Any testing and results with HEVC and Handbrake is of interest to me. I opted for the i7, but still wonder about the i9. In macOS the most intensive thing I do most often are Handbrake encodes. Windows is for games, but if the performance difference is stark I might consider using it for Handbrake. There may be trade-off missing out on CoreAudio though.
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.
Decision time i7 vs i9 for new 2020 iMac?
- Thread starter BasilFawlty
- Start date
- Sort by reaction score
If the most intensive thing you do is encode to H265 then stick with the i7 and make sure you upgraded to the 5700xt, 265 encoding is much faster using the 5500XT in windows handbrake compared to the i7 CPU by a factor of around 2-3 times, while in macOS the 5500XT was around twice as fast as CPU only for handbrake 265 export. However I don't know how much of that was to do with the t2 chip which I know doesn't work in bootcamp for 265 encoding which may explain why the AMD card was faster under bootcamp.
If early reports of the i9 only giving around 10-15% extra performance over the i7 then neither CPU will be as fast as using the AMD graphics card for 265 handbrake exports.
Not to derail the thread, but I've always used CPU encoding, my (limited) understanding is that GPU encoding sacrifices quality for speed. I'm willing to wait for a little for an advantage in bitrate or file size. I do have the 5700 XT, let me know if I'm missing anything.
I think each successive generation of graphics cards are closing in on the quality gap compared to CPU encoding. I haven't really done a frame by frame pixel by pixel comparison but when setting to a fixed bit rate I hardly noticed any difference between CPU and GPU encoded videos.
What I have noticed is that using an Nvidia 2070 super eGPU compared to the AMD 5500xt as well as the CPU, the Nvidia encode is least efficient creating files that are twice the size of the AMD and CPU encodes when exporting at any given RF setting. The AMD was surprisingly the most efficient making the smallest file sizes compared to the CPU without any noticeable difference in quality but then I wasn't exporting UHD stuff so results may vary depending on the project.
All this means is that the debate about i7 vs i9 can sometimes be overshadowed by other bits of hardware that do the heavy lifting even better then either CPU such as graphics cards and T2 chips.
Funny, I made an appointment to try a 2nd hand Quadrifoglio (with manual gear). Unfortunately it was sold before I could see and try the car. I now have a Giulia 2.0 Super with standard 200hp. I had the option to upgrade/tune to Veloce level (280hp) or max (320hp); did the last one. The engines of the 200hp and 280hp are 100% identical, the difference is just a couple of settings in a computer. Tuning has been done by Squadra Tuning (see Squadra-tuning.com). He (Stephan, owner) has also added the "race" setting (security systems off!). The Super weights 100kg less than the Veloce (2 versus 4 wheel drive). I currently also own a Spider 916 (3.0 V6). Totally incomparable cars, but both great. I'm driving Alfa since 1988 (2x33, 2x155, 166V6, Spider, Giulia). My observation is that Alfa is more reliable than most German cars.Which model Giulia do you have? I was just looking at some Stelvios with the same 280bhp engine and wondered how much difference would a remap make. An extra 40bhp should be quite noticeable. What's amazing is that the humble 200bhp engine can also be tuned to 320 hp! Maybe that's the best solution for low cost and insurance purposes
I really would love to get a Guilia Quadrifoglio and 3 years ago I very nearly did but common sense kicked in and I gave it a pass as the car was too much of an unknown quantity regarding reliability and residuals at the time. I test drove both a 280bhp Veloce and the Quad and it was no comparison. The Veloce was nice and handled well but lacked all the drama and fun of trying to tame that wild Quadrifoglio and it's 500+ horses, the car kicked so much at every turn it was like riding a bucking bronco and the noise she made was incredible. It was wet and windy when I had the test drive but I spent most of it with the window down so I could hear the engine and exhaust roar.
