|
[Sponsors] |
![]() |
![]() |
#21 | |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
Quote:
Anyway, it worked out well since you opened my eyes for the memory topology of the amd motherboards. Maybe I should, but I didn't know about that, thanks for pointing it out! About 4, would that be due to the cpu overcocking on the 7820X? In my mind, in that specific case, it makes sense that it would have a more significant impact than in more memory starved setups as there aren't as many cores to make use of all the bandwidth. That being said, maybe I should consider trying to overclock the 9800X just a tad. ![]() Thank you for taking the time to help me out, once again! This pretty much solves it, the Ryzen setup is starting to get too many "ifs" for a lower maximum potential performance for my taste. The 3600 would have been killer in terms of price/performance if it came any closer to the 9800X, but it hasn't been found to have great results so far, so I gotta go with what I have so far. I might as well just pay a little more and be more confident that I'll get even better results. I'll still wait for zen 3 just in case, but I'll keep an eye out for good deals on RAM and the x299 stuff. |
||
![]() |
![]() |
![]() |
![]() |
#22 |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
Hello again!
In my quest for buying the components, I have found a deal that's intriguing me: a used Asus ROG X299 E-Gaming + i7-7800x for 450 euros, and the components seem to be in good shape. Would you consider this a good deal? I have seen this cpu being recommended here, but I'm not sure if it is still worth it. |
|
![]() |
![]() |
![]() |
![]() |
#23 |
Senior Member
Join Date: May 2012
Posts: 551
Rep Power: 16 ![]() |
Since you give this statement I assume that you have physical access to the setup. I suggest that you run a few benchmarks with OpenFOAM for instance to assess the performance. The 7800X is probably starved by low core count. However, on the bright side is that you will probably see improvements when overclocking the CPU. To reach high overclocks you might need to delid the CPU and that may not be something you like to do. If you are not afraid to damage your hardware then it is well worth it if you plan on overclocking. |
|
![]() |
![]() |
![]() |
![]() |
#24 |
Super Moderator
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,426
Rep Power: 49 ![]() ![]() |
The CPU, while not necessarily a bad choice for a budget CFD workstation, is worth less than 150. I would not spend 300 on a used X299 motherboard.
|
|
![]() |
![]() |
![]() |
![]() |
#25 | |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
Quote:
Oh, I'd be up for overclocking the cpu mildly, but deliding is going a bit too far for me. Thank you both for helping me out here again. |
||
![]() |
![]() |
![]() |
![]() |
#26 |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
So, Zen 3 has kinda exploded into the scene. From what I have read, the memory controller is still the same, but it can go up to 3800MHz or 4000MHz when you get a good sample. It seems it was a big leap comparing to zen 2 in pretty much everything, but I don't know what to make of it for CFD.
Do you think there's a reason for me to reconsider my purchasing decisions, or am I still good because of the bandwidth provided by the quad-channel in the 9800x? Edit: There is an interesting FEM benchmark here: https://www.techpowerup.com/review/a...-5800x/10.html. There's a good performance increase, apparently. That being said, I think I've seen this benchmark once in this forum, and I'm not sure it scales with memory bandwidth as much as most CFD codes. |
|
![]() |
![]() |
![]() |
![]() |
#27 |
Senior Member
Join Date: May 2012
Posts: 551
Rep Power: 16 ![]() |
I think 9800X is still a good value.
The Euler3D benchmark is not very representative for the common FVM codes. Also, many of the presented result just don't make sense. Finally, looking at the other results and comparing them to other sites, it seems the reliability of the site you link is not very high, imho. Unfortunately the entire 5000 series sold out in less than one minute in Sweden upon release. Otherwise I would have presented you with some real numbers. The next shipment is in the beginning of December. If you cannot wait then perhaps you can lurk about on hardware forums and try to get someone to run the OpenFOAM benchmark for you. ![]() |
|
![]() |
![]() |
![]() |
![]() |
#28 |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
For sure, I think so too! Actually, with the price increases by AMD, I'm finding it a bit more expensive to build a zen 3 machine than an x299 one right now (with 8 cores). I also believe that the 9800x probably has some more headroom for improvement with overclocking of both cpu and memory (speculating a bit here), but in case it becomes very very close between the two, the 5800x would become the better option (damn it, AMD, I wanna order the thing!).
Alas, we'll only know with numbers, and thank you for offering to provide those! ![]() In the mean time, I'll try to ask around, as I've just ordered a motherboard I found a good deal on, but that can still be returned to the store, if need be. As for the benchmark/website, I agree with you. It was a mistake by me, thank you for pointing it out. |
|
![]() |
![]() |
![]() |
![]() |
#29 |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
Ok, so I've decided to go X299 after all this time.
