Apple interjects, bigger caches , and updated cores. ) .
Question is whether Apple Silicon could be competitive with the 7800/7900. The 7900 will likely be better. If there is a 7900 die with Infinity Fabric then double card of that will definitely be for highly parallel loads no matter what Apple does.
Then they just don't put them up. Apple didn't put up W6800X Duo graphs against the Studio Ultra for the workloads the W6800X Duo shines at. There is only a limited number of graphs they need to up up over a subset of workloads. On a CPU core only raytrace render a 40 Core Mac Pro would likely decimate a 28 core Intel Mac Pro. If several developers have had 8-10 months to implement the optimizations that Apple covered at WWDC 2022 then there will probably be at least 6-8 extremely fine tuned Apple Silicon specific apps (and associated workloads) that Apple could pick that would be a large market differentiation between the new Mac Pro and the old one.
If Apple commissioned a W7800X then they'd want to sell it to at least a limited set of Mac Pro users squatting on Intel systems. . Somebody at Apple is drinking a giant bucket of kool-aid if they don't think that next Mac Pro isn't going to get measure against a Threadripper W5000 and a 7800/7900 ( or two) on DaVinci Resolve , Adobe , and several other multiple platform apps. They better bet is to control the narrative to some workload their relatively ancient CPU in a 7,1 is faded to the background and it is a more GPU battle (where probably pretty close if using the same basic silicon for the GPU package. )
All the more if they come out with a "half sized" Mac Pro with something like 1-2 PCI-e v4 slots. The notion that the only primary differentiation between the old and new Mac Pro is primarily solely going to be on GPU "horsepower" is likely not true from everything leaked so far. There is lots of talk in this Mac Pro forum that goes down the rabbit hole that maximum CPU core count and GPUs are the only significant things that matter. It is a multiple dimensional subproduct differentiation issue.
There have been several posts over the last 2+ years since the Mac Pro 2019 launch with users with Nvidia 2000/3000 series slapped in them to run over in bootcamp Windows context. Is the Mac Pro 7.1 going to loose it bootcamp ability when the Apple Silicon version comes out.
Apple is all too keen for reviewers to stack up the latest Samsung , Google Pixel , etc. phone up against iPhone when it comes out. Why are truly independent reviewers going to duck general comparatives for the Mac Pro.
If Apple doesn't do a 7800/7900 product is far more likely driven by costs than by website sales comparatives . It costs an order (or two) of magnitude less to basically "do nothing" on the current Mac Pro ecosystem as far as configuration changes/updates. If all Apple has is a 1-2 slot "wonder" box then they will herd a subset of Mac Pro buyers into a older, very expensive box they probably would not touch otherwise at higher margin for Apple. And other substantive subset are just going to walk away from macOS. As long as there is growth in the rest of the Mac product line up they probably won't care ( swapping people who do like their products for those that don't. ) .
If doing cost comparative if Apple prices their solution $3-5K more than the alternative they fail that way also. ( the MPX mark up tax is going to have a problem in with price corrected GPUs in the mainstream market. Same with markup for Apple Silicon )
I think a W7800X is a money loser for Apple. High investment, new GPU drivers that have to be maintained, not many sales. I don't see Apple commissioning a new set of GPU drivers (along with the required multi year OS release maintenance of said drivers) just for one machine. The 6000 series GPUs were used across the lineup. A 7000 release would be substantial investment on multiple fronts for a subset of a subset of users.
Can't see Apple going for it.
Best case is that maybe AMD gets together with a third party and convinces Apple to take the driver changes. That way Apple could relegate a 7000 series card to a third party level of support.
Apple's answer is going to be "the new AS Mac Pro is so amazing that why waste your time upgrading your Intel."
I don't know how much comparisons to PC products bother Apple. They made very bold claims about the Mac Studio. The claims didn't pan out compared to PCs. They don't seem to care. I think the only thing that's going to concern them is if a new top end Mac Pro running macOS gets beaten by an old top end Mac Pro running macOS. I think they have a casual interest in converting PC users but I'm not convinced they have a strong interest in going after the PC workstations of the world anymore.
Last edited:
Bottom line, all the excuses you give to apple, all the top end houses leave apple. Apple loses it's halo players. Halo players start questioning the rest of the "ecosystem". They find alternatives. They start recommending those new alternative/eco systems to others. And the decay to apple begins.
When the snowball is little, everyone will laugh at it. By the time it gets down the mountain, no one will be left to laugh.
When the snowball is little, everyone will laugh at it. By the time it gets down the mountain, no one will be left to laugh.
This already happened in 2013-2018. The top Mac Pro was bad and people left. Apple was fine and still is. In 2017 when Apple released the iPhone X it was very popular. The Mac line up was VERY VERY bad in 2017-2019 and yet Apple still made huge sales of the iPhone. The overall Mac line up now is great with sales increasing and strong demand for Macs.
