i5 9600k Benchmarks.

[MasConejos]

That's what I ordered. I too find it hard to financially justify the i9 + Vega + SSD that so many people seem to be ordering. I ordered an i5/Vega/512GB SSD, with a student discount, and I still feel bad about how much I spent.

 

[purduealum91]

I purchased the $2,099 edu model - 9th gen core i5 w/ 8GB RAM, 2 TB Fusion, 8GB 580X and I added 32GB of Crucial RAM. I returned my Corsair Vengeance because it was not registering as 2666.

 

[Bohemien]

I re-ran Cinebench 10x in a row and this time including the fan speed. Max speed was 1850 rpm.
[doublepost=1554047292][/doublepost]View attachment 829651

Here's a screen capture

So what score did you get in Cinebench?

 

[propower]

Thanks much for these!

So - 4.2GHz 6-core sustained speed (at 92 or so degC). In my work (real time audio recording) raw speed is super important for low latency tracking. My 3.8 2017 i5 has sustained full load speed ~4GHz. This 2019 would be a 50% increase in plugin running ability. Looks like a great choice for many.

 

[purduealum91]

My Cinebench R20 score was 2595.

 

[kaintxu]

If I'm looking for longer durability and I had to chose between I9 or Vega 48, which one would mean the iMac would last slightly longer (1-2 years)? I guess the CPU?

 

[Zdigital2015]

You can always upgrade the GPU with an eGPU, though I think they are crazy expensive

The upside is that the Vega 48 gets to use all x16 PCIe 3.0 lanes directly off of the CPU versus an eGPU which will only get x4 PCIe 3.0 lanes hanging off of the PCH and ultimately going through the DMI 3.0 bus to communicate with the CPU.

Although the Vega 48 is not a cheap BTO item, once it’s done, it’s done and makes way more sense than hanging an eGPU off of the iMac and eating up a Thunderbolt 3 port, even if it’s down the road.

Besides, you will spend at least $450 to build a Vega 56 or above eGPU, much more if you go with a Radeon VII. Ditto for an NVIDIA GPU, if you are going that direction. Just my 2¢.
[doublepost=1554066693][/doublepost]

If I'm looking for longer durability and I had to chose between I9 or Vega 48, which one would mean the iMac would last slightly longer (1-2 years)? I guess the CPU?

Get the Vega 48 and upgrade the CPU later...the cost for the 9900K CPU should come down a bit over time and make your iMac last even longer once you swap it out with the 9600K.

 

[kaintxu]

Saw on posted by someone on Reddit for Cinebench on the 9600k. Weird that it thinks there are 3 cores and six threads. Should be 6 cores. FYI, the Core i9 scores 4069 with the same Vega CPU compared to 2487 on the Core i5. That's quite a difference!

I just ran Cinebench on my i5 9600K w/ 40GB RAM and the 8 GB Radeon 580X. I can confirm Cinebench thinks there are 3 cores and 6 threads. My geekbench single core score is 5773, multicore score is 23633, and Open CL score is 119723. I upgraded from a late 2013 21.5" iMac with core i7.

Is this something that could mean a problem down the line a benchmarking program thinking is 3 cores. Could it also be that other actual non-benchmarking software thinks it's a 3 core and works as such?

Cheers

 

[Colonel Blimp]

That looks pretty impressive, even compared to the i9 which tops at about 3.7 GHZ under full load.

But the i9 is an eight-core CPU. If you max out only six of the i9’s cores, it should top out at a higher frequency than when maxing out all eight cores.

If we look at the first chart on this page at AnandTech, we see that the i9 should run at least 4.7 GHz when restricted to 95W with only six cores maxed out. But we also know that the i9 in the 2019 iMacs is outperforming AnandTech’s prediction of 3.6 GHz with all eight cores maxed out. (The results I’ve seen posted here show the i9 topping out at 3.8 to 3.9 GHz when maxing out all eight cores.) So perhaps the i9 might run even faster than 4.7 GHz when maxing out only six cores?

What would probably be more informative than my armchair theorycrafting, however, would be actual benchmarks. 🙂

If I'm looking for longer durability and I had to chose between I9 or Vega 48, which one would mean the iMac would last slightly longer (1-2 years)? I guess the CPU?

I’d be inclined to think the GPU, not the CPU, partly for the reasons Zdigital2015 mentions above, and partly because a lot of software is being written (and rewritten) to take more and more advantage of GPU compute power, but mostly because you say that you do like a bit of gaming, and for that the GPU is going to give far the best advantage.

 

[Zdigital2015]

But the i9 is an eight-core CPU. If you max out only six of the i9’s cores, it should top out at a higher frequency than when maxing out all eight cores.

If we look at the first chart on this page at AnandTech, we see that the i9 should run at least 4.7 GHz when restricted to 95W with only six cores maxed out. But we also know that the i9 in the 2019 iMacs is outperforming AnandTech’s prediction of 3.6 GHz with all eight cores maxed out. (The results I’ve seen posted here show the i9 topping out at 3.8 to 3.9 GHz when maxing out all eight cores.) So perhaps the i9 might run even faster than 4.7 GHz when maxing out only six cores?

What would probably be more informative than my armchair theorycrafting, however, would be actual benchmarks. 🙂



I’d be inclined to think the GPU, not the CPU, partly for the reasons Zdigital2015 mentions above, and partly because a lot of software is being written (and rewritten) to take more and more advantage of GPU compute power, but mostly because you say that you do like a bit of gaming, and for that the GPU is going to give far the best advantage.

The article at Anandtech qualifies as a classic for me.

It certainly makes the case for Apple restricting the TDP of the 9900K to 95w at the sacrifice of some performance. Intel has played fast and loose with TDP for a while now, and with unrestricted TDP, it really shows in the 9900K...165w dissipation., that's insane!

There is no way Apple could have done anything but restrict it and use the existing chassis. I know everyone wants a redesign, but I don't think we will see it until October of this year, at the earliest.

The money quote from the article though is this - "The Intel Core i9-9900K, in normal operation, scores an efficiency rating of 32.9. This rises to 44.2 if the processor is fixed to 95W."

Apple knows this will buy them a bit more time using the existing chassis. If the redesign of the iMac is as radical as I think it may be (no more HDDs at any price point), they needed to put out just one more iteration for those who are going to be put off by the inevitable price increase coming our way. I suspect that this iMac or a incremental update will be on sale for a year or more after the "new" iMac is announced and released, whenever that may be.

I am curious as to whether Comet Lake (10-core S-Series) will also be able to have its TDP restricted to 95w. I assume that it will, but until it is announced, speculation is the name of the game.

 

[kaintxu]

I re-ran Cinebench 10x in a row and this time including the fan speed. Max speed was 1850 rpm.
[doublepost=1554047292][/doublepost]View attachment 829651

Here's a screen capture

Is it me or the i5 9600K seems capped at 80W, while it's meant to be a 95W CPU?

 

[Jimmdean]

Seems like one of those things we shouldn't be worrying (or complaining) about. If someone wants to have full control over their voltages and thermals they're just going to have to build their own machine.

 

[kaintxu]

ignore me - Those were CPU results.

 

[gusping]

I purchased the $2,099 edu model - 9th gen core i5 w/ 8GB RAM, 2 TB Fusion, 8GB 580X and I added 32GB of Crucial RAM. I returned my Corsair Vengeance because it was not registering as 2666.

Return it and get a smaller SSD. Fusion drives are incredibly slow now, let alone in 3-4 years time. Even Sata SSDs are slow by comparison to NVME SSDs.

 

[Colonel Blimp]

Return it and get a smaller SSD. Fusion drives are incredibly slow now, let alone in 3-4 years time.

No, they aren’t (at least not the 2 and 3 TB models with 128 GB of flash storage, which is what purduealum91 ordered).

See this post, this post, and this post (and this reply), written by persons who have actually used Fusion Drives.

 

[quagmire]

Figure I would post this here as well.... I limited Cinebench to 6 threads to see how the 9900K would perform at a task that can only take advantage of 6 threads/cores. The 9900K ran at 4.5 Ghz and got a 2749 score on Cinebench. So it can still outperform the 9600K going at full tilt. Whether that performance boost is worth the cost of the 9900K is of course up to you.

 

[Colonel Blimp]

Figure I would post this here as well.... I limited Cinebench to 6 threads to see how the 9900K would perform at a task that can only take advantage of 6 threads/cores. The 9900K ran at 4.5 Ghz and got a 2749 score on Cinebench. So it can still outperform the 9600K going at full tilt. Whether that performance boost is worth the cost of the 9900K is of course up to you.

Wouldn’t you need to limit Cinebench to 12 threads to max out six cores on the i9, since (unlike the i5) the i9 has Hyper-Threading? Limited to six threads, you’d only be maxing out three cores, no?

Edit: No. Now that I’m not tired, I realize that limited to six threads, the i9 should assign each thread to a separate core, even with Hyper-Threading.

 

[quagmire]

Wouldn’t you need to limit Cinebench to 12 threads to max out six cores on the i9, since (unlike the i5) the i9 has Hyper-Threading? Limited to six threads, you’d only be maxing out three cores, no?

The 9600K is only capable of 6 threads. Even if the 9900K was only using 3 cores during that run, still faster than the 9600K running full on at 6 cores.

I do not know the logic of hyperthreading though whether it will peak the physical cores first before trying to send another thread through the same core.

EDIT: I did a run limited to 12 threads. I saw 4.0-4.1 Ghz with a score of 3829.

 

[kaintxu]

The 9600K is only capable of 6 threads. Even if the 9900K was only using 3 cores during that run, still faster than the 9600K running full on at 6 cores.

I do not know the logic of hyperthreading though whether it will peak the physical cores first before trying to send another thread through the same core.

EDIT: I did a run limited to 12 threads. I saw 4.0-4.1 Ghz with a score of 3829.

So this means for the i9 we have:

- 6 core/6 threads: 2749 at 4.5GHz
- 6 core/12 threads: 3829 at 4.1 GHz
- 8 core/16 threads: 4069 at 3.8 GHz

Cinebench for the i5 was at 4.1 GHz (someone correct me if I'm not wrong):

- 580X: 2595
- Vega 48: 2487


So we can assume, that as much as the I9 is capped, it's still, not just slightly, but extremely more powerful than the i5, and way more future proof, probably being able to outlast the i5 for a good couple of years?

Did this to have it all together in one place.

 

[propower]

So this means for the i9 we have:

- 6 core/6 threads: 2749 at 4.5GHz
- 6 core/12 threads: 3829 at 4.1 GHz
- 8 core/16 threads: 4069 at 3.8 GHz

Cinebench for the i5 was at 4.1 GHz (someone correct me if I'm not wrong):

- 580X: 2595
- Vega 48: 2487


So we can assume, that as much as the I9 is capped, it's still, not just slightly, but extremely more powerful than the i5, and way more future proof, probably being able to outlast the i5 for a good couple of years?

Did this to have it all together in one place.

Cool summary!!!! Thanks!
Limiting the i9 to 6/12 looks like a real winner. For the audio work I do speed is very important - but when mixing - cores/threads are just as important. The simple chart shows precisely how the Core/Thread benefit is slowing down at 6/12. 1060 points to go 6/6 to 6/12. 240 points for 6/12 to 8/16. Being able to limit the i9 Core/Thread counts for the task at hand is a very cool trick and may be the thing that makes me trade up from my 2017 i5.

 

[Bohemien]

Being able to limit the i9 Core/Thread counts for the task at hand is a very cool trick and may be the thing that makes me trade up from my 2017 i5.

As I understood it, it's only possible in Cinebench because the program allows it. So unless Logic lets you restrict itself to a number of cores to run on this won't help you there I'm afraid.
[doublepost=1554242509][/doublepost]

So we can assume, that as much as the I9 is capped, it's still, not just slightly, but extremely more powerful than the i5, and way more future proof, probably being able to outlast the i5 for a good couple of years?

Thanks for putting this together, nice to have all the figures in one place!

I really don't understand Apple's CPU strategy. IMHO, looking at the benchmarks, it would've made more sense to configure the 27" models as follows:

1. base: i5-8500 (as it is)
2. mid-tier: i5-9600K, with an option to go for the i7-8700 (which is only available in the 21.5" top-tier model)
3. top-tier: i7-8700, with the i9-9900K as an option

As it is, to me it makes absolutely no sense to opt for the i5-8600 in the mid-tier model, only if you want the TDP of 65W. Performance and price-wise, once you choose an SSD, the difference makes getting the i5-9600K model a no-brainer. Too bad they didn't choose the i7-8700 as an option for the 27", it sits perfectly between the i5-9600K and the i9, and it has the 65W TDP.

Anybody here who got the maxed-out 21.5" (i.e. i7-8700 with Vega 20 GPU)? Could you maybe post some Cinebench results from that machine?

 

[propower]

As I understood it, it's only possible in Cinebench because the program allows it. So unless Logic lets you restrict itself to a number of cores to run on this won't help you there I'm afraid.

X-code (instruments)= choose # of cores and HT On/OFF (though i have to double check the HT part - been a while since I used that 🙂... yes it still does...

 

[Bohemien]

Cool, I didn't know that! 😎

 

[Colonel Blimp]

I do not know the logic of hyperthreading though whether it will peak the physical cores first before trying to send another thread through the same core.

EDIT: I did a run limited to 12 threads. I saw 4.0-4.1 Ghz with a score of 3829.

Now that I’m not tired, I realize that limited to six threads, the i9 should assign each thread to a separate core, even with Hyper-Threading.

Limited to 12 threads, four of the eight cores would have been running two threads each, and the other four cores would each have been running a single thread. Your 12-thread test scored higher than your six-thread test partly because you had 33% more cores at work.

Note, however, that your 12-thread benchmark score of 3829 was 39% higher than your six-thread score of 2749, so it would seem that Cinebench benefits from the Hyper-Threading in the i9.

 

[kaintxu]

While I know this is an i9 video, talking about the vega 48, what do you guys think, will the performance for this games be the same with the i5? Have a look at this video.