And that will be the end of the Mac.
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 Waiting for Mac Pro 7,1 A1991 (no more)
- Thread starter Mago
- Start date
-
- Tags
- macpro 2016
- Sort by reaction score
And that will be the end of the Mac.
...but Apple doesn't seem to really care about performance on their computers. ("Designed to Throttle")
Its Intel CPUs which are designed to Throttle
.Somehow when Skylake launched the Throttling was almost unseen on MBPs. Throttling started from Kaby Lake, and Coffee Lake only increased it. Why?
If you design chassis for years to come, you assume certain power envelope in which processor can operate at most optimal settings. You cannot do anything if your sole supplier chips are breaking their own specs.
One could easily enforce the specs. One of my HPE ProLiants has the following management settings:Its Intel CPUs which are designed to Throttle
Somehow when Skylake launched the Throttling was almost unseen on MBPs. Throttling started from Kaby Lake, and Coffee Lake only increased it. Why?
If you design chassis for years to come, you assume certain power envelope in which processor can operate at most optimal settings. You cannot do anything if your sole supplier chips are breaking their own specs.
Also your argument ignores other "fail to perform" issues like why the MP6,1 is still stuck using E5-x6xx v2 CPUs - when Intel has introduced three newer generations of Xeons.
The 6,1 is still using those old CPUs because of its fundamental design flaws. If it had been designed from day one with rather higher thermal limits (it might even have handled its existing specs properly that way) it would most likely have been upgraded to similar or higher specs than currently used in the iMac Pro, and it might still exhibit throttling- either through Apple's design compromises, Intel's, or both.
Yes, as I said, "...but Apple doesn't seem to really care about performance on their computers".
Koyoot's diatribe about x64 power consumption is missing an important fact.
That "fact" is that by definition "turbo" mode is based on the idea that "if the temperature of the CPU is cool enough, the CPU will exceed the base clock and power".
I think that there is some confusion about the meaning of "turbo" and "throttling".
The turbo clock speeds are peaks based on CPU temperature. If you have a single core task, it's common that it can run at full turbo boost without exceeding the CPU TDPlimits.
If you have a multi-thread task on a system with barely adequate (for the TDP) cooling, then the CPU may throttle back to the base GHz. If the system has good cooling, then the CPU may stay in the turbo zone even if it's above the TDP.
Why artificially cut the wattage when a CPU has adequate cooling? Why does Apple often fail to provide adequate cooling?
(A: Thin. Form over function.)
Last edited:
Not arguing with that. I was actually agreeing, it was indeed designed to throttle, in multiple senses. Judging by the number of people who had 6,1s repaired or replaced due to video cards dying, that thermal corner they painted themselves into was very small indeed. The unified thermal core was designed for one fairly hot CPU, two fairly cool GPUs, and that was it. Too hot a CPU= overloaded. One GPU, even with the same power/heat output as the two= unbalanced, and therefore overloaded. Couple of Vega 64s= where's the fire extinguisher? Oh, and a 450W power supply, so even if the cooling was adequate there wasn't the headroom to power more demanding GPUs. For me, and perhaps many other users, it was a great concept- the cheesegrater's a great big box, but it's also a great big heavy box, not helpful any time you need to move it around, so building something smaller and (much) lighter wasn't a bad idea. Unfortunately, the execution was so flawed.....
Of course they can enforce specs.
But thanks to stupid policy of Intel breaking specs, so their marketing can put the pants back on, it would be a disaster from marketing perspective, for any company that would stick to specs.
Again, we are talking about a 6 core(8700K), inefficient CPU running in 45W thermal envelope. How high clocks you would maintain there? 2.8 GHz?
8 core, Skylake(Coffee Lake Core i9-9900K) CPU requires 125W of power, and equally capable cooling solution, to run at spec'd 4.7 all-core Turbo. The clocks and core counts are what is selling parts. Try running it in 95W power and cooling envelope, and you suddenly wake up with just 4.3 GHz All-Core Turbo. What message that does send to buyers?
Its all marketing. You seem to forget about it. And its not just Apple, but every other company.
That it's not a heavy Apple tax (at least for the base configs) isn't all that surprising. The question is whether Apple will update it rapidly when the new chips are available.
The 2013 Mac Pro and 2006 Mac Pro before it were very price-competitive with comparable PC workstations (with the caveat that there's more potential options on the PC side, especially with the tube Mac Pro.) That price competitiveness eroded with each update (or lack thereof.)