The Mac Pro is a blip in Apple's spread sheet. Many pros have already moved on to PCs where Nvidia is king.
I know many of my friends that use PCs but still use iPhones because they WANT to. The point is the decay will not happen. This is not Apple Computer anymore, Apple's right now is mainly a mobile company. iPhone, iPad and MacBook. The desktop Macs while important are not huge money makers and not many people buy them. The Mac Pro is a niche within a niche.
I STILL want Apple to make a great Mac Pro and I know many people here would be angry about a 6,1 like Mac Pro with no PCI-e and I don't want that to happen.
The top houses have money and they will pay whatever Apple demands but it has be VERY impressive as in modular and expandable over time. Having no or barely any PCI-e slots would end the Mac Pro line for many that buy it.
Last edited:
AMD will launch Zen4 desktop processors Aug/Sep. Naturally RDNA3 GPUs have to follow after that. At the earliest it's likely to be around Oct/Nov. So I expect the launch schedule for RNDA3/Radeon 7000 series will follow the same pace as Radeon 6000 series. If Apple is going to release its own W7900x, it'll be a long wait of another 12 months from now. MacOS (x86_64) driver support may appear a couple of months earlier, perhaps in Ventura 13.3 or 13.4.
MacPro8,1 will launch Fall 2022 and start shipping early 2023. This gives Apple plentiful time to avoid direct comparison between M2 Extreme GPUs and a potential W7900x (or even RX7900XT due to lack of MacOS driver) if it does matter to Apple (which I think it doesn't because Apple has a bigger problem).
The bigger problem for Apple's GPUs is clearly demonstrated in M1 series that its design/software is not scalable:
We don't know what exactly caused the sub-par performance in higher core counts. It's perhaps lack of software optimisation. It's perhaps 'poorly' designed hardware that caused thrashing in translation look-aside buffer.
Apple was obviously targeting M1 Ultra GPU at W6900x in their planning. And the unreleased M1 Extreme GPU was planned to be equivalent to two W6900x. Both didn't work out as planned. I would speculate M1 Extreme GPU would score the equivalent of ONE W6900x in GB5 Metal. It was perhaps one of the reasons not released to avoid any obvious embarrassment.
Apple may conveniently tap into AMD's Radeon roadmap for business intelligence on where the industry is heading. By working on W7900x and W8900x is excellent convenience. On the other hand, MacPro8,1 appears not to be a direct replacement of the Intel Mac Pro. It's illogical to discontinue 7,1 immediately. GPU refreshes for 7,1 are minimal cost to Apple to hold everything together for a smooth transition and not abruptly demolishing the existing Mac Pro ecosystem.
BTW, Metal 3 support for Radeon 5000 and 6000 series seem to be another indication that AMD GPUs won't be abandoned on Intel Mac Pro anytime soon.
Not all people left. THere were some amount that stayed, partly because of the apology tour which I think was around 2017 as well. The last will leave now if apple fails them. Apple is also a much bigger ship than in 1996. It has more room to coast once the halo players leave.
I am preparing my workflow towards Windows/Linux just in case. If Apple fails to deliver a CPU is soldered and dGPUs that are weak in regards to RDNA 3 and RTX 4000 I personally will leave and I believe many here will too.
I recently brought an Xbox series S for my gaming needs/TV streaming in the living room and I am suprised at how powerful it is for $300. The Apple TV costs $199 and comes with a measly 64GB storage and does not even come close to the Series S and the Xbox comes with 512SSD as standard.
Similarly, I expect the new Mac Pro to be expensive beyond belief and there will be MUCH cheaper alternatives that are faster and better and thats why I am looking at other platfroms.
That is very telling table. I put up a quick graph and extrapolated. Everything (GPU core, EU, ALU) doubles at the same interval, but the Metal score doesn't.
M1 Extreme (Dual Ultra) would have given ~122000.
M1 Unbelievable (Dual Extreme) would have given ~144000.
So maybe that's why Mac Pro M1 was scrapped, and they are working on M2 Mac Pro now.
Or maybe it's the UltraFusion that made things problematic, and they need some another more advanced bridge layer between the SoCs.
Surely I hope they get that line straightened out and up.
And at least in the meantime, please give us dGPUs.
Last edited:
Seems consistent to me with a new architecture picking the low-hanging fruit - figuring out faster / more efficient ways to do easy stuff, and then running into a wall when the work actually becomes hard. "We found a brilliant new way to skin a cat, but we forgot to take into account the cat doesn't want to be skinned".
