[SOLVED] Can anyone tell me which would perform better out of these cpus?

zenonithus

Distinguished
Aug 12, 2009
81
0
18,630
Hi, I am currently running a dell PC which has a dual cpu setup. So 2 x Intel Xeon E5-2687W 0 @ 3.10GHz, 3101 Mhz, 8 Core(s). So 16 cores in total

I was thinking of upgrading to a single cpu system that has an AMD Ryzen 7 5800X, 8C/16T, 3.8GHz – 4.7GHz

Obviously the Rizen will be more powerful that one of the Xeons. Though is it more powerful that two of the Xeons? Is there much of a difference?
 
Solution
That depends on your workload. If your workloads are memory bandwidth sensitive, you are likely to see a performance drop, especially if you intend to - Assuming E5-2687W v3 - keep your old DDR4 memory in your new build. 1x dual-channel DDR4, no matter how hightly clocked, is just not going to compete with 2x quad-channel DDR4.

Actual throughput may also see a downgrade, even though single-thread performance may nearly double.

A closer comparison in computational throughput might be had in 5900X with 12 cores, though it's limited to the same dual-channel DDR4 memory interface.

Both of them are going to consume only a small fraction of the power your old Intel Xeon system did, however.

EDIT: And if you wanted something really...

Howardohyea

Commendable
May 13, 2021
259
64
1,790
AMD would be faster, single threaded as well as multi threaded compared to two Xeons.

Also are you sure about the 5800X? I heard that the i7 11700 (non-k) is about the same speed (maybe slightly slower, depending on the task) of a 5800X and it's cheaper
 

JWNoctis

Respectable
Jun 9, 2021
443
108
2,090
That depends on your workload. If your workloads are memory bandwidth sensitive, you are likely to see a performance drop, especially if you intend to - Assuming E5-2687W v3 - keep your old DDR4 memory in your new build. 1x dual-channel DDR4, no matter how hightly clocked, is just not going to compete with 2x quad-channel DDR4.

Actual throughput may also see a downgrade, even though single-thread performance may nearly double.

A closer comparison in computational throughput might be had in 5900X with 12 cores, though it's limited to the same dual-channel DDR4 memory interface.

Both of them are going to consume only a small fraction of the power your old Intel Xeon system did, however.

EDIT: And if you wanted something really comparable to your old system in terms of market positioning, then you can get a new Xeon workstation build, or wait for Zen 3 Threadripper.

EDIT2: You may also run short of PCIe lanes, if you use case needed more than what typical consumer-grade hardware provides these days ( i.e. 20 lanes + whatever the chipset provided sharing another 4 lanes, of PCIe 4.0).
 
Last edited:
  • Like
Reactions: Howardohyea
Solution
Even if memory bound, so it can benefit from quad channel bandwidth, wouldn't any application also have to be very aware of the CPU it's running on so it doesn't incur latencies should a thread be moved between processors? So wouldn't it depend entirely on the application for that dual Xeon system to make any sense?

In other words, it would almost have to be written for a dual CPU system. That's not uncommon as many enterprise-scale applications are written with specific, and usually Intel, system arch's in mind.
 

Jacob 51

Notable
Dec 31, 2020
555
20
915
Hi, I am currently running a dell PC which has a dual cpu setup. So 2 x Intel Xeon E5-2687W 0 @ 3.10GHz, 3101 Mhz, 8 Core(s). So 16 cores in total

I was thinking of upgrading to a single cpu system that has an AMD Ryzen 7 5800X, 8C/16T, 3.8GHz – 4.7GHz

Obviously the Rizen will be more powerful that one of the Xeons. Though is it more powerful that two of the Xeons? Is there much of a difference?
That Ryzen is pretty powerful. You'll notice a lot of difference
 

chaoyang

Reputable
Oct 24, 2019
384
44
4,790
Hi, I am currently running a dell PC which has a dual cpu setup. So 2 x Intel Xeon E5-2687W 0 @ 3.10GHz, 3101 Mhz, 8 Core(s). So 16 cores in total

I was thinking of upgrading to a single cpu system that has an AMD Ryzen 7 5800X, 8C/16T, 3.8GHz – 4.7GHz

Obviously the Rizen will be more powerful that one of the Xeons. Though is it more powerful that two of the Xeons? Is there much of a difference?
You could try using this to compare GPU (Although it may not be 100% accurate somethimes):
https://www.game-debate.com/hardware/compare?hardwareType=cpu
 

zenonithus

Distinguished
Aug 12, 2009
81
0
18,630
Thanks guys for taking the time to reply. I'm not too technical though I think I can understand what you mean. The energy consumption is a concern with the duel processors so a single one would be beneficial. I work in zbrush and there are operations that can be very cpu intensive so I have no idea if it is taking advantage of both.

