Jump to content

I think AMD has a problem


Recommended Posts

I have a 3900x that hits a max multiplier of 45.5 on 3 cores, 45 on another and 44.3-44.8 on the rest. The bus clock is consistently shy of 100 at 99.8, which is the only variable that really seem to change between motherboards. I have to wonder if the boosting logic will allow the multiplier to overshoot 46 to make up for boards that have underclocked bus speeds. Currently overclocking my board's bus clock to 101 breaks boosting. Its really in the spirit tinkering and testing for me as I'm not sweating it that much.

Link to comment
Share on other sites

The bus clock being not at 100 is the reason of the post of mine above.

 

HWbot banned most/all Ryzen results due to that..as it affects ALL Benchmark results due to squirked time measurements

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

Because the benchmark programs arent coded properly to handle dynamic Bus Clks.

Windows 10 Pro, Ryzen 2700X @ 4.6Ghz, 32GB DDR4-3200 GSkill (F4-3200C16D-16GTZR x2),

ASRock X470 Taichi Ultimate, XFX RX6800XT Merc 310 (RX-68XTALFD9)

3x ASUS VS248HP + Oculus HMD, Thrustmaster Warthog HOTAS + MFDs

Link to comment
Share on other sites

Is the advertised Ghz/