Whats wrong with AMD? Benchmarks 10900 vs 3950x vs i7 NUC

Hi guys I would like to give you an idea of my plotting performance because I am actually quite disappointed by my 2 AMD 3950x rigs. Check this out:

3950x (16 Cores, 32 Threads) , 64GB 3200, Two 2TB Corsair MP600 Force in a RAID 0, 2TB Samsung 870 Evo for tmp2, Windows 10 and Plotmanager with these settings:

size: 32
bitfield: true
threads: 6
buckets: 128
memory_buffer: 4500
max_concurrent: 12
max_concurrent_with_start_early: 12
stagger_minutes: 30
max_for_phase_1: 4
concurrency_start_early_phase: 4
concurrency_start_early_phase_delay: 0

Here’s my 2nd 950x, 64GB 3200, Two 2TB Corsair MP600 Force in a RAID 0, tmp2 is set to the final ext HDD, Windows 10 and Plotmanager with these settings:

size: 32
bitfield: true
threads: 4
buckets: 128
memory_buffer: 4000
max_concurrent: 15
max_concurrent_with_start_early: 15
stagger_minutes: 30
max_for_phase_1: 6
concurrency_start_early_phase: 5
concurrency_start_early_phase_delay: 0
temporary2_destination_sync: true

My i9 10900 (10 Cores, 20 Threads) , 128GB RAM, two 1TB MP 600 Pro in RAID 0, tmp2 is set to the final ext HDD, Windows 10 and Plotmanager with these settings:

size: 32
bitfield: true
threads: 4
buckets: 128
memory_buffer: 6000
max_concurrent: 9
max_concurrent_with_start_early: 9
stagger_minutes: 30
max_for_phase_1: 4
concurrency_start_early_phase: 4
concurrency_start_early_phase_delay: 0
temporary2_destination_sync: false

and finally my NUC i7 6 Core, 12 Threads, 32GB RAM, One 1TB MP600 Pro, Windows 10 Plotmanager with these settings

size: 32
bitfield: true
threads: 4
buckets: 128
memory_buffer: 6000
max_concurrent: 4
max_concurrent_with_start_early: 4
stagger_minutes: 40
max_for_phase_1: 2
concurrency_start_early_phase: 2
concurrency_start_early_phase_delay: 0
temporary2_destination_sync: false

Is this just a matter of the performance difference between the MP600 Pro and the Force or am I doing anything wrong? It seems that the Intel systems are just so much faster.

I also try out different settings on the AMDs all the time but I can’t get anywhere near those Intel numbers.

Any clue or can anyone confirm these experiences?

2 Likes

I dont know about the difference between force and pro, but mp600 is not very fast anyway.

In any case, both amd systems are using 2tb disk, intel 1gb disks.

I think that is the biggest difference, the systained write speed of 1 and 2 tb disks is about the same from the tests i saw on tomshardware, but the 2tb has to write double the data.

1 Like

well the Pro is a PCIe Gen4 x4 and faster but why should a 2TB write double the data? This is something I’ve never heard of especially when you look at the various setups here in the forums. Most of the users are going for two 2TB M.2

And yes, the MP600 are fast!

By the look of it your not starting them far enough apart and your disk space is being overlapped after a certain amount of time.

1 Like

thank you… that’s a good point. I will try to go for 40 mins. Good point!

I mean that if a1 tb and 2tb drive have the same sustained write speed of 800mb/s and you write 3 plots to the 1tb model and 6 plots to the 2tb one, your writing double the data to the 2tb model but with the same total write speed limit

According to the review from toms, the mp600 has quite a low sustained speed. And I’ve seen more than a few ppl report its not very fast.

But the actual plot times per system sometimes seem to be governed by some darkmagic formula…so

1 Like