If they aren't going to regularly update the iMac Pro, then they haven't solved any problem learned from the tube Mac Pro, which was ultimately a failure not because of its more narrow focus, but overarchingly because they just didn't update it.
The thing with iMac Pro is that no matter what it will be doing, soon Mainstream platforms will have better CPUs, or not worse, than those in iMPro...
16 cores vs 18 cores. 8 core CPU with 28 PCie lanes, vs 8 core CPU with 32 PCie lanes...
Its amazing how technological progress is going lately. Apple is starting to follow, not set the tech trends.
16 cores vs 18 cores. 8 core CPU with 28 PCie lanes, vs 8 core CPU with 32 PCie lanes...
Its amazing how technological progress is going lately. Apple is starting to follow, not set the tech trends.
Last edited:
I gave up waiting and purchased a iMac Pro 10 core. Now 2 weeks old. I ran out of USB ports day one with no way to plug in all my accessories so I upgraded to a Caldigit TS3 Plus which now only has 2 USB 3 ports free. However it runs great and I am relieved that I have moved forward. I was selling my cMP 5,1 but now I have to keep it until I can update all my older perfectly good Firewire drive drives. There is no way to connect old stuff (except USB) to the new world. Since I work with clients who do not have the latest Thunderbolt 3/USB 3.1 macs I have to keep the old and new working.
So far the only issue with the iMac Pro is the 3Dconnection space mouse dongle loses connection on restart but works fine if you pull it and re-insert it.
So far the only issue with the iMac Pro is the 3Dconnection space mouse dongle loses connection on restart but works fine if you pull it and re-insert it.
If you can do what you want on a G5 now, I don't see how any pro Mac was ever worth the money for you, nor do I understand how ARM processors would materially affect you.
I do a lot of audio work in Logic. My current pro was handed to me as a gift, but I still have old copies of my softsynths and plugins on my G5 back home in the states, and I could happily go back to it. In all honesty, aside from a few ease of life features, the newer versions of DAWs haven't added much feature wise anyway, aside from programming bloat...
In all fairness, the difference is here is that I already invested years ago into a G5 setup, so a lot of what made the G5 'pro' at the time, I'm already accustomed to and would be able to use. But outside of audio, my computer usage is mostly pretty lightweight, except that I enjoy tinkering. You can't tinker with a SOC.
... and since ARM processors likely means SOC, which likely means no expandability, therefore eliminating my ability to have loads and loads of storage and backups, my secondary use for my Mac Pro. At the present moment I use it for two things, my audio workstation and my backups.
I should add: You technically CAN tinker with certain kinds of SOCs, like raspberry pis and the sort, but it simply isn't as fun. Part of it is simply hobbyist enjoyment.
Actually, the point of the video is that after one year the iMP is running much faster than after launch.
Apple is pushing a lot of their narrative over the integration of sw and hw, mostly as an excuse to build thermal throttled machines and show they run almost as big ugly ones. Yet, it seems to be working, at least on the iMP.
Apple is pushing a lot of their narrative over the integration of sw and hw, mostly as an excuse to build thermal throttled machines and show they run almost as big ugly ones. Yet, it seems to be working, at least on the iMP.
Technically no. Far more likely there were no CPU updates because Apple didn't do any upgrades of CPU+chipset. Xeon E5 16xx v3 and v4 could have readily been used in an updated system. Pointing at the CPU being a principal blocker is fundalmentally flawed.
The following link will pull up the v2 , v3 , v4 versions of the 1680.
https://ark.intel.com/search?q=xeon+e5-1680
They are 120 , 140 , 140 W TDP respectively. The majority of the +20 of the v3 and v4 is because Intel moved more power management onto the package. In other words most of that +20W is sitting right next to the v2 option in the Mac Pro 2013 (6,1). Same thermal core would be principle mechanism there too.
Intel's Xeon W is also in the same !40W range, but by that time Apple was already onto the idea of the iMac Pro. ( They were putting lots of work into that solution.).
Even if Intel fab progression process hadn't gotten stuck the workstation CPUs probably would be still in the 140W range ( e.g., just add more cores/cache/etc until get to 140W .. the bottom of the line up might run a bit cooler but the mid-high range would stay around the same target with just "more". )
If there was a bigger envelope they could have handled some of the other components that had an uptick in TDP but that was primarily not the CPU options. ( there were some exotic > 12 core options that were outside the 1600 class that can try to point to as 160W boogey man , but Apple didn't particularly need those. )
Apple could have pushed the literal desktop Mac Pro design to an upgraded literal desktop solution, but wouldn't have left much room for the iMac Pro. Apple got a far amount of push back that some people didn't want a literal desktop computer. Some were OK with it. Pushing two solutions at the latter group ("OK with it") probably wouldn't work.