Also the Rizen7 5800 is surprisingly cheap CPU compared to the Rizen9. Infact it's pretty much the same price as one second hand Xeon E5-2687W 0, looking online.

Also I have 128gb memory in my dell which is described as Quad channel; up to 512GB 1600MHz or 1333MHz ECC RDIMM memory; 16 DIMM slots (8 per processor) (512GB config uses 1333MHz 32GB LRDIMMs)

I take it 32gb DDR4 would be better than this in the new system?
 
The Xeon has 8 cores, but 16 processing threads and a passmark rating of 9699.
That is when all 16 threads are fully utilized. Only some apps can do that. The single thread rating is 1717. Since you have two of them, you effectively have 32 threads and a rating of 2x 9699 or 19398.
The 5800X has 16 threads and a rating of 28656/3507.
It will be much stronger.
In part because of the better single thread performance which is important in driving all threads.
A technical reason is "amdahl's law"
https://en.wikipedia.org/wiki/Amdahl's_law
 

zenonithus

Distinguished
Aug 12, 2009
81
0
18,630
Ah ok thanks Geofelt, that makes sense. So my program may not even be using all the 16 threads. The higher rating pretty much tops the xeon. Ok cheers guys I think I will go with the AMD. IS there a general preference when it comes to apps like Maya or Zbrush when using AMD or intel? Or are they all the same?
 
...
Few, if any, application-level softwares actively preferred one over the other, to my understanding.
One would like to think so...maybe not actively preferred.

I've been aware that some are though, MatLab is one that comes to mind right off. My understanding is they are compiled with Intel compiler that by default disables switches which, if enabled, would tell the SW to check for an AMD processor and enable the full feature set of the processor when run. These aren't by any means 'main-stream' type applications, mostly scientific, math, engineering applications that I've read about. There are also cases where you can force the software to enable the features anyway at run-time, but not everyone will find out about or do that. Many will just be disappointed and think it an inadequacy in their new system when it fails to perform as expected because something like SSE2 or AVX2 instructions are ignored.
 
Last edited:
One of my friends uses matlab.
To date, it is very much single threaded.
To that end, look for the best single thread performance you can find.
5800x and 5900x have the same single thread performance, differing onlyby the number of processing threads, namely 16 vs.24

Intel 11th gen is also an option.
16 threads for i9-11900K with a rating of 28082/3764.
I7-11700K 16 threads and a rating of 26044/3573
 
  • Like
Reactions: Phaaze88

JWNoctis

Respectable
Jun 9, 2021
443
108
2,090
One would like to think so...maybe not actively preferred.

I've been aware that some are though, MatLab is one that comes to mind right off. My understanding is they are compiled with Intel compiler that by default disables switches which, if enabled, would tell the SW to check for an AMD processor and enable the full feature set of the processor when run. These aren't by any means 'main-stream' type applications, mostly scientific, math, engineering applications that I've read about. There are also cases where you can force the software to enable the features anyway at run-time, but not everyone will find out about or do that. Many will just be disappointed and think it an inadequacy in their new system when it fails to perform as expected because something like SSE2 or AVX2 instructions are ignored.
And that is fixed, a few months after it was found out.

I do concede that for softwares with a lower profile, such ... intrigue ... might never be found out, or fixed if it was.

All the more reason to read user experiences, and benchmarks where such is available.
 
...
All the more reason to read user experiences, and benchmarks where such is available.
For enterprise-scale apps, which tend to be very expensive, you should also stay really close to hardware configurations the software designers recommend or certify. In many cases you really have to in order for them to honor their performance guarantees. At several thousands dollars a seat for licenses alone that can be a costly oversight.
 
Last edited:
  • Like
Reactions: JWNoctis