You have no idea about Threadripper and EPYC. They have different purposes. AMD does not have Xeon-W series CPU but AMD Threadripper and EPYC outperform them.
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 All Models Is Apple ****ing themselves over the ThreadRipper 3 yet?
- Thread starter macjunkie2013
- Start date
- Sort by reaction score
You have no idea about Threadripper and EPYC. They have different purposes. AMD does not have Xeon-W series CPU but AMD Threadripper and EPYC outperform them.
Of course they have different purposes, that is basically what I am saying. Don't try put the words in my mouth. Threadripper doesn't support more than 256GB of memory compared to EPYC, which supports up to 4TB each socket with Registered ECC support, making EPYC more suitable for HPC purposes. But Threadripper will operate in much higher frequency compared to EPYC, making it suitable for workstations.
With Xeon-W on Mac Pro supporting up to 1.5TB of RDIMM, it actually have a use case beyond what Threadripper system could provide, which doesn't even have RDIMM support.
Exactly, other than adding some industry specific expansion cards my employees will never open the case on their 7,1s. When we decide additional hardware capability is warranted we'll replace the units.
And how is AMD's Instruction set different than Intel's Instruction set...?
If people will switch from Apple platform solely based on the fact that it uses AMD CPUs instead of Intel it will show only one thing.
That they are stupid, and should be stripped of their voting rights, because they are incapable of making the right decisions in their lifes. /joke
There's a reason why Tim Cook is running a trillion dollar company and you are not.
If people left Apple because Apple used AMD CPUs it would show only one thing: That Apple was stupid and missed out on profits because they switched.
Kpjoslee
The problem is not all software needs that amount of RAM and Apple has only a few professional software which is a joke already. Also, Apple used to provide cheap Mac Pro around $3000. What you are thinking is for high-end users. Most software with Threadripper is already enough.
Since EPYC exists, there is no reason for using Xeon W. Higher clock speed isn't important for workstation especially comparing between 14nm vs 7nm.
The problem is not all software needs that amount of RAM and Apple has only a few professional software which is a joke already. Also, Apple used to provide cheap Mac Pro around $3000. What you are thinking is for high-end users. Most software with Threadripper is already enough.
Since EPYC exists, there is no reason for using Xeon W. Higher clock speed isn't important for workstation especially comparing between 14nm vs 7nm.
Higher clock speed has always been important for a workstation and one of its main value propositions through the years. With ever increasing core counts, clock speed, while still important, has to be balanced with heat, energy used and the practical size of the CPU which plays into the cost.
Whether or not the end user needs 1.5TB of DRAM is highly debatable, but so is the practical limit to how many cores are actually useful at the end of day. Trading clock speeds for core count isn’t the end all, be all which is what I am hearing from the majority of users here extolling the virtues of Threadripper.
Let’s bring a little reality into the debate on core counts - https://www.pugetsystems.com/labs/articles/What-is-the-Best-CPU-for-Video-Editing-2019-1633/
Threadripper and EPYC aren’t the answer for Apple...yet. Intel is right for right now, but they sure need a good swift kick in the ass.
EDIT:https://www.pugetsystems.com/labs/articles/What-is-the-Best-CPU-for-Photography-2019-1620/
Last edited:
Adobe Premiere Pro obviously is, as Puget is stating up to a 20% increase and the Export/Transcoding, the TR3 wins pretty easily, but it looks like that’s the only place Threadripper is really running away with things, which is logical when you put 18-cores up against 24- or 32-cores.
You would think though that Adobe After Effects customers would be screaming bloody murder to have AE optimized for higher core count CPUs, right? Or to offload more to the GPU. I guess Adobe figures they don’t need to spend the time or effort.
DaVinci Resolve is offloading some work to the GPU, but TR-3960X isn’t running away with it given 33% more cores over the 10980XE.
Photoshop seems to run out of juice once you past 8-cores, but this is already pretty well known from previous Puget Systems articles from years past.
Lightroom Classic CPU and passive work scores do definitely benefit from Ryzen and TR.
That being said, the proof in the pudding will be how FCP X, DaVinci Resolve and Premiere Pro interact with the new Mac Pro, the Vega II and the Afterburner card. I am incredibly eager to see comparisons between a TR3960X and a Xeon W-3265M on Premiere Pro and DaVinci Resolve workloads. I suspect it is going to be more interesting than anyone is expecting.
Even PP is not fully optimized more than 8 cores except for exporting. Go and check Adobe forum about it.
Adobe is well known for having poor optimization for a long time since CC released as they don't optimize their own software instead of adding new stuff from CC versions due to the monopoly.
Like Adobe, if the software is not optimized for higher core counts, it's not Threadripper or EPYC's fault.
Adobe is well known for having poor optimization for a lot longer than since the release of CC.
The entire reason I never upgraded from CS2 was because the product wasn't seeing much in the way of performance optimizations. The features that were added were, meh, at least for me.