As unbelievable as it might be, I want to believe (the audacious ambitions, not the performance)
Last edited:
haha, exactly. Apple sillicon is the GOLD standard for mobile, tablet and laptop in pref/w and efficiency. It's awesome for these devices.Seems consistent to me with a new architecture picking the low-hanging fruit - figuring out faster / more efficient ways to do easy stuff, and then running into a wall when the work actually becomes hard. "We found a brilliant new way to skin a cat, but we forgot to take into account the cat doesn't want to be skinned".
When I move to PC for desktop later I will always watch Apple progress in the laptop space and maybe in 3 years buy an Mx Macbook. I 100% believe they designed Apple sillicon for laptops/notebooks and desktop was just an "add-on".
This can be be seen with the M1 Ultra. The Ultra is just two laptop SoC's glued. I see why Apple chose this path, mac dekstops were dead when they ditched Nvidia and Intel and modularity was not possible with Apple's designs.
One way, Apple might dig itself out of the GPU bottle neck is perhaps go the dedicated accelerators route, thereby circumventing certain advantages that traditional GPUs have enjoyed.
These accelerators can perhaps find their way into MBPs, Mac studios etc.
These accelerators can perhaps find their way into MBPs, Mac studios etc.
There may be a very small window where the 7,1 is available after the 8,1 launches. But I don't think it will be very long.
Apple will want to discontinue support for Intel as soon as they can. The longer the 7,1 is on the market, the longer they have to keep shipping macOS upgrades for it. It's the last Intel Mac, and it's going to create support overhead as long as it exists. People are going to expect if they buy an Intel Mac, they get macOS updates for at least a few years. Keeping the 7,1 on the market will draw that out.
If Apple keeps MPX on the 8,1 - there could at least be a transition path where Apple ships MPX cards that work on both models for a while. But if Apple really does go a different direction with the 8,1 - I'd expect they're going to bury every trace of the 7,1 as quick as they can.
Chuckle. When Zen4 launch in Sep it will actually ship. Going to get announced in Aug and ship/launch in Sep. When Apple "launches" is it one or two Quarters before you can get it. 'Launch' and can't buy the product is P.T. Barnum misdirection spin.
As far as RDNA3 order. Decent chance that the monolithic. RX 7600 will launch qucker than the 7800. It is significantly less complex to make. The die size probably isn't that much bigger than the N4 GCD (graphics complex die) and doesn't need 3D interposers and ultra precision packaging.
Leaked info so far:
7900 Q3 '22 (probably Q4 to actually buy in substantive volume ).
7800 Q1 '23. ( again likely a slide. Even more likely if initial demand bubble for 7900 soaks up MCD chiplets).
7600. Q4 '22 ( doesn't depend upon MCD or 3D interposer or more complex packaging constraints. Still on super mature N6 and mostly not competing for N5 wafer starts. ).
AMD Navi 31 GCD (Graphics Complex Die) is reportedly 350 mm²+ in size - VideoCardz.com
AMD flagship RDNA3 GPU reportedly has a small graphics die, but there is a reason The upcoming Navi 31 GPU reportedly features 350 mm² graphics die, significantly smaller than the rumored figures from last year. According to Greymon55, AMD Navi 31 GCD (Graphics Complex Die) is only 350 mm² in...
The RX7900XT may not even be an option for the Mac Pro if the board sizes are out of strict PCI-e specs.
AMD is mainly trying to get the 7900 before Nvidia to "stick it to them" because Nvidia is probably going to struggle ( both with a larger glut of cards they dumped into the crypto-craze market that is about to reguritate on them. And also the throw perf/watt Nvidia is throwing out the window to post numbers to still capture the 'king of the hill' crown at any cost. ). The 7900 and 4090 are in a pissing contest to a large enough extent that they are largely going to remove themselves from the Mac Pro (even if Apple wanted to do drivers which probably don't. ).
Lining up the 7800 intel drivers with the new Mac Pro or not isn't a make or break timing issue. Apple shoveling the Mac Pro onto market before the the 7000 series arrives. That window is blown. Probably never was true even in pre-pandemic internal roadmaps. The notion that if Apple merely blocks the driver , that will disable any direct comparisons is goofy. If Davinci Resolve runs on both platforms then comparing. A to B is entirely possible. That is how multiple platform benchmarks work. And those multiple platform benchmarks exist.
In another thread today someone tossed out this link about how Intel was suffering in Perf/Watt against the M2.
Dell XPS 13 Plus review: XPS plus, battery minus
A killer screen that kills the battery life.
It is pure "emperor's new clothes" time (and hypocrisy) to think that when the numbers don't favor the M-series that is all going to be hidden away and the new Mac Pro is going to get a free pass (only can compare it to multiple year old previous Mac Pro configurations).
That is more indicative that there is something wrong with the Apple GPU hype train than anything wrong with the hardware. The expectation was created for perfect linear increases was lots of hope more than historical insight or industry exposure..
Broad range of Metal Geekbench scores.
The W6800X and W6800X Duo get about the same score. So scalable benchmark? Nope.
