|
|
Yeah I'll take a look. I'm not looking at OCing, but I do want to know why my machine is running so cool and underperforming under load.
Real programmers use butterflies
|
|
|
|
|
AMD had always the heat as negative point with their processors, they have got way better in this topic.
I run a Ryzen 7 5800xt with a Dark Rock 4 Pro tower as CPU cooler.
Having played some games and done my standard usage, never heard the CPU fan going high (yet).
M.D.V. 
If something has a solution... Why do we have to worry about?. If it has no solution... For what reason do we have to worry about?
Help me to understand what I'm saying, and I'll explain it better to you
Rating helpful answers is nice, but saying thanks can be even nicer.
|
|
|
|
|
I do not own a Ryzen but have been using AMD for a long time.
I bought one Ryzen 3 some two years ago for my brother and the temperatures were close to yours while benchmarking.
From the top of my head, if your problem is just with single core performance, it is probably a motherboard configuration issue.
AMD motherboards usually have a configuration to unlock single core frequency. In my 11 year old Asus motherboard it is called "core unlocker".
What it does, when enabled, is to allow the cores to independently increase their frequency (up to the turbo frequency) without increasing for all cores.
It might have happen that, during your benchmark, you probably have that feature disabled (the default) and the CPU thought that it hasn't necessary to increase the frequency of the entire CPU just so that your single task on that single core would run faster.
I suggest you enable that feature for the benchmark only because it can cause (not confirmed) system instability and shorten the life span of the CPU to have it enabled all the time, due to points (location of each core) in the CPU die heating much more than the rest of the CPU.
Alternatively, run more things while running the benchmark so that more cores are busy and the CPU throttles its frequency for all cores.
|
|
|
|
|
In anticipation of heat issues, I chose bequiet dark rock pro 4 coolers for both my itx boards, and love them. Overkill, they keep mine at close to human body temperatures, even in socal summer heat. Huge and spendy, but quiet -- you can check their site to see if it'll fit your board and case headroom. One of them was nicely discounted, snagged from Amazon warehouse; presumably returned by someone that neglected to check headroom before purchasing. Stay cool my friends. ~jm
|
|
|
|
|
I won't be upgrading the cooling while I'm using this mobo. 65 degrees under load is cool as a cucumber. Too cool really, given that my CPU is underperforming @ userbenchmark.com
Real programmers use butterflies
|
|
|
|
|
"Too cool really"
I am struggling to comprehend this possibility . . . 
Hope ye get it sorted, cheers!
~jm
|
|
|
|
|
Since no one else has asked, is your machine running Win 10 or Win 11? Win 11 has performance issues with AMD systems.
The difficult we do right away...
...the impossible takes slightly longer.
|
|
|
|
|
I'm running 10. I won't touch 11 for maybe another year or so.
Real programmers use butterflies
|
|
|
|
|
On both my asus boards, the default speeds/timing were really very slow, until I went to the memory manufacturer's to find proper timings, and changed these settings to match the claimed memory timings, and changed from default *power saving* profiles.
Perhaps explore the bios to check/adjust CPU speed, Dram frequency and timing, and FCLK frequency,and power settings; and under bios' Monitor tab, to verify true temps? It does no harm to look around enough to get familiar with default settings to become comfortable in there. I cautiously explore each tab and change only one thing at a time, reading the page explanations under each setting as i go to figure things i am unsure about. cheers
~jm
|
|
|
|
|
65 °C is very good if the CPU is running with all CCXs under full load. In recent years, for me and for friends and family, I have always assembled machines with Ryzen (from the 1000 version up to the latest 5000). I didn't experience any stability issues, except when the RAM was used out of specification. Sometimes 100 KHz too much on the RAM is enough to make a system unstable: then you have to play with the overclocking parameters to bring the system back to stability, but in my opinion, it is almost never worth it.
But, always always always, before buying the pieces, check their compatibility and glean in the forums looking for any problems with certain combinations. Obviously, opinions on the web must always be taken "cum grano salis".
For Ryzen CPU / APU with 65W of TDP usually, the supplied heatsink is sufficient. For versions with 105W of TDP, if the CPUs are always under stress, it is better to adopt more performing heatsinks.
However, at around 95 degrees (or slightly less) the CCXs slow down themselves.
|
|
|
|
|
I have an (I think related) issue that *is* a problem.
My CPU is significantly underperforming in benchmarks compared to other Ryzen 7 4750s.
It's not anything obvious like cheap RAM, or windows throttling. It's like the board refuses to push the chip to anything near capacity. It's not about the clockspeed - it clocks full under load. The EDC goes to 100% as well.
My board can change its performance profile I think but when I tried one of them I got a reboot under load too quickly for me to figure out for sure what happened.