The entire reason I never upgraded from CS2 was because the product wasn't seeing much in the way of performance optimizations. The features that were added were, meh, at least for me.
What's unfortunate from a Mac product line standpoint is Threadripper's modest RAM capacity. Sure, 256 GB is a lot by any normal standard - but it's modest by the standards of Mac Pro-level workstations. Most people don't need >256 Gb of RAM, but many Mac Pro customers do...
Right now, an AMD workstation requires one of two compromises...
Threadripper, which is very fast, and is even fast in poorly threaded applications, because each core boosts to a very high speed - but has a relatively modest maximum RAM capacity...
OR EPYC, which will take absurd amounts of RAM (isn't it something like a 4 terabytes per socket, and dual socket designs are possible) - but it's slow unless your application is well-threaded, because it has a low clock speed (a lot of cores, but no fast ones).
Sure, you could put Threadripper in an iMac Pro, which has a 256 GB RAM capacity - but how do you cool it (maybe a new iMac Pro case based on the XDR display design could) - and, regardless of whether the Mac Pro used EPYC or Xeon-W, the iMac Pro would usually be faster. If the Mac Pro were EPYC, the iMac Pro would clock much higher, and only applications that could use more than 32 cores would be faster on the Mac Pro (assuming a 32 core Threadripper and 64 core EPYC). If the Mac Pro stays with Xeon-W, only the very fastest 28-core is even in the running.
Right now, an AMD workstation requires one of two compromises...
Threadripper, which is very fast, and is even fast in poorly threaded applications, because each core boosts to a very high speed - but has a relatively modest maximum RAM capacity...
OR EPYC, which will take absurd amounts of RAM (isn't it something like a 4 terabytes per socket, and dual socket designs are possible) - but it's slow unless your application is well-threaded, because it has a low clock speed (a lot of cores, but no fast ones).
Sure, you could put Threadripper in an iMac Pro, which has a 256 GB RAM capacity - but how do you cool it (maybe a new iMac Pro case based on the XDR display design could) - and, regardless of whether the Mac Pro used EPYC or Xeon-W, the iMac Pro would usually be faster. If the Mac Pro were EPYC, the iMac Pro would clock much higher, and only applications that could use more than 32 cores would be faster on the Mac Pro (assuming a 32 core Threadripper and 64 core EPYC). If the Mac Pro stays with Xeon-W, only the very fastest 28-core is even in the running.
Last edited:
Yes, you are correct on the memory for Eypc.
About those clock speeds.......
W-3275: 28 cores/56 threads @ 2.50Ghz. (The non-M variant addresses 1Tb, the M variant 2Tb)
Eypc 7542: 32 cores/64 threads @ 2.9Ghz.
Added bonus - 1/2 the price of the non-M Xeon & 1/3 of the M version.
About those clock speeds.......
W-3275: 28 cores/56 threads @ 2.50Ghz. (The non-M variant addresses 1Tb, the M variant 2Tb)
Eypc 7542: 32 cores/64 threads @ 2.9Ghz.
Added bonus - 1/2 the price of the non-M Xeon & 1/3 of the M version.
The EPYC has a boost speed of 3.4 GHz, while the Xeon boosts to 4.4 GHz.
That's a big difference in a well-cooled application where boost matters more than base. The Mac Pro had better be a able to run its processor right up to boost clock most of the time - what else is all that fancy cooling that supposedly dissipates 500 watts for?!
There are situations where a laptop could be faster than that EPYC - if the application stays under 8 cores, the 9980HK's ability to boost to 5 GHz (briefly and only on a core or two), and hang out around 4 GHz or more should beat the EPYC. Of course, the EPYC will win in a well-threaded application, and by a lot in a very well-threaded application when the power of 32 cores comes into play...
My 9980HK (according to Intel Power Gadget) stays around 4.7 GHz going as high as 4.8 on one core, is around 4.6 sustained on two, 4.5 on four, and 4.1 (occasionally 4.2) on all eight... Even on eight cores sustained, it has a 20% advantage over EPYC (assuming a poorly threaded application so EPYC can't go beyond eight cores).
The big Xeon-Ws, with the right cooling, can come pretty darned close to max turbo on all cores - I suspect the 3275 will be somewhere over 4GHz on all cores in the Mac Pro all day long. I'm not saying that the Threadripper 3970x won't do the same thing (and cheaper) - with 500 watts of cooling, it probably will. What won't is an EPYC with a max boost of 3.4 GHz
That's a big difference in a well-cooled application where boost matters more than base. The Mac Pro had better be a able to run its processor right up to boost clock most of the time - what else is all that fancy cooling that supposedly dissipates 500 watts for?!
There are situations where a laptop could be faster than that EPYC - if the application stays under 8 cores, the 9980HK's ability to boost to 5 GHz (briefly and only on a core or two), and hang out around 4 GHz or more should beat the EPYC. Of course, the EPYC will win in a well-threaded application, and by a lot in a very well-threaded application when the power of 32 cores comes into play...
My 9980HK (according to Intel Power Gadget) stays around 4.7 GHz going as high as 4.8 on one core, is around 4.6 sustained on two, 4.5 on four, and 4.1 (occasionally 4.2) on all eight... Even on eight cores sustained, it has a 20% advantage over EPYC (assuming a poorly threaded application so EPYC can't go beyond eight cores).
The big Xeon-Ws, with the right cooling, can come pretty darned close to max turbo on all cores - I suspect the 3275 will be somewhere over 4GHz on all cores in the Mac Pro all day long. I'm not saying that the Threadripper 3970x won't do the same thing (and cheaper) - with 500 watts of cooling, it probably will. What won't is an EPYC with a max boost of 3.4 GHz
Bingo. Finally - some sense to this discussion re: the mac environment.
Do you have test results and benchmarks between EPYC and Xeon-W?
Oh btw, Intel CPUs have a terrible thermal paste applied in CPU while AMD CPUs are soldered.
You don't need test results and benchmark when it is hard limit on both EPYC and Xeon-W.
https://ark.intel.com/content/www/u...n-w-3275m-processor-38-5m-cache-2-50-ghz.html
https://www.amd.com/en/products/cpu/amd-epyc-7742
That shouldn't concern you unless you are enthusiast and intend to overclock your CPUs. EPYCs and Workstation Xeons are guaranteed to run at those specifications at rated power.
https://www.overclock3d.net/news/cp...intel_s_3000_28-core_xeon_w-3175x_processor/1
It's not about overclocking. Intel was criticized for applying thermal pastes on their CPU for a while. It's a huge issue and this is because YOU HAVE TO REPLACE IT after a few years because the thermal paste does not last long. As a result, the temperature can not spread properly if you don't replace it. Do people have to risk it? Wow... So, I have no idea what you are suggesting.
Meanwhile, almost all AMD CPUs are soldered from the beginning that you need to worry about it.
In the real-world, the spec sheet does not represent real results with actual softwares. What you are showing is completely pointless.
That Xeon W-3175x is enthusiast CPU with "Xeon" name on it. It is compatible with Enthusiast motherboard with intention for overclocking. They are meant to be easily delidded for extreme overclocking purposes. It is not even comparable with Xeons that are going to be used on Mac Pros.
The issues that you are stating happens with i7s and i9 mainstream CPUs, which are often pushed beyond rated specs via overclocking. There aren't any issues with "thermal paste" with Xeons period.
Your argument was about boost clocks on EPYCs and Xeons, not the benchmarks on softwares. EPYCs and Xeons will run at those rated speed.
Last edited:
Thermal paste does not last forever and you need to replace it someday no matter what types of thermal paste you used. W-3175x is not the only example and even soldered Intel CPU isnt great with controlling high temperature. Like I said, it's not about overclocking and yet you didn't read it properly. Xeon SP series also have a thermal paste inside. Why is that?
Not true. Because of thermal throttling at all cores, it's impossible to sustain at the maximum clock speed.
If you can show me any issues related to that on Xeon based workstations (not i7s, i9s, HEDT X i9s, or W-3175x), I am more than glad to look. I haven't heard anyone talking about needing to replace "thermal paste" on Xeon based workstation (be it Apple, HP, Dell, Lenovo, or any other vendor). If not, you are just banging on non-existent issue just to bolster your argument.
just let you know, unless you are constantly overclocking. Any half decent thermal paste will outlive the lifetime of your CPU. Since You are not going to overclock workstation, it will not going to be a problem for the lifetime of the machine.
Yes, it will run at those rated speed, how many times do I have to repeat that. The boost clock is only for single core turbo. Neither EPYC or Xeon will operate all-core turbo on boost speed. Workstations Xeons (or Threadrippers) are more ideal for workstations because those have much higher boosts on fewer cores utilized on applications not utilizing all cores.
Last edited:
You know no model from Apple currently available can sustain the theoretical boost clock defined by Intel due to thermal limit even iMac Pro. I’m interested to see if Mac Pro can sustain it, but given that HEDT processor require decent cooling even in custom PC to sustain full boost nowadays, well we will soon have the performance report on Mac Pro 7,1.
Let’s hope that with that price tag and marketing centric design, the processors in MP 7,1 can fly without reaching 100 degree C like other Macs.
Let’s hope that with that price tag and marketing centric design, the processors in MP 7,1 can fly without reaching 100 degree C like other Macs.
I wish I had the cubic bucks required to build a decent Mac Pro rig and the equivalent Threadripper rig to actually test myself.
Maybe Snazzy Labs has the cash to get a 24-core Xeon and a 24-core Threadripper equally decked out and can put them through their paces...
Looking forward to the real world tests and where the rubber really meets the road.