But I have one last question: I've found a few used 10900x with warranty at about 100 euros over the 9800x (new). I've noticed it has a better memory controller than the 9900x and the 9800x. However, I'm not sure this is worth it. What do you think? I do realize I'm stepping a bit above my original budget, though. I can overclock the 9800x a little, and it might compensate for the 2 extra cores, but if the memory controller is better I can most likely see an improvement in performance with the 10900x. Other than that, it's the old thing of used vs new. |
|
![]() |
![]() |
![]() |
![]() |
#30 |
Senior Member
Join Date: May 2012
Posts: 551
Rep Power: 16 ![]() |
Sorry, I'm not sure about this. For only 100 Euro (with warranty!) then the 10900X may be a better choice. I think you will be happy with either, as long as you have reasonable expectations on the performance
![]() Make sure to get good memory and cross your fingers that the IMC of your CPU is a good one. |
|
![]() |
![]() |
![]() |
![]() |
#31 | |
New Member
Francisco
Join Date: Sep 2018
Location: Portugal
Posts: 27
Rep Power: 8 ![]() |
Quote:
![]() I believe the 9800X is enough for what I need and if I need to go above that, there are workstations at the uni for that. Also, I'll post some benchmark results in the benchmarks thread. This cpu has been suggested many times here, so I'm sure it can come in handy. Off topic: If you're still getting the 5800X, It'd be great to find out just what that alternative can do, even if I'm not getting it myself. Once again, thank you a bunch! |
||
![]() |
![]() |
![]() |
![]() |
#32 | |
Member
Join Date: Aug 2012
Location: Italy
Posts: 68
Rep Power: 14 ![]() |
Quote:
In an international eshop each CPU is sold at 50€. The motherboard would be bought as new (Asus WS c621 Sage) at 500€. 8x12 = 92 GB DDR4 ECC 2666 MHz memory would cost around 300€. Adding a fast drive and a good PSU, it all would cost around 1100/1200€. I would be extremely interested to hear your qualified opinion. Thanks so much in advance! Last edited by Stabum; May 5, 2024 at 14:32. |
||
![]() |
![]() |
![]() |
![]() |
#33 | |
Member
Join Date: Aug 2012
Location: Italy
Posts: 68
Rep Power: 14 ![]() |
Quote:
thanks in advance |
||
![]() |
![]() |
![]() |
![]() |
#34 |
Super Moderator
Alex
Join Date: Jun 2012
Location: Germany
Posts: 3,426
Rep Power: 49 ![]() ![]() |
Not sure if I can give much more input here.
These CPUs are surely a bargain at 50$ each. The cheap CPUs are offset by the expensive motherboards, but that's just how it is with these setups. Be sure to get reg ECC memory. Unbuffered memory may or may not work, it's just not worth the risk. And add 2 CPU coolers, a case, and a graphics card to your total cost. |
|
![]() |
![]() |
![]() |
![]() |
#35 | |
New Member
Join Date: Sep 2018
Location: Saarbrücken, Germany
Posts: 9
Rep Power: 7 ![]() |
Quote:
I don't know about the availability of that system in your country, but you could try finding a Thinkstation P720 with Dual CPU configuration for a good price. There are also Dell Precision 7820 or HP Z8 G4 workstations with corresponding Dual Xeon Gold (Xeon SP 1st Gen) CPUs equipped, but they tend to sell for higher prices than P720s online. Anyways, I got a cheap deal on a used P720 two months ago and just finished upgrading it today. I couldn't benchmark it yet since I'm exclusively working with Fluent at the moment and so far have only worked with OpenFOAM two times on systems someone provided to me ready to use. But I could observe 4.5 faster calculation times on my current simulation job compared with my old i5 6600k + 4x 8GB DDR4 @ 2400MHz. (Normally, I should expect a speed up factor of up to ~8 according to the results in the benchmarking thread: OpenFOAM benchmarks on various hardware, so I need to look into this at a later point.) Hardware prices incl. shipping costs:
I can probably sell the Quadro P600, 2x16GB RAM, 690W PSU and 2x Xeon Gold 5122 for at least 30 + 40 + 80 + 40 = 190 €. Maybe even more. That reduces the net total to around ~ 750 €. From what I gathered online, those systems are even cheaper in the US than in Europe, but imho that's still a great value. It will also be possible to upgrade the CPUs to a 2nd Gen Xeon SP when those get cheaper on ebay. This should noticeably accelerate simulation times around ~20%, especially when paired with slightly faster DDR4-2933 instead of DDR4-2666, according to the benchmarking thread. |
||
![]() |
![]() |
![]() |
![]() |
#36 |
Senior Member
Join Date: Jun 2011
Posts: 205
Rep Power: 16 ![]() |
||
![]() |
![]() |
![]() |
![]() |
#37 |
Senior Member
Join Date: May 2012
Posts: 551
Rep Power: 16 ![]() |
I think the OF benchmark used here is a good one.
If you cannot run that then it is probably easiest to just look at the peak memory bandwidth of the system and construct a rating from that. After you have sorted the list you could also sort it according to CPU architecture age as newer architectures generally provide better results for the same memory bandwidth. Finally you could add a column for CPU cache as that has a positive benefit. The Intel tool for loaded latency gives some indication of all this, but it is of course better to run the OF benchmark, simply due to the references presented there. https://www.intel.com/content/www/us...y-checker.html |
|
![]() |
![]() |
![]() |
|
|
![]() |
||||
Thread | Thread Starter | Forum | Replies | Last Post |
CPU choice | zero_custom | Hardware | 3 | January 12, 2016 07:23 |
Multi 5820k CPU in a desktop or workstation? | sharonyue | Hardware | 5 | December 23, 2015 14:35 |
Superlinear speedup in OpenFOAM 13 | msrinath80 | OpenFOAM Running, Solving & CFD | 18 | March 3, 2015 05:36 |
Star cd es-ice solver error | ernarasimman | STAR-CD | 2 | September 12, 2014 00:01 |
OpenFOAM 13 Intel quadcore parallel results | msrinath80 | OpenFOAM Running, Solving & CFD | 13 | February 5, 2008 05:26 |