There were several things that could have incrementally fixed ( incrementally open up the thermal envelope ), but Apple choose not to do that. Or open up a second track to run in close parallel with the iMac Pro development.
Even without the iMac Pro the other apparent Apple OCD component is that it could just be incremental updates to the CPU. The SSD , the GPU , Thunderbolt ... there seems to have to be some grand conjunction of several components to move forward.
[doublepost=1548110992][/doublepost]
Again... not really. The CPU is in the 120-140 range. Three times 140 is 420W. Mac Pro 2013 Power cap 450W. In fact, Apple's comments about the design corner were:
" ... The triangle you mentioned, the thermal core, is designed to have three fairly similar loads – similarly balanced in power. And so the overall size of the product and the fan, that defines the overall thermal capacity for the enclosure. ..."
https://techcrunch.com/2017/04/06/t...-john-ternus-on-the-state-of-apples-pro-macs/
The problematical issue was far more so that the GPUs were hotter than the CPUs. In Fact if look at the TDPs of the 7000 series https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units#Radeon_HD_7000_Series
D300 -- Pitcarin based -- nominal 130-175
D500 -- approx Tahiti LE --- nominal 185
D700 -- Tahiti --- nominal 240-260
Apple was not running their versions at the nominal rates but neither were the vast majority of the workstation Intel CPUs up in that range either.
Were Apple more went off the road was
"... We designed a system that we thought with the kind of GPUs that at the time we thought we needed, and that we thought we could well serve with a two GPU architecture… that that was the thermal limit we needed, or the thermal capacity we needed. ..."
What Actually happened with the AMD GPUs is that they went even higher TDP. Somewhat similar to the swamp that Intel is currently bogged down in AMD's strategy was deeply reliant on process improvements to allow to offer "more" in the same power zone. When there was a big hiccup from 28nm to the next iteration the 6,1 design ran into a 'wall'. AMD didn't get off of 28nm until 2016 timeframe ( which at that point the iMac Pro was probably being spun up. )
Apple could have gotten around the issue. One GPU and putting some addition thermals in the other slot. That would have offended the "symmetry" OCD design rules. A 200-250W GPU with more surface area and more air flow would work in approximately the same design.
But again a one workstation CPU and one GPU literal desktop wouldn't leave much room for the iMac Pro. The iMac Pro keeps a far amount of the same thermal constraints.
If they really wanted to do something to "un patin themselves" into the corner, one significant factor would be to not constrain themselves to the "Mini or less footprint" , literal desktop constraints.
It is pretty likely that Optical Drives and 5.25" bays will be gone. Weight and volume for that will probably be pruned. Apple has nuked HDDs from the Mini and iMac Pro. Four 3.5" bays .... probably won't happen either.
There is a balance Apple needs to find with the next Mac Pro in that it is not focused on being a container of lots of things ( something Apple doesn't want to do) and not containing enough to be flex used in a number of contexts. ( more than one storage drive. more than one "card" (whether 2nd GPU or something else), more than one bank of DIMMs slots. )
[doublepost=1548112252][/doublepost]
Its Intel CPUs which are designed to Throttle
Somehow when Skylake launched the Throttling was almost unseen on MBPs. Throttling started from Kaby Lake, and Coffee Lake only increased it. Why?
Their fab progression is a major contributing cause. Intel pushing what were 4 core baselines out to 6-8 cores is another factor. However, this is a bit of the pot calling the kettle black since AMD's GPU were bogged down in largely the same swamp from 2013-2016. It isn't immediate doom.
Intel's power envelope for the workstation class really hasn't significantly changed since 2012-2013.
Really hasn't beeen a Kaby Lake or Coffee lake workstation class solution. The upcoming derivatives on Cascade Lake probably won't be much higher the range it has been ( unless Intel is desperate and just overclocks everything across the board. ). They have other options to lean on. ( and performance to get back by getting rid of OS fix-it kludges for security holes. )
Last edited:
Those are desktop processors, not workstation/server processors. Wrong chart.
Hasn't changed, you say?