Real programmers use butterflies
|
|
|
|
|
Being a Ryzen 7 4750 I guess the scenario is that of a laptop. In that case, having never owned one with a Ryzen, I can't give you any suggestions. However, 65 ° C on the die is a more than the acceptable temperature for a fully loaded Ryzen 7. You have to see if it is a problem with your unit, or if other models of the same series are running all at that speed. In the first case, of course, your hardware has problems. Otherwise, in the latter case, it's a badly born laptop model built with components that are not harmonized with each other. Speaking of desktop PCs instead, I noticed that chipsets make a difference. For example, the X570 with a Ryzen 5000 has no problem pushing RAM to the max with the most extreme configurations while remaining stable all the time. Other chipsets, X470 and B550, on the other hand, are more sensitive to the type of RAM installed and in some cases, I had to lower the RAM clock a little. Personally, I prefer to mount MSI brand motherboards, the PRO or MAX series, but I also found the GigaBytes stable. However, lately, my experience is limited to AM4 Moterbords as I never mounted an Intel machine again after the first Ryzen 1000 has appeared.
|
|
|
|
|
The profile is set to standard stuff. It was already selected in the bios when I bought the thing. I haven't changed anything there other than the one time I set CM5 Aggressive? or something and it caused a reboot as soon as it went under load, so I changed it back.
I have all stock cooling and the CPU itself is OC resistant anyway so I haven't really fiddled with overclocking anything.
Real programmers use butterflies
|
|
|
|
|
Your messages are getting flagged so I'm replying here because I can't stand the wait! =)
I can still read them via email though.
It's a Pro 4750G not a 4750 sorry. Not a laptop. Something of a monster desktop chip.
Here are excerpts from my CPU-Z report (sorry for the length)
Socket 1 ID = 0
Number of cores 8 (max 8)
Number of threads 16 (max 16)
Manufacturer AuthenticAMD
Name AMD Ryzen 7 PRO 4750G
Codename Renoir
Specification AMD Ryzen 7 PRO 4750G with Radeon Graphics
Package Socket AM4 (1331)
CPUID F.0.1
Extended CPUID 17.60
Core Stepping RN-A1
Technology 7 nm
TDP Limit 65.0 Watts
Tjmax 95.0 °C
Core Speed 3539.6 MHz
Multiplier x Bus Speed 35.75 x 99.0 MHz
Base frequency (cores) 99.0 MHz
Base frequency (ext.) 99.0 MHz
Instructions sets MMX (+), SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, SSE4A, x86-64, AES, AVX, AVX2, FMA3, SHA
Microcode Revision 0x8600106
L1 Data cache 8 x 32 KB (8-way, 64-byte line)
L1 Instruction cache 8 x 32 KB (8-way, 64-byte line)
L2 cache 8 x 512 KB (8-way, 64-byte line)
L3 cache 2 x 4 MB (16-way, 64-byte line)
Max CPUID level 0000000Dh
Max CPUID ext. level 8000001Eh
FID/VID Control yes
# of P-States 3
Memory SPD
-------------------------------------------------------------------------
DIMM # 1
SMBus address 0x50
Memory type DDR4
Module format UDIMM
Module Manufacturer(ID) TEAMGROUP Inc. (7F7F7F7FEF000000000000000000)
SDRAM Manufacturer (ID) SpecTek Incorporated (7F7FB50000000000000000000000)
Size 16384 MBytes
Max bandwidth DDR4-2401 (1200 MHz)
Part number TEAMGROUP-UD4-3200
Serial number 0202D900
Manufacturing date Week 47/Year 20
Nominal Voltage 1.20 Volts
EPP no
XMP yes, rev. 2.0
AMP no
JEDEC timings table CL-tRCD-tRP-tRAS-tRC @ frequency
JEDEC #1 10.0-10-10-25-35 @ 750 MHz
JEDEC #2 11.0-11-11-27-38 @ 825 MHz
JEDEC #3 12.0-12-12-29-41 @ 900 MHz
JEDEC #4 13.0-13-13-32-45 @ 975 MHz
JEDEC #5 14.0-15-15-34-48 @ 1051 MHz
JEDEC #6 15.0-16-16-37-52 @ 1126 MHz
JEDEC #7 16.0-16-16-39-55 @ 1200 MHz
JEDEC #8 17.0-16-16-39-55 @ 1200 MHz
JEDEC #9 18.0-16-16-39-55 @ 1200 MHz
XMP profile XMP-3200
Specification DDR4-3200
VDD Voltage 1.350 Volts
Min Cycle time 0.625 ns (1600 MHz)
Max CL 16.0
Min tRP 12.50 ns
Min tRCD 12.50 ns
Min tRAS 25.00 ns
Min tRC 37.50 ns
Min tRRD 3.75 ns
XMP timings table CL-tRCD-tRP-tRAS-tRC-CR @ frequency (voltage)
XMP #1 15.0-19-19-38-57-n.a @ 1500 MHz (1.350 Volts)
XMP #2 16.0-20-20-40-60-n.a @ 1600 MHz (1.350 Volts)
XMP #3 17.0-20-20-40-60-n.a @ 1600 MHz (1.350 Volts)
DIMM # 2
SMBus address 0x51
Memory type DDR4
Module format UDIMM
Module Manufacturer(ID) TEAMGROUP Inc. (7F7F7F7FEF000000000000000000)
