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.
2020 iMac Benchmarks Surface Online [Updated]
- Thread starter MacRumors
- Start date
- Sort by reaction score
I have a i5 27-inch Retina Late 2015. OK, the 2020 version is double the speed in multicore because it has double the cores but that's about it, in 5 years! Biggest speed improve is the SSD but i will need slower external HDD's to make up for the loss in capacity. Not sure what I'm going to do.
iMac (27-inch Retina Late 2015) Benchmarks - Geekbench
browser.geekbench.com
Because storage is prone to failure no to mention corruption and viruses like ransomeware, if it fails you can just replace it for couple $100 and life goes on. If its soldered, you might as well buy a new machine. RAM, Screens, USB ports...those hardly fail, very rare case but with storage they actually tell you to make backups in the 3-2-1 method.
I'm guessing the multi-core boost is largely due to the Comet Lake i5 having HyperThreading, and the Coffee Lake i5 not having it. So some tasks can perform in 12 simultaneous threads despite just having 6 cores.
Other than that, these results aren't that great (unsurprisingly). Comet Lake is hopefully Intel's last "sorry, hopefully we'll have a new microarchitecture soon" releases, at least as far as Skylake / 14nm goes.
In practice, though,
(edit)
OK, so someone responded with a failing SSD after three years.
Er. Wha? At worst, you need to wipe the drive in that case. No hardware changes needed.
Last edited:
New update says that the SSD is in fact not removable, but the 4 and 8 TB models have an expansion slot. So I guess you were the early bird who got fooled.This
It is soldered unless you get the 4 or 8 tb storage, but nice quick apple defense. Well done.
Pay attention when they post scores, they play games with Geekbench 4 vs. 5 to pump the numbers up and skew opinion on folks that just take a glance. 3-5% speed increase core to core is what Intel is capable of doing right now. Last iMac, Intel's last gasp at 14nm.
Yes, correct. It will die on you after thousand upon thousands of writes. Expect death in 15+ years, oh wait, you won't have the computer in 15 years... Context is key.
The SSD part of my 2013 iMac Fusion drive failed. Two weeks before the 3 year warranty limit, so got a free replacement from Apple. When they get fails on the new machines they have to dump the whole main board...
I would be curious if that was due to the machine keep on transferring data between the SSD and HD. Which I know small SSD’s are less tolerant to a lot of writes. My 2013 iMac with the 512 GB SSD was good for the 5 years I used it before upgrading to the 2019.
If not that, probably the controller died. Also understand that’s an SSD’s biggest weak point is the controller than the storage medium themselves. And if the controller goes, the data is lost as well to my understanding. Having the T2 being the controller may improve reliability in that regard since it is based off an A series SoC.
I don't use the internal soldered SSD of my mac mini and my MBP, I use a Samsung X5 external Thunderbolt3 drive that I boot from .
At work we have iMacs , and we all have an X5 So we can sit wherever we want and boot from it. The MBP's original internal SSD is configured but I use it mostly as backup
Another good thing about it is that I can buy a new Mac and start working from it right away. The speed of the X5 on my 2015 MBP is around 2000 MB/s read and 1900 write, so you won't lose performance.
We do this for years, in the past (2012-2015) we used USB-3 SSD's (Samsung T5) , and in situations where those became too slow we copied our data/work folders to the faster internal drive. But now with the X5 and TB3 there is no need to use the internal SSD at all
Here's the list of GB results for the real trainspotters:
Core result: https://browser.geekbench.com/v5/cpu/search?utf8=✓&q=iMac20,1
Compute: https://browser.geekbench.com/v5/compute/search?utf8=✓&q=imac20,1
Core result: https://browser.geekbench.com/v5/cpu/search?utf8=✓&q=iMac20,1
Compute: https://browser.geekbench.com/v5/compute/search?utf8=✓&q=imac20,1
Apologies for delayed reply - been away.
Fairly heavy use of the machine - ~3 working days a week as a developer, so lots of small files changing often, also a big photo library. It's possible the fusion drive algorithm overused the SSD, or I just got unlucky. But by the time it broke I ended up splitting the fusion drive to confirm it was just the SSD part broken - the HDD bit was fine.
That repaired machine then got 1 more year of life as a primary machine, before the next 3 years being a secondary to a 2017 iMac. It still works fine and was finally switched off yesterday as the 2017 machine became the secondary, as the new 2020 iMac has become primary.
Ran GeekBench 5 on the new machine (3.8GHz 8-Core i7, AMD Radeon Pro 5700 8 GB, 72GB RAM, 4TB SSD) and got the following results:
CPU: 1177 / 7927 - https://browser.geekbench.com/v5/cpu/3327012
GPU:
OpenCL: 44174 - https://browser.geekbench.com/v5/compute/1356337
Metal: 49980 - https://browser.geekbench.com/v5/compute/1356339
81% multi-core boost when doubling the core count isn't bad at all.
But the single-core result is fairly weak, yes. Keep in mind the CPU is just a(nother) mild revision of the 2015 Skylake. The 2017 is Kaby Lake, and this one is Comet Lake; neither changes the microarchitecture nor the process node. This winter's Rocket Lake (new microarchitecture*) or next year's (or 2022's?) Alder Lake (new process node, too) will probably make much more of a difference. Will either of those ever end up in an iMac? Maybe not.
*) With other goodies, such as PCIe 4.
It is rather funny that a $399 iPhone SE is a quarter faster at single-core tasks.
I suppose the 2017 machine had a higher base clock speed (4.2GHz) so an 81% increase on 2x cores is reasonable. Now to find some software that can make use of them...
For once I did a fresh install rather than a Time Machine migration - got rid of about 500GB of cruft. But I'm still getting annoying hangs with things like 'Contacts' as it locks up while trying to do a sync with cloud.
XCode is noticeably faster once it gets going on a reasonable sized project though, but that is SSD vs HDD rather than anything else.
I'm looking forward to Apple silicon machines for performance, but needed a new one *now* and am not looking forward to soldered RAM at Apple prices...
But my projects aren't big enough! A ~1500 file iPad project of ~200k lines of Objective-C code only takes 23 seconds to build from clean now...
Scarily a 60 file, 15k line Swift project also takes 23 seconds (23.985 vs 23.353 for the other one) as well.
Is the RAM a mix of third party and the RAM it came with from Apple? What slots were the RAM installed in? Is it running at full speed?
There seems to be an issue when mixing the RAM that came with from Apple and third party RAM which can cause performance issues and may explain the lower scores. The 10700K should be scoring in the mid to high 8000's for multi core and high 1200's to low 1300's for single core.