Anyway back to the iMac, I've been using an i7 iMac 2020 for almost a week now and performance is good, much better then using a 16 inch MBP i9 which thermal throttled like crazy. Running FCP and and using neatvideo plugin the MBP 16 inch would thermal throttle down to 1.5ghz from a base of 2.4ghz while the i7 iMac keeps things running at over 4ghz. Only downside is that the AMD graphics card in the iMac doesn't seem to be supported by neat video so it's only using the CPU. Hopefully this will change as they add support.
Geekbench 5 on the 16 inch i9 MBP with 32gb ram out of the box scored 7000 multicore while the iMac i7 with only 8gb ram scored 7500. I added an 8gb stick of ram to it and its score went up to 7880, would probably be higher if I added matching pairs of 8gb sticks rather then 3 sticks of 2x4gb and 1 8gb, in which case I would expect scores to hit above 8000. But geekbench is a synthetic benchmark that lasts about 1 minute so doesn't take into account thermals in which case I'm sure the difference between the 16 inch MBP i9 8 core and the iMac i7 8 core would be much greater.
As for i7 vs i9 iMac, I suspect the thermals of the iMac would severely limit that i9's ability to stretch it's extra legs out over the i7 in a pure speed test. The extra 2 cores would come in handy if you're running lots of virtual machines, running multiple intensive apps e.g rendering in FCPX while exporting another project in Compressor, or using something that multi-thread such as Handbrake.
Speaking of Handbrake, using the Cpu to compress video is marginally faster in the windows version under bootcamp compared to using it on mac. The 5500xt also seems to export a little bit faster in windows under AMD VCE settings compared to when the AMD card is used by videotoolbox in MacOs. Under bootcamp, using handbrake to export h265, the AMD 5500xt was faster then an Nvidia 2070 super thunderbolt 3 EGPU. I would expect the 2070s would outperform the 5500xt had it not been constricted by the thunderbolt 3 interface. Did a load of test and benchmarks which I will writeup in a thread I made before about the iMacs and eGPUs.
Back to the iMac. I'm still waiting....
With regards to the memory: if you mix sizes or brands, you won't get the optimal speed. You can check this with Geekbench version 4. If you look in the Geekbench 4 database (search for iMac20,) you will find many samples of sub-optimal memory configurations. Best is to insert, depending on your needs, a new set of 2x16 or 2x32, and sell or store the original memory). My new bought 2x32Gb is waiting for an iMac 2020 to arrive...
I use my current iMac also to run VMs in Parallels (v15). That works amazingly well. I also have an HP Microserver Gen8 running VMware. But in this forum I've read that some people want the I9 because it has more cores; assuming that's better for running VMs.
What I noticed with Parallels is that if you configure e.g. 2 cores (out of the 4 cores on my iMac) for Windows 10, with full load in Windows, all 4 real cores will be about 50% busy. So there is no relation between the number of physical cores and and the number of logical cores (which is actually very good!).
So Parallels could make the setting more variabel, allowing me e.g. to set 3.14 cores for my Windows VM
Conclusion: the real number of cores in your iMac isn't of much importance for Parallels.
Last edited:
10% faster for $400 doesn't seem unreasonable to me. As for the other important parts, upgrade them too.
Maybe price/performance is better for the 8 core but its much worse for the iMac Pro 10 core.
I have had no problem repeating my build time. Your machine is probably less thermally constrained than an iMac but perhaps the SSD is not as fast.
I tested further and my times seem to vary between 178 to 188 seconds.
Builds are typically not constrained by IO at SSD speeds (I see less than 100MB write/sec at all times during this benchmark), and the NVME drive I'm using is actually rated pretty highly anyway. I'm not convinced that's the distinguishing factor.
I'm also not convinced we're not running into some limits with Xcode or how this benchmark is constructed.
Anyway, sort of interesting, but probably not going to mess with it further.
MaxTech ran the benchmark on a Machine almost identical to mine except for the SSD, it recorded much slower times. Your times are broadly similar to mine which suggests that the differences between the T2 controlled SSD and yours are not significant. The build times are too short for thermals to be a factor.