SDRAM Manufacturer (ID) SpecTek Incorporated (7F7FB50000000000000000000000)
Size 16384 MBytes
Max bandwidth DDR4-2401 (1200 MHz)
Part number TEAMGROUP-UD4-3200
Serial number 0202D8CF
Manufacturing date Week 47/Year 20
Nominal Voltage 1.20 Volts
EPP no
XMP yes, rev. 2.0
AMP no
JEDEC timings table CL-tRCD-tRP-tRAS-tRC @ frequency
JEDEC #1 10.0-10-10-25-35 @ 750 MHz
JEDEC #2 11.0-11-11-27-38 @ 825 MHz
JEDEC #3 12.0-12-12-29-41 @ 900 MHz
JEDEC #4 13.0-13-13-32-45 @ 975 MHz
JEDEC #5 14.0-15-15-34-48 @ 1051 MHz
JEDEC #6 15.0-16-16-37-52 @ 1126 MHz
JEDEC #7 16.0-16-16-39-55 @ 1200 MHz
JEDEC #8 17.0-16-16-39-55 @ 1200 MHz
JEDEC #9 18.0-16-16-39-55 @ 1200 MHz
XMP profile XMP-3200
Specification DDR4-3200
VDD Voltage 1.350 Volts
Min Cycle time 0.625 ns (1600 MHz)
Max CL 16.0
Min tRP 12.50 ns
Min tRCD 12.50 ns
Min tRAS 25.00 ns
Min tRC 37.50 ns
Min tRRD 3.75 ns
XMP timings table CL-tRCD-tRP-tRAS-tRC-CR @ frequency (voltage)
XMP #1 15.0-19-19-38-57-n.a @ 1500 MHz (1.350 Volts)
XMP #2 16.0-20-20-40-60-n.a @ 1600 MHz (1.350 Volts)
XMP #3 17.0-20-20-40-60-n.a @ 1600 MHz (1.350 Volts)
Real programmers use butterflies
|
|
|
|
|
Seems like a good hardware configuration at a first look. What profile did you select in the BIOS for the RAM? I see your memory has a 3200 MHz XMP profile available. Did you select that profile in the BIOS?
|
|
|
|
|
Ryzen is spec'd to 95C. However standard practice for silicon is 65C if you want it to last. The lifetime vs heat is not a linear curve, you can kill a chip very quickly as you approach 125C.
|
|
|
|
|
I think it probably should run a little hotter under heavy load. My CPU isn't giving me performance numbers I expect of a Ryzen 4750G. The numbers aren't awful - they're within the ballpark(ish), but it's kind of like one of the cylinders isn't firing sometimes.
So when I look at the cooling numbers (running cooler than normal) and combine that with the performance numbers (running slower than normal) it leads me to believe my chip is being underutilized by my mobo and/or bios.
Real programmers use butterflies
|
|
|
|
|
|
If you're running Windows 11, there a a lot of driver issues that the chip manufacturers are working through yet. Also remember that you might be comparing your cpu to people that have tweaked theirs to maximize its performance, including not running the TPM module which has been shown to lower performance quite a bit.
|
|
|
|
|
All modern desktop grade CPUs have been running at those temperatures for a long time now. Intel included. Big SIs like Dell or HP only use the bare minimum needed cooling so this is also something to keep in mind.
65 C is a fantastic temp under load. My R5 3600XT hits 75 C under Prime95 stress test with a CoolerMaster EVO-212 black edition on it. If you are dissatisfied with your heat levels, you would need to invest in either a very high end air cooler like something from Noctua, Scythe or BeQuiet; or you would need to go water cooling with an AIO. All of this assumes you can even fit the coolers into your system. I custom build all my PCs so I know I can put in whatever cooler I might need to go to in the future.
As for voltages, the funny thing with Ryzen is that they actually do better with less voltage in many cases. I ran a full per CCX/core tune on my system and found that performance went up, some voltages were set lower and overall heat didn't change any that mattered.
|
|
|
|
|
I'd be thrilled with the temp if I wasn't also lagging in CPU performance compared to other setups with the same CPU. I suspect something with my mobo/bios configuration.
Real programmers use butterflies
|
|
|
|
|
|
I'm already running 2x16.
Anyway, I've got a bigger issue now, and I'll probably need a whole new case and cooling system.
Real programmers use butterflies
|
|
|
|
|
.. with working on my own home project for many years in a stop and go fashion.
Every time I stumbled on something I don't like, I improve it!
Even though it's my own code, 6 months later, there is always plenty of thing that need improvement it turns out.
And the end result grow increasingly functional yet simple at the same time. Loving the look of it! 
modified 23-Dec-21 1:42am.
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
