F5F Stay Refreshed Hardware Desktop Comparing dual and quad-channel RAM for data science: the key differences.

Comparing dual and quad-channel RAM for data science: the key differences.

Comparing dual and quad-channel RAM for data science: the key differences.

G
Gunner3212
Member
159
01-31-2020, 11:54 AM
#1
You're weighing options for a data science computer this summer. You're considering Ryzen 5900X or 5950X versus the upcoming Zen 3 Threadripper with 16 cores. The 5900X/5950X offer solid performance at a lower price, but the Threadripper supports quad-channel memory which could boost speed. Your workloads involve heavy parallel processing with libraries like future and multiprocessing, so memory bandwidth matters. Quad-channel memory may provide noticeable gains in those scenarios. As for extra reasons to choose Threadripper, it could handle large datasets more efficiently and future-proof your setup. Let me know if you need more details!
G
Gunner3212
01-31-2020, 11:54 AM #1

You're weighing options for a data science computer this summer. You're considering Ryzen 5900X or 5950X versus the upcoming Zen 3 Threadripper with 16 cores. The 5900X/5950X offer solid performance at a lower price, but the Threadripper supports quad-channel memory which could boost speed. Your workloads involve heavy parallel processing with libraries like future and multiprocessing, so memory bandwidth matters. Quad-channel memory may provide noticeable gains in those scenarios. As for extra reasons to choose Threadripper, it could handle large datasets more efficiently and future-proof your setup. Let me know if you need more details!

A
Atlas_343
Member
103
01-31-2020, 03:22 PM
#2
You might think about using dual channel to fully utilize the available memory capacity.
A
Atlas_343
01-31-2020, 03:22 PM #2

You might think about using dual channel to fully utilize the available memory capacity.

A
Acew25
Member
51
02-02-2020, 01:59 PM
#3
Your system handling most of its memory in bulk could benefit from quad channel, but other factors matter too. Your CPU’s workload versus memory operations might shift the balance—consider profiling to confirm. Using Python introduces the Global Interpreter Lock, which could limit thread benefits if data isn’t processed quickly enough. Also, define the time constraints for these tasks; even modest gains might not justify the cost if speed improvements are modest. A thorough analysis of your workflow and possibly a performance test is essential before deciding.
A
Acew25
02-02-2020, 01:59 PM #3

Your system handling most of its memory in bulk could benefit from quad channel, but other factors matter too. Your CPU’s workload versus memory operations might shift the balance—consider profiling to confirm. Using Python introduces the Global Interpreter Lock, which could limit thread benefits if data isn’t processed quickly enough. Also, define the time constraints for these tasks; even modest gains might not justify the cost if speed improvements are modest. A thorough analysis of your workflow and possibly a performance test is essential before deciding.

J
Josue47
Junior Member
30
02-03-2020, 10:47 PM
#4
The key questions to consider are essential. I have a project I aim to expand, and from my observations, data transfer between subprocesses is frequent. I haven’t analyzed the code yet, but it looks like quad-channel could offer clear gains for some of my work. A minimum improvement of around 10% across all projects would be needed to justify the upgrade. I’m hoping for a more definitive result, such as a 50% increase. (The maximum possible is 100% based on my understanding.) I’m confident quad-channel would help, but many articles suggest limited advantages. It’s tough to find solid benchmarks. So I’m wondering if others in a similar situation can share their experiences. I’d even consider spending more for an octo-channel Threadripper Pro if someone can prove it delivers big benefits. [0] https://www.pcworld.com/article/2982965/...mance.html
J
Josue47
02-03-2020, 10:47 PM #4

The key questions to consider are essential. I have a project I aim to expand, and from my observations, data transfer between subprocesses is frequent. I haven’t analyzed the code yet, but it looks like quad-channel could offer clear gains for some of my work. A minimum improvement of around 10% across all projects would be needed to justify the upgrade. I’m hoping for a more definitive result, such as a 50% increase. (The maximum possible is 100% based on my understanding.) I’m confident quad-channel would help, but many articles suggest limited advantages. It’s tough to find solid benchmarks. So I’m wondering if others in a similar situation can share their experiences. I’d even consider spending more for an octo-channel Threadripper Pro if someone can prove it delivers big benefits. [0] https://www.pcworld.com/article/2982965/...mance.html

K
KateKitKat
Member
138
02-04-2020, 08:58 PM
#5
Combining scientific computing with ample RAM makes ECC a suggested necessity.
K
KateKitKat
02-04-2020, 08:58 PM #5

Combining scientific computing with ample RAM makes ECC a suggested necessity.

P
Pyrus344
Junior Member
24
02-24-2020, 05:15 PM
#6
At this stage, any possible profits remain uncertain until your code is active and you can analyze its performance. If you already have a clear system design for your process, you can be fairly confident about the benefits depending on how you structure your code. Another idea that came to mind is that Threadripper offers ECC memory support, which could be useful if your work involves long-running tasks, providing error correction that might be valuable.
P
Pyrus344
02-24-2020, 05:15 PM #6

At this stage, any possible profits remain uncertain until your code is active and you can analyze its performance. If you already have a clear system design for your process, you can be fairly confident about the benefits depending on how you structure your code. Another idea that came to mind is that Threadripper offers ECC memory support, which could be useful if your work involves long-running tasks, providing error correction that might be valuable.

P
piff2222
Member
57
02-24-2020, 11:31 PM
#7
Thanks for your feedback. ECC is crucial for personal projects, though it may not be necessary for professional work. Regarding Threadripper, only Threadripper Pro officially supports ECC, while regular models do not. For evidence of improved performance, a recent blog post by Gamer's Nexus shows that dual-channel offers about a 17.7% boost in certain tasks, though results vary. The 2014 article is still relevant but may need updating for current comparisons. You might find the referenced piece useful: [Gamer's Nexus article](https://www.gamersnexus.net/guides/1349-...le-channel).
P
piff2222
02-24-2020, 11:31 PM #7

Thanks for your feedback. ECC is crucial for personal projects, though it may not be necessary for professional work. Regarding Threadripper, only Threadripper Pro officially supports ECC, while regular models do not. For evidence of improved performance, a recent blog post by Gamer's Nexus shows that dual-channel offers about a 17.7% boost in certain tasks, though results vary. The 2014 article is still relevant but may need updating for current comparisons. You might find the referenced piece useful: [Gamer's Nexus article](https://www.gamersnexus.net/guides/1349-...le-channel).