Ryzen 9 5900HX competitive with M1

[pshufd]

This is a Mini PC with the Ryzen 9 5900HX and it is 8 cores/16 threads. The 5900HX is a 45 watt part (compared to probably 20 for the M1) but it outperforms the M1 in Geekbench 5 multicore while being somewhat slower in single-core. The Mini PC takes up to two NVMe SSDs and 64 GB of RAM and comes with generous ports (by M1 mini standards). The price appears to be $899 but it doesn't exactly say what you get for that.

AMD can do this because they're using TSM's -1 process. I don't think that Intel can do something like this on their 10nm but they could prove me wrong. Of course the M1 is Apple's low-end chip. I don't think that anything else out there can touch the M1X in a SFF or mobile chip.

 

[mi7chy]

A lot of these Indiegogos are scams. By the time it ships, if it ships at all, it'll be obsolete with RDNA2 APU, DDR5, etc. coming out soon. Cooling looks mediocre, too, compared to even laptops. I wouldn't waste $900 on this. CPU performance wise, though, even 5800H is up to nearly double the M1 depending on workload while 5900HX is slightly upclocked version.

 

[mi7chy]

Better off just building a SFF PC with AMD 5700G APU on AM4 socket that's more flexible with regards to CPU and dGPU upgrades.

https://videocardz.com/newz/minisfo...d-ryzen-9-5900hx-cezanne-mini-pcs-performance

$360 AMD 5700G CPU
$200 Asrock B550-I mobo
$79 Crucial 16GB DDR4-3200
$135 Corsair SF600 PSU
$104 Cooler Master NR200P case
$120 WD Black SN750 1TB NVMe SSD

 

[LeeW]

Don't really see the point in comparing M1 to what is available in the non-Apple market. There are plenty of CPUs and systems that do their job just fine whatever your needs.

The real and relevant comparison as far as M1 goes is its performance vs other, older Apple devices. Which is significant and that is really all that should matter.

 

[k27]

In another forum, a DaVinci Resolve user compared an M1 against a Windows laptop (Asus g15: Ryzen 5900hs, 16GB Ram, 1TB SSD and 6GB NvidiaRTX 3060). The M1 is significantly slower and doesn't stand a chance with 4K 50fps material and multiple filters.
Everything stutters on the M1 and runs at maybe 10 FPS. The Asus runs almost smoothly. He also uses RAW material in some of his projects..
He says it takes an m1x or m2 for him to look at an Apple machine again that can compete.
He just laughs at the YouTubers because they are cherry-picking and testing things that have nothing to do with his day's work.

Edit:
I think it is this laptop: https://rog.asus.com/de/laptops/rog-zephyrus/2021-rog-zephyrus-g15-series/

 

[pshufd]

Don't really see the point in comparing M1 to what is available in the non-Apple market. There are plenty of CPUs and systems that do their job just fine whatever your needs.

The real and relevant comparison as far as M1 goes is its performance vs other, older Apple devices. Which is significant and that is really all that should matter.

The reason is to compare if you are deciding on one or the other. If the Windows version has significantly less performance, than the M1 is the stronger choice all else being equal.

 

[pshufd]

In another forum, a DaVinci Resolve user compared an M1 against a Windows laptop (Asus g15: Ryzen 5900hs, 16GB Ram, 1TB SSD and 6GB NvidiaRTX 3060). The M1 is significantly slower and doesn't stand a chance with 4K 50fps material and multiple filters.
Everything stutters on the M1 and runs at maybe 10 FPS. The Asus runs almost smoothly. He also uses RAW material in some of his projects..
He says it takes an m1x or m2 for him to look at an Apple machine again that can compete.
He just laughs at the YouTubers because they are cherry-picking and testing things that have nothing to do with his day's work.

Edit:
I think it is this laptop: https://rog.asus.com/de/laptops/rog-zephyrus/2021-rog-zephyrus-g15-series/

I was doing conversions last night and was wondering why they were slow and then noticed that the CPU was working pretty hard. If you have an 8, 12 or 16 core machine, it can beat the M1. I'm looking at the M1X when it comes out as that should be competitive with the best desktop chips from AMD.

 

[mi7chy]

He just laughs at the YouTubers because they are cherry-picking and testing things that have nothing to do with his day's work.

The mistake is equating hardware video encoding performance with general performance but maybe it was intentional FUD like 8GB on M1 is equivalent 16GB on other architectures. People with common sense don't claim PC laptops are 10x faster when they can play 8K60 AV1 while M1 can't.

 

[k27]

The mistake is equating hardware video encoding performance with general performance

I intentionally do not use the hardware video encoding of the M1, but software encoding on the CPU.

Hardware video encoding is usually worse in quality. A positive exception is perhaps Nvidia NVENC.
Hardware video encoding is only fast. When it comes to quality or file size, it's usually bad. But these YouTubers don't pay attention to that. My impression is that most of them are not particularly competent either. But I don't know all of them and I don't usually watch these superficial tests without depth and real knowledge.

Edit
"Seeing exact same thing. I haven't compared the VT vs x265 quality yet extensively but eye test, the VT version had to be encoded at 8K+ BR to yield comparable quality to x265 which was closer to 2K BR. And of course the resulting 9GB file vs 2.3GB file."

Mac Mini M1 H.265 encoding?

Seeing exact same thing. I haven't compared the VT vs x265 quality yet extensively but eye test, the VT version had to be encoded at 8K+ BR to yield comparable quality to x265 which was closer to 2K BR. And of course the resulting 9GB file vs 2.3GB file.

forums.macrumors.com

"You'll probably want to use software encoding if you care about quality and size. Feedback on M1 that I've seen hasn't been good with regards to quality and filesize."

With the M1, is there any encode by quality option in video toolbox? - HandBrake

M1 hardware video encoding: Obviously really poor quality. Or very large files.
I don't want either. Then it's better to use the CPU and a software encoder like x264 or x265, wait a little longer and get very good quality without huge file sizes.
But then the benchmarks are probably no longer so impressive...

 

[diamond.g]

Better off just building a SFF PC with AMD 5700G APU on AM4 socket that's more flexible with regards to CPU and dGPU upgrades.

https://videocardz.com/newz/minisfo...d-ryzen-9-5900hx-cezanne-mini-pcs-performance

$360 AMD 5700G CPU
$200 Asrock B550-I mobo
$79 Crucial 16GB DDR4-3200
$135 Corsair SF600 PSU
$104 Cooler Master NR200P case
$120 WD Black SN750 1TB NVMe SSD

Shouldn't you go with at least DDR4-3600 RAM because iGPU is bandwidth thirsty?

 

[mi7chy]

Shouldn't you go with at least DDR4-3600 RAM because iGPU is bandwidth thirsty?

You have to decide if the difference is worth the premium. iGPU would only be a temporary stop gap until one can find a dGPU and the difference is tiny with only a frame or few more so it's better to put the money towards a dGPU.

 

[diamond.g]

You have to decide if the difference is worth the premium. iGPU would only be a temporary stop gap until one can find a dGPU and the difference is tiny with only a frame or few more so it's better to put the money towards a dGPU.

That is fair. Originally I was going to write something about 1:1 fclk, but removed it because I wasn't sure there was enough of a performance benefit (I think Cinebench gives higher numbers with higher fclk).

 

[Kung gu]

People with common sense don't claim PC laptops are 10x faster when they can play 8K60 AV1 while M1 can't.

You can compare it. If a PC has AV1 encoding and M1 does not it means in that sense the PC is faster.

This what Intel and AMD do in their marketing materials. If a CPU has better h.265 playback, Intel will tell you its X times than before.

and the vice versa can be done with M1.

 

[Kung gu]

In another forum

Can you share the forum link?

a DaVinci Resolve user compared an M1 against a Windows laptop (Asus g15: Ryzen 5900hs, 16GB Ram, 1TB SSD and 6GB NvidiaRTX 3060). The M1 is significantly slower and doesn't stand a chance with 4K 50fps material and multiple filters.

You know that the M1 is a 10-25W SoC. It's also a 4 Performance and 4 low power chip. It does not even have full FAT Performance 8 cores.
Of Course the AMD 5900HS will beat the M1 in performance, it has 8 Performance cores and the default TDP is 35W and it will go higher when it boosts.

The 5900HS scores 13158 in Cinebench r23 when using 80Watts of power.
The M1 scores around 7700 in Cinebench r23 around 7700.

The M1 scores around 7700 in Cinebench r23 around 7700.

The RTX 3060 will also definitely beat the M1 in every thing expect efficiency. The M1 GPU is around a GTX 1050/ti.

I would wait to compare it with the 14" and 16" MBP later this year.

 

[DeepIn2U]

Better off just building a SFF PC with AMD 5700G APU on AM4 socket that's more flexible with regards to CPU and dGPU upgrades.

https://videocardz.com/newz/minisfo...d-ryzen-9-5900hx-cezanne-mini-pcs-performance

$360 AMD 5700G CPU
$200 Asrock B550-I mobo
$79 Crucial 16GB DDR4-3200
$135 Corsair SF600 PSU
$104 Cooler Master NR200P case
$120 WD Black SN750 1TB NVMe SSD

you left out the crucial part the dGPU cost.

 

[Kung gu]

you left out the crucial part the dGPU cost.

The AMD APU comes with an iGPU.

 

[mi7chy]

The 5900HS scores 13158 in Cinebench r23 when using 80Watts of power.
The M1 scores around 7700 in Cinebench r23 around 7700.

80W for the 5900HS is a typo since the default TDP is 35W. They probably duplicated the 80W dGPU TDP.

5800H configured for ~24W scores about the same as M1 in Cinebench R23 and it's only 7nm. So, 7nm Zen 3 mobile is equivalent to 5nm M1 in efficiency but benefits of Zen 3 mobile are you don't lose x64 compatibility and have higher CPU and dGPU performance ceiling. And, performance per $ PC laptops with dGPU destroy M1 MacBooks.

 

[Kung gu]

80W for the 5900HS is a typo since the default TDP is 35W. They probably duplicated the 80W dGPU TDP.

The TDP rises when the AMD boosts.
This is the same for Inyel CPUs too.

5800H configured for ~24W scores about the same as M1 in Cinebench R23 and it's only 7nm. So, 7nm Zen 3 mobile is equivalent to 5nm M1 in efficiency but

If the 5800H scores 7829 with 24W, that's bad.

Why are you happy with that??
The 5800H is a full 8 core processor and yet it scores the same as a M1 which has 4 full cores and it scores 7830. Shouldn't the multicore of the 5800H be much higher considering the AMD has more big cores at around the same watts.