What an awesome thread haha. Just read the entire thing. Still not sure on what to buy. Even though the price/quality ratio of the i7 clearly wins, I have not seen clear evidence that the 10% performance boost I see in the multicore Geekbench (and 12% in Cinebench) is very far from the real life application in my case.
I am a music producer using Ableton. Large, CPU-intensive sessions. Max Tech did do a Logic Pro benchmark, but that benchmark does not reflect real life performance at all if your ask me. The problem is never: "oh too bad I can't add the 157th track in my session before I get an error." The bottleneck is all the different (third party) plugins you use on those tracks for a long sustained period, and as a result get a CPU overload with audio latencies, increased buffer sizes, audio glitches and a general slow and overheat machine. And that drives me absolutely nuts lol. If I can get even close to 10% extra stuff I can throw at it before that happens, it's worth the price difference for me. It's clear what Ableton says on their website about it (here):
"Which is more beneficial, a faster CPU speed or more cores?
Both are important. If your budget allows it we recommend getting the fastest processor and maximum amount of cores that you can afford. Here's a breakdown of the pros and cons of both:
More cores, slower clock speed
Based on the above, I am leaning towards the i9. It feels like a necessary guess to some point, and I'm the first to admit that I know jack **** about how exactly processors work, but I have to go with it because a representative music production benchmark just doesn't exist. And then I would rather be sure to have the best thing there is, of the newest generation processors. If anybody thinks otherwise or has reasons to believe a different benchmark would be more representative for music production, please let me know. All thoughts are welcome!
I am a music producer using Ableton. Large, CPU-intensive sessions. Max Tech did do a Logic Pro benchmark, but that benchmark does not reflect real life performance at all if your ask me. The problem is never: "oh too bad I can't add the 157th track in my session before I get an error." The bottleneck is all the different (third party) plugins you use on those tracks for a long sustained period, and as a result get a CPU overload with audio latencies, increased buffer sizes, audio glitches and a general slow and overheat machine. And that drives me absolutely nuts lol. If I can get even close to 10% extra stuff I can throw at it before that happens, it's worth the price difference for me. It's clear what Ableton says on their website about it (here):
"Which is more beneficial, a faster CPU speed or more cores?
Both are important. If your budget allows it we recommend getting the fastest processor and maximum amount of cores that you can afford. Here's a breakdown of the pros and cons of both:
More cores, slower clock speed
- Pros
- Live supports multi-threading, therefore the more cores are available, the more efficient it will be when working with larger sets with higher track counts, or when working with large instrument or effect racks.
- You'll probably be able to run more apps in conjunction with Live without seeing performance drops.
- Cons
- Lower single-threaded performance than a higher clock speed processor."
Based on the above, I am leaning towards the i9. It feels like a necessary guess to some point, and I'm the first to admit that I know jack **** about how exactly processors work, but I have to go with it because a representative music production benchmark just doesn't exist. And then I would rather be sure to have the best thing there is, of the newest generation processors. If anybody thinks otherwise or has reasons to believe a different benchmark would be more representative for music production, please let me know. All thoughts are welcome!
Last edited:
I also run Ableton and have compared my i7 against the i9 performance benchmark from the Ableton Live forum. They run pretty much identically. A Logic benchmark test for a session of bussed channels over at Gearslutz revealed they run identical amounts of tracks and plugins but the i7 ran about 10° cooler - this is the best real world test I've seen comparing the processors.
As you know, audio doesn't utilise cores like video rendering does, due to how an audio buffer processes in real time, less and higher frequency cores are sometimes more beneficial than more lower frequency cores. This is obviously a fine balance depending on one's specific workflow.
Last edited:
Could be hitting a memory bandwidth issue (which is much more hidden/hard to measure as opposed to just pure core performance). Doesn't really make any sense that an i9 wouldn't scale pretty well in this respect.
Thanks for this! Could you direct me to those benchmarks and the exact scores if possible? That Ableton benchmark probably uses only native plugins?