Hasn't changed, you say?
Does Xeon E5 or Xeon W appear on that list? No. the TDP limit isn't moving (which is different that the power consumption the graph is primarily charting). Is there a 165W TDP item on this list of options
https://ark.intel.com/products/series/125035/Intel-Xeon-W-Processor
Again, no. There are overlaps with the dies used in E5/W solutions so can pull some useful information from you attempt at misdirection.
First, from the Core i9 7900X on up higher on the diagram the marked TDP is 140W and all except the 7900X neatly fits under the that limit. So no, a properly design Mac Pro 2013 that took the TDP and added a 5% safety margin ( 147W) would have worked out just fine. Yes Intel's CPUs have been creeping closer to the 140W threshold mark, but no they haven't been going significantly over the line. The 149W of 7900X would basically performance pretty close to it marks if had a hard firmware limit set at 147W max power draw. (e.g., the iMac Pro does more than reasonably well with it its clock limit down below 140W. )
If what you are arm waving about is that the actual consumption has shifted up , shouldn't be the more material aspect. Apple should not have designed the Mac Pro core design around the specific instance of just 1-2 CPUs and 1-2 GPU of a specific implementation generation while completely ignoring the TDP guidelines. The TDP guidelines are there so that your system can work with a family of products of the same class. If Apple was planning to use the MP 2013 baseline design for next 8-10 then molding the design tightly to just an extremely narrow snapshot of instances would have been a thoroughly bozo move.
Second, the 7980XE is a corner case. That isn't the whole line up. The High Core Count dies that Intel pulled down into the W series can't base clock at the highest limits in the classic 140W TDP range. For the Skylake X 'refresh' Intel goosed the base clocks higher, which will run them over the edge of the TDP, but these are all 'unlocked' CPUs anyway so more than a few folks will be running them outside the TPD range anyway. Apple's standard procedure it not to run stuff way outside the TDP range. So pointing to stuff they won't pick is pointing at a corner case that likely won't exist.
That's is why for the Xeon W they just don't run the clocks that high when have all the cores cranking.
i9 7980XE ( 18 cores base 2.6 GHz / T 4.2GHz) 165W ---- vs ---- W 2195 ( 18 cores base at 2.3 / T 4.3GHz )
i9 7960X ( 16 cores base 2.8 GHz / T 4.2GHz) 165W ---- vs ----- skipped by W class. (no power bust there at all possible)
i9 7940X ( 14 cores case 3.1 GHz / T 4.3GHz) 165W ---- vs ---- W 2175 ( 14 cores base at 2.5 / T 4.3GHz)
the i9's have a "Intel® Turbo Boost Max Technology 3.0 Frequency ‡ which is 4.4-4.5GHz which W's don't have. Which is something else to bust the TDP limit when coupled to overclocking. Again not present in the W/E5 line up. The W options have a better than i9 nominal turbo mark ( so still more than good on running mixed user interaction workloads. ), but don't chase that for maximum power draw past limits. And when get into the comfort range of the HCC die (just 14 out of 18) even the turbos even out.
When pushed hard, the W 2195 probably technically 'busts' 140W with a marginal cooler, but also probably not in the 190W range either.
Third, as I said before which didn't get quoted at they core count goes up but the process fab process doesn't progress they are going to run closer (to a bit over) their TDP limits.
Finally, If you want to bring the discussion back to what would have been a "Power Buster" of the Mac Pro 2013 (6.1) design the Threadripper is an every bigger bust than any of the E5 or Intel W options that came along at 175-180W budge for a 140-147W budget limit. That's present on that chart too.
AMD's Workstation was a power bust once they eventually got something and the AMD GPUs were power bust from 2013-2016.
Since Intel is using those high end i9s to compete with the Threadripper far more than the Ryzen options there. that's partially what is pulling them off the 140W line. [ And yes some folks will use those in what they'll call a "workstation" but Apple hasn't chased the overclocking, max power consumption crowd before ... and aren't likely to in any revised Mac Pro either. ]
Did Apple need 25-30% margins around the CPU and GPU TDPs (e.g, could have run up from 140 to 180 W ) ? That's a judgement call. On the one hand, if give the component vendors that much slop they can get careless. ( 'brute force' when design way into a corner. ). On the other hand waaaay too tight isn't good either (like under the TDP). At an empty PCI-e slot that Apple doesn't 'control' that would be good. At the components they do select the tolerances will probably be much tighter than that.