-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
***The Official 5900X \5950X owners thread***
- Thread starter easyrider
- Start date
It's been around that price for quite some time. As this CPU competes very well with the latest Ryzen gen 5 CPUs for gaming I don't think it's likely the price will go down much more for the foreseeable future.
Last edited:
Associate
Anyone seen the 5950x at a good price? I feel aggrieved paying over £500 for one where for an extra £50 I could get a 7950x (albeit with the additional cost of motherboard and ram)
Associate
- Joined
- 7 Nov 2017
- Posts
- 1,923
Unless you go second hand £529 on OC is the cheapest you'll find.
Associate
Yeah, thought as much. Will most likely keep good of my 3900x until the next round of AM5 CPUs come out.
Motherboards and ram should be a bit cheaper by then too
Motherboards and ram should be a bit cheaper by then too
Hi,
With the prices of new platforms as they are at the moment I have decided to replace my good old 3900x with 5950x that I was able to grab for £430. The idea is to skip the first AM5 series entirely.
I do have temperature problems with the new CPU though.
I'm using Alphacool 280mm AIO in a well ventilated case.
All 32 thread load is ~70*C with the fans not even spinning fast.
HW Info shows: CPU package power at 145W, CPU VDD 100A, clocks ~4375MHz for all cores under load.
With single thread load its 77*C but with 2 threads running its hitting 90*C that behavior is the same all the way to 5 threads. Then with 6+ threads under load the temperature starts dropping again.
The fans are at full speed in that situation which is loud and annoying.
I have put fresh Arctic Silver 5 on the new CPU and the cooler is fully tightened (it has springs on the mounting screws so pressure should be even)
I'm feeling like I'm already on AM5 lol. Any ideas why I'm seeing 90*C on that thing?
With the prices of new platforms as they are at the moment I have decided to replace my good old 3900x with 5950x that I was able to grab for £430. The idea is to skip the first AM5 series entirely.
I do have temperature problems with the new CPU though.
I'm using Alphacool 280mm AIO in a well ventilated case.
All 32 thread load is ~70*C with the fans not even spinning fast.
HW Info shows: CPU package power at 145W, CPU VDD 100A, clocks ~4375MHz for all cores under load.
With single thread load its 77*C but with 2 threads running its hitting 90*C that behavior is the same all the way to 5 threads. Then with 6+ threads under load the temperature starts dropping again.
The fans are at full speed in that situation which is loud and annoying.
I have put fresh Arctic Silver 5 on the new CPU and the cooler is fully tightened (it has springs on the mounting screws so pressure should be even)
I'm feeling like I'm already on AM5 lol. Any ideas why I'm seeing 90*C on that thing?
Caporegime
90 is a little high, but 70 all core load isn't bad at all. Have you tried setting -10 or -15 all cores in Curve Optimiser?
Last edited:
Soldato
I've got a H115i on mine (from 2016) and get similar temps in those scenarios but never getting much more than 85 absolute worse case. I'm using PBO on auto though, not got into refining it yet.
I did use TM from 2016 though so about to re-paste with Kryonaut and see if that helps, then I can optimise PBO properly.
I did use TM from 2016 though so about to re-paste with Kryonaut and see if that helps, then I can optimise PBO properly.
What load are you using to tests temps? Is it something like prime95 and is it an avx load?
What memory speed are you running also.
How old is the cooler? I know on my artic AIO they released an offset bracket after the 5000series released for better temps.
I'm using prime95 custom test with 1344K FFT size without disabling AVX2. Tested just now with disabled AVX2 and then p95 uses just AVX - temps are few *C lower on 2 threads ~86*C instead of constantly hitting Tjmax.
The AIO was bought in July 2020 so not very old.
Memory is a Corsair Vengeance 2x32GB kit 3200 and runs without OC
Last edited:
Cooler is old enough that the cold plate may not be making great contact with the core chiplets due to them being a decent bit off centre, an updated offset mount may exist for it. Artic must have released one for a reason.
I asked about the memory as if you loaded up an XMP profile it may have set the memory controller voltage higher than needed which all adds to the heat you are trying to dissipate.
Yes, XMP profile is loaded and that indeed raises voltage above default. The modules run at 1.35V
There is no offset mounting bracket for my pump-block combo.
CPU AIO
Komplettwasserkühlung für AMD- und Intel-Prozessoren mit Kupferradiatoren und Schnellverschlüssen, geeignet für alle CPU-Sockel.
www.alphacool.com
Last edited:
May be worth looking at the VSOC voltage if you can find it in your BIOS or you could use a program like zentimings and post up a screenshot. It's what I meant by memory controller voltage.
I don't think it's going to make a huge difference dropping it though, it's why I kept asking about the cooler.
Caporegime
Voltages look fine to me.
I have a feeling something is not entirely right with the cpu itself or how the mobo drives it.
Attaching HWinfo screens comparing 2t and 32t load
This illustrates that the cooling is doing just fine with 145W 32 thread load. Way lower fan speeds and nice temps compared to 2 thread load that is just bonkers - with cpu package power ~100W although concentrated in one chip/CCD I worry that the issue could be more chip-to-ihs thermal dissipation than it is ihs-to-waterblock or AIO efficiency problem.
Attaching HWinfo screens comparing 2t and 32t load
This illustrates that the cooling is doing just fine with 145W 32 thread load. Way lower fan speeds and nice temps compared to 2 thread load that is just bonkers - with cpu package power ~100W although concentrated in one chip/CCD I worry that the issue could be more chip-to-ihs thermal dissipation than it is ihs-to-waterblock or AIO efficiency problem.
Last edited:
Associate
I didn't realise until I tried your custom test just now, but it seems Prime95 now sets the core affinity to the first 'x' cores by default, so a 2 thread P95 run will be locked to core 0 and core 1. At least compared to an old v29 download I had, which just left it up to the Windows scheduler, which in turn mostly used the "best" cores.
So, it's a bit of both - the IHS has to get all the heat out of a very, very small area and if the waterblock isn't positioned exactly over that area it will struggle too. That's where the offset mounts will come in (at least for one CCD anyway).
Hitting 90C isn't good, but it's an unsual load - not much will insist on running flat out on two cores right next to each other.
Associate
IMHO, your screenshot show a normal 5950X to me. With a stock PPT, i'd most certainly expect a 2T workload to be higher temperature than a 32T workload.
Under heavy multicore a 5950X is super power limited, so the clocks and voltage drop. Even though peak power is higher, the power denisity is much much lower.
You can see it in your screenshots.
2T = 63.7 W core power / 2.3 busy cores = ~27.7 W per core
32T = 111.8 W core power / 16 busy cores = ~7.0 W per core
Easy to see why the 2T CPU core temps are way way higher. Your cooler has the cooling capacity, but the heat density is so high the heat doesn't get out of the chip fast enough through the STIM/IHS. This is a 'feature' of modern CPUs unfortunately.
That is a valid point with power denisity. And I have expected higher temp with less cores loaded. I guess i did not expect to be as high (hitting limit)
So for comparison i have rerun the same test forcing p95 to run on CCD2
Here is how it compares CCD1 vs CCD2
CCD2 clocks and voltages differ a little but I think it is still comparable power. Temps way lower on CCD2