Last edited:
Yeah sure, here's the post with the results of the Logic benchmark test
The results:
i9: 21 Plugins, average clock speed: 4.12 ghz, max: 87°
i7: 21 Plugins, average clock speed: 4.40 ghz, max: 78°
Here's the Ableton test for the i9. I ran the same test with my i7 and it performed marginally better for some track counts and slightly worse for others, making it pretty identical. However, this isn't as good a real world test, as it doesn't contain sub groups like the logic test (assuming you use grouped tracks in your processing) It still goes a long way to showing the CPU performance in Ableton is very similar for each processor.
I think you'll be great with either processor, depending on how you work, these iMac's are very powerful and more than enough for most producers. I opted for the i7 because I saw no decerinble performance gain in the i9. The i7 runs at a higher sustained frequency which benefits audio and looks to run cooler, as you can see above.
These benchmarks aren't highly dependent on memory, so I don't think that's the reason. DAWs don't spread a serial signal chain (bussed tracks) across cores effectively, so the higher sustained clock frequency of the i7 really helps.
Thank you, that's interesting! Like the poster says, the test method itself does seems geared towards trying to benefit more from the higher clock speed of the i7 then a comparison of a realistic session by putting an unrealistic number of plugins on the master buss (so for one thread). The i9 could outperform the i7 in a project with the plugins spread more over the session, who knows. But still, this is a meaningful result. Also because you will need a lot of processing power for one track sometimes.
It would be so nice if there would be a benchmark with a truly realistic session (40 tracks, 5 plugins per track, let's say 5 sub groups, some FX send/returns, and one mix buss), with a nice mix of plugins from some of the biggest companies (Waves, Plugin Alliance, Soundtoys, hell maybe some CPU-monsters like Diva or anything Acustica Audio) running, just to see the CPU load on that. Most of the third party plugins support multithreading, but each in a slightly different way I suppose so those results would be very interesting.
Yeah this test does not look very precise to me. All it says is that somewhere between 256 and 512 tracks the sound gets funny. Could be at first at 257, could be 512. Tough to compare that and draw conclusions. And these demo files are no representation of real life, almost nobody just uses stock plugins. What was your result?
Thanks to pointing out that Gearslutz thread. Going to read the whole thing now lol.
I've had my 10-core for almost a month, and I'm still debating whether I made the right choice. But I just realized Apple's return policy is 14 days, not 30. So I guess I'm "stuck." Oh well.
But to follow up on my RAM post from before, I tested every possible configuration with Apple's two 4 GB chips and my two 32 GB Timetec chips, and then I ordered four 16 GB Timetec chips. Indeed, mixing chips dropped the RAM performance, so 32-0-32-0 performed better. But filling all the slots with the same amount of RAM (16-16-16-16) performed the best. Plus, that actually saved me about $30 (before Timetec dropped the price of the 32 GB chips).
But for my workflow (a couple VMs, multimedia editing, some gaming under Windows, etc.), I think I should have actually gotten the 8-core i7.
But to follow up on my RAM post from before, I tested every possible configuration with Apple's two 4 GB chips and my two 32 GB Timetec chips, and then I ordered four 16 GB Timetec chips. Indeed, mixing chips dropped the RAM performance, so 32-0-32-0 performed better. But filling all the slots with the same amount of RAM (16-16-16-16) performed the best. Plus, that actually saved me about $30 (before Timetec dropped the price of the 32 GB chips).
But for my workflow (a couple VMs, multimedia editing, some gaming under Windows, etc.), I think I should have actually gotten the 8-core i7.
Well besides all the rational arguments, being able to say you have a 10-core CPU in a bar is still a hell of a pick up line so completely worth it.
How do you guys think the i7 vs the i9 would compare when you count the sustained CPU load as a factor? I know to little about how CPU's work to guess anything on that. Max Tech said in his video that the i9 could have an edge with sustained loads (I believe because the workload is spread more), but on the other hand the i7 seems to run cooler (like Jobinhosyntax pointed out in his post above).
Any (rational) ideas?