The 6800 and 6600 have a ratio of 'core' count of 1.88 (3840/2048) and score ratio of 1.62 ( 138267/85580) -13% drop across those two ratios.
The 6900 and 6600 have a ratio of 'core' count of. 2.5 (5120/2048 ) and score ratio of 1.95 (166946/85480 ) -22% across those two ratios.
There is drop off on AMD also. Can check OpenCL across AMD,Nvida,etc also. linear core inscreases don't perfectly scale with Geekbench. Not sure why there would be an expectation that there are no scalar components to even a mostly parallel benchmark.
That is highly doubtful. Because it only worked for the narrow corner case they were they 'won'. Apple execs have made comments are several points of surprise at the results got out of the M1 early on. There is some stuff that worked better than they thought. They also likely didn't have exact 6900 numbers when designing the Ultra. The Ultra's GPU cores are exactly the same as the M1 Pro/Max GPU cores which launched even earlier than the Ultra did.
They are a slightly different than the plain M1 cores but not by much.
The Pro and Max GPU cores were vastly more likely designed to take the laptop GPU targets. Whatever they "happen to cover" with an Ultra was just relatively cheap 'gravy'. (yeah had to 'pay' with ultrafusion but the additional GPU core design cost is pretty close to zero. )
Same is likely true for this upcoming 'quad'. The GPU cores probably are no different from what is being rolled out to the laptops. The laptop space is the primary target and they 'reuse' what comes out for desktop.
Piled higher and deeper flawed logic. The W6800 Duo is a bigger threat to the Extreme than the W6900X. The W6900X is more to suck the maximum money out of folks pockets. A 68 Duo is cheaper than a 69. Two dies (with some cores turned off) cost less than just one fully enabled die? Probably not. ( 69 and 68 are the same die:
just a difference in binning. )
The W6900X was an opportunity to goose even fatter profits out the Mac Pro ecosystem. It probably wasn't a target.
How? First, the systems with 6000 series in them are no where near Vintage/Obsolete stage. Apple was still selling iMac 27" models while Metal 3 was in 'beta'.
Second, Metal 3 beings some efficiencies that are being deployed (or have been deployed ... Apple playing 'catch up') in other APIs already. A direct from storage to GPU API update? Join the club ( it is in DirectX and Nvidia APIs already). Some structure to support raytracing abstractions. Again join the club... already existed elsewhere.
Better debugging and tracing/profiling support? That is something should hold back? (again is it really any giant leap over what others already provide in their interfaces? )
Keeping the Metal API mostly the same across both ranges of legacy GPUs just makes Apple's job more earily in the larger scope. Lowers port differences costs on both "sides" of the macOS port. (lowers costs for 3rd party developers also doing "universal/fat binary" building. ). The bigger issue for "support" for the Mac Pro 2019 is if Apple leaks that "zero 3rd party GPU driver" constraint of macOS on Apple Silicon back over onto the macOS on Intel side as well. Nothing new on side A so therefore constrain nothing new on side B. That is going to be a very large expectation mismatch for lots of legacy Mac Pro users.
When the 8,1 launches, I guess its details will be as interesting as the fate of 7,1 immediate discontinue or not, for how long it'll be quietly on sale..
M1 GPUs have the same GPU clock (I believe, haven't checked). I also think memory bandwidth not an issue in GB5 metal benchmark. Have a look at @mikas plot above based on my data retrieved from GB browser. The scaling tapers off logarithmically. While I don't expect exact linear scaling, the curve does indicate something gone wrong in programming guideline, hardware design..
Take a look at Radeon 6000 GPUs natively supported in MacOS (x86_64):
and the corresponding plot, GeekBench 5 Metal score vs number of GPU shaders:
AMD RDNA2 scales really well. Don't they? Go figure what's gone wrong with M1 GPUs.
What’s interesting about the geekbench link that @deconstruct60 shared is that there are three ‘dual’ GPU ‘cards’ in that list (third one is more than a GPU featuring much faster interconnect than the first two )
- Vega II/ Vega II duo
- Radeon w6800x/ Radeon w6800x duo
- Apple M1 max/ Apple M1 Ultra.
The GB5 score of the first two variety is almost identical between the solo and dual cards (= the benchmark doesn’t detect/ignores the 2nd GPU)
While in Apple’s case it offers some 45-50% improvement between the ‘solo’ and the dual version.
Are there other GPU benchmarks that can shed some light ?
Example octane benchmark suggest near linear scaling from w6800x to w6800x duo.
Maybe blender/ octane users can do some test between m1max and ultra ?
See this redshift test : the ultra is about 78% faster than the max (370 seconds vs 668 seconds)
Mac Studio and Studio Display — Prolost
prolost.com
Not quite logarithmic slow down in all benchmarks perhaps…
Last edited: