Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The other bit about the imac pro is that it has dual TB3 lanes for those that might need the extra bandwith in that area.
 
The other bit about the imac pro is that it has dual TB3 lanes for those that might need the extra bandwith in that area.

yes. Also quad-channel memory vs dual-channel bus.

We'll have to wait for results, but I believe that in many real world tasks, the entry level 2017 iMac Pro will beat the high-end iMac 2019.

Despite Geekbench scores.
 
Any thoughts regarding the performance of a base 2017 iMac Pro vs. top spec 2019 iMac, assuming similar RAM, drive, etc. The primary use is heavy computations, mostly in Excel. The price difference is relatively narrow and it will be employer purchased.

Hi there,

Unfortunately, the better purchase will depend entirely on your excel spreadsheet. There are a number of threaded operations in Excel, most notably "sort" and some GUI elements.

https://docs.microsoft.com/en-us/office/client-developer/excel/multithreaded-recalculation-in-excel

You will have to determine if your document meets the threading criteria to know which platform will give you better performance.

If this is going to be a workstation with mission critical, financial, medical, scientific, or other consequential processing, then go for the iMac PRO because it features ECC DRAM and an encryption ASIC. Having uncompromising agency over your data is far more important than performance.
 
  • Like
Reactions: mikehalloran
There is no 2019 iMac Pro. They offered two BTO options: 256GB RAM and Vega 64X (whatever that means—Radeon VII?). Everything else under the hood and on the outside is the same.
No way. The X letter in Radeon means factory overclocking. So this is still the same 64 with a minimally noticeable increase. (disappointment)

Besides, the Vega 48 is also based on the already old Vega 10 process. (another disappointment). Its computing power goes from 580 to only 10-15%, according to forecasts. The only real advantage is HBM2 memory
 
Last edited:
No way. The X letter in Radeon means factory overclocking. So this is still the same 64 with a minimally noticeable increase. (disappointment)

Besides, the Vega 48 is also based on the already old Vega 10 process. (another disappointment). Its computing power goes from 580 to only 10-15%, according to forecasts. The only real advantage is HBM2 memory

Shaders alone dictate a 25% increase not mentioning architectural differences and more bandwidth between Polaris (580) and Vega 48.
 
So the only 6+ Core including HT cores is the i9 option right? Because all the other CPU's have no HT? Or am I totally wrong?

The Core i7-8700 BTO offered in the 21.5” is 6c/12t. The 9th Gen Core i7 CPUs i7-9700K, is an 8c/8t, no hyper threading. Only the 9th Gen Core i9 is 8c/16t. The benefit of no HT is the mitigation of side channel attacks, so take it with a grain of salt. I suspect HT will continue to give way to core count as Comet Lake S-Series (desktop CPUs) are reported to have 10-cores heading our way.

Core i5s have not included hyperthreading since they went from 2-cores to 4-cores.
 
Not trying to be a wise guy, but do you have a link? I have been looking and the usual suspects have said dink about basin Falls Refresh Xeons...still wondering about that rumored 22-core version. My gut tells me
that Apple will ditch the 8-core as the base version and make the 10-core the $4999 base CPU model.

Unless Intel radically changes the prices of the Xeon W line up, then they probably will keep the 8 core if also trying to keep the $4999 price point. The other option is for Apple to crank up the base base (if the Intel price adjustment is smallish ) to something like 5,499.

I think Intel will both cut the Xeon W prices and "unkneecap" the L3 cache for the 8 core model and put a clock bump on it. Between the i9 going 8 core and the Ryzen 3000 going 8+ (and thread ripper just even higher core count), the W can't support the same price premium it did back in 2017.

22 is unlikely for Cascade Lake iteration of W that Apple would likely use (i.e, with the 2066 socket with the nominal Xeon W line up). Leaked information about the SP versions show they didn't get a count count bump (max count still is 28) .

22-core may drop for next iteration of "extended" Xeon W. I wouldn't count on Apple touching it if it did. ( it is a socket and chipset change ... and Apple is extremely unlikely just to build a single board with a different socket just for a BTO option. ). Intel shoved that W 3175X into the line up at 28 ( with a 3647 socket / C621 chipset ). There is a decent chance they may also dribble out a 22 ( and maybe 20 ) versions. At the moment, motherboards made for the 3175x can only take one CPU package model. .... which is way off the 'norm'. Even if that was two ( a 28 and 22 model), Apple probably walk away from that as an option.
 
That 255W TDP for the 3175X (compared to the 145W for the 2195) would certainly impact how quiet the iMac Pro could run (I think it could handle it with the fans cranked).
 
I cannot find the specific link (it was the the "Waiting for Mac Pro 7.1" thread), but based on a search, Cascade Lake-X is supposed to be announced at Computex with Glacier Falls (the Xeons) due by end of year: https://hothardware.com/news/intel-cascade-lake-x-hedt-cpus-glacier-falls-x399-late-2019-launch

That article is dated back in Nov 2018.

Article from same site but in the current year (2019 )

https://hothardware.com/news/intel-cascade-lake-cpus-rumored-april-cascade-lake-x-computex


But that is only Cascade Lake X; technically not Xeon W. Intel's logjam on 14nm production is clearing up, but it is unlikely they will try to launch both -X and W ( which is largely same die with different features turned on/off) at the same time at high volume. The details for W will probably get much clearer, but I suspect those will come later (as the HEDT crowd is much more 'manic' about getting things before others do. )

That "late 2019" timeline back in 2018 was probably a combination of
i. fuzzy ness on how quickly could get 10nm started and 14nm logjam resolved. ( maybe more single tracking through fab)

ii. Hope that AMD would stumble greatly with getting their 7nm processors out the door ( can coast longer because competition )

iii. Comet Lake not clear path to production.

iv. Muddled idea of if doing a Cooper Lake -X and/or Xeon W Cooper Lake) or not ( goes back to how long until 10nm+ gets online). [ target socket change from 2066 in early 2020 or late 2020 . If late 2020 probably would going to try to kick the can on Cascade longer. ]


Intel need to do something sooner rather than later though. The top iMac BTO covering the low end iMac Pro isn't just an Apple problem. Lots of other workstation vendors have the same issue with the "midrange' workstation powered by Xeon w.
[doublepost=1553377981][/doublepost]
That 255W TDP for the 3175X (compared to the 145W for the 2195) would certainly impact how quiet the iMac Pro could run (I think it could handle it with the fans cranked).

It isn't just "quiet". The total cap for power in the iMac Pro is a bit under 500W. Another 80W would be eating into some other component's power budget. Apple just probably threw a higher TDP Vega64x in the enclosure too. And some "magical" Navi future 'big' GPU is highly unlikely going to claw back 80W either.

different socket
Different chipset
outside the power budget

(and also probably can't use the extra memory channels. since DIMM space a somewhat of a premium. )

It doesn't match at all to the iMac Pro. ( even if went to "bigger" with a wider screen ... it still doesn't probably fit in the line up).
 
Last edited:
  • Like
Reactions: mikehalloran
Unless Intel radically changes the prices of the Xeon W line up, then they probably will keep the 8 core if also trying to keep the $4999 price point. The other option is for Apple to crank up the base base (if the Intel price adjustment is smallish ) to something like 5,499.

I think Intel will both cut the Xeon W prices and "unkneecap" the L3 cache for the 8 core model and put a clock bump on it. Between the i9 going 8 core and the Ryzen 3000 going 8+ (and thread ripper just even higher core count), the W can't support the same price premium it did back in 2017.

22 is unlikely for Cascade Lake iteration of W that Apple would likely use (i.e, with the 2066 socket with the nominal Xeon W line up). Leaked information about the SP versions show they didn't get a count count bump (max count still is 28) .

22-core may drop for next iteration of "extended" Xeon W. I wouldn't count on Apple touching it if it did. ( it is a socket and chipset change ... and Apple is extremely unlikely just to build a single board with a different socket just for a BTO option. ). Intel shoved that W 3175X into the line up at 28 ( with a 3647 socket / C621 chipset ). There is a decent chance they may also dribble out a 22 ( and maybe 20 ) versions. At the moment, motherboards made for the 3175x can only take one CPU package model. .... which is way off the 'norm'. Even if that was two ( a 28 and 22 model), Apple probably walk away from that as an option.

Yeah, they will only touch a 22-Core if it’s LGA-2066, and it fits the existing iMac thermals. After the Core X-Series refresh in October failed to include a higher core count CPU, I resigned myself to the likelihood of 18-Core being the max in a refresh to the Xeon W.

I don’t see the 3175X going in the Mac Pro either. Really wish Apple would get us a preview, tired of all talk, no action.
 
You can either have a hair dryer in your face, or peace and quiet, when working hard.
I tend to prefer the latter and so do the guys who work near me.
 
  • Like
Reactions: mikehalloran
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.