How do you guys think the i7 vs the i9 would compare when you count the sustained CPU load as a factor? I know to little about how CPU's work to guess anything on that. Max Tech said in his video that the i9 could have an edge with sustained loads (I believe because the workload is spread more), but on the other hand the i7 seems to run cooler (like Jobinhosyntax pointed out in his post above).
Any (rational) ideas?
If there are no thermal constraints the i9 10900K will outperform the i7 10700K in single core too because of the higher boost clocks. 5.3GHz vs 5.2GHz, but inside the iMac, the boost clocks for both are capped at 5.0GHz. With the thermal system and constraints the i7 10700K will likely beat the i9 10910 in single-core.
You see this play out with the 10600K which has a base clock of 4.1GHz over the i7 or i9 in more constrained systems where the boost clocks cannot be sustained.
You see this play out with the 10600K which has a base clock of 4.1GHz over the i7 or i9 in more constrained systems where the boost clocks cannot be sustained.
No thermal constraints when running 1 core 100%. According to datasheets at ark.intel.com the max. clock speed for the i7 is 5.1 GHz en 5.0 for the i9. Apple hasn’t capped the clock speed; they do use tailored versions of the CPUs.
8thNote - would you mind expanding on this a little please? How much better was 4 x 16 compared to 2 x 32 and how did you test them/tell the difference?
About to order 64GB and wondering which to get. I'd get 2 x 32GB if no real difference as that would allow possibility for upgrade later but if 4 x 16 is noticeably better I might do that - also wondering if 4 x 16GB might run a little cooler.
thanks
If you chose 2x32 GB, take Crucial certified for Mac RAM. It's the only RAM that guarantees you long term upgradability when adding another kit. It will always use Micron DRAM chips and be 100% compatible, unlike OWC, Kingston and others that mixes Hynix and Micron chips with time.
There is virtually no difference between 4x16 or 2x32, really.
Thanks - I'm going for the Crucial but not the Mac certified one - this one (CT2K32G4SFD8266)
Reason being it is 2/3 the price and with 64GB its unlikely I'll need to upgrade anyway (which is why I'd happily do 4x16 if there was a noticable improvement).
According the the iMac 2020 RAM compatibility thread, the only difference between this one and the Mac certified one is use of lead and halogen which does not meet Apple's environmental policies and users have reported it works fine.
[Aside] - I bought my first Power PC Mac in about 1995 and I remember 8 MB of RAM cost me £200 (about £330 in today's money). I wondered at the time if/when I might ever be able to afford 64 MB to really make Photoshop fly... I'm now buying 64 GB of RAM for £213 - which makes it about 8000x cheaper!
Reason being it is 2/3 the price and with 64GB its unlikely I'll need to upgrade anyway (which is why I'd happily do 4x16 if there was a noticable improvement).
According the the iMac 2020 RAM compatibility thread, the only difference between this one and the Mac certified one is use of lead and halogen which does not meet Apple's environmental policies and users have reported it works fine.
[Aside] - I bought my first Power PC Mac in about 1995 and I remember 8 MB of RAM cost me £200 (about £330 in today's money). I wondered at the time if/when I might ever be able to afford 64 MB to really make Photoshop fly... I'm now buying 64 GB of RAM for £213 - which makes it about 8000x cheaper!
Last edited:
Yep you are good with this one too.Thanks - I'm going for the Crucial but not the Mac certified one - this one (CT2K32G4SFD8266)
Reason being it is 2/3 the price and with 64GB its unlikely I'll need to upgrade anyway (which is why I'd happily do 4x16 if there was a noticable improvement).
According the the iMac 2020 RAM compatibility thread, the only difference between this one and the Mac certified one is use of lead and halogen which does not meet Apple's environmental policies and users have reported it works fine.
[Aside] - I bought my first Power PC Mac in about 1995 and I remember 8 MB of RAM cost me £200 (about £330 in today's money). I wondered at the time if/when I might ever be able to afford 64 MB to really make Photoshop fly... I'm now buying 64 GB of RAM for £213 - which makes it about 8000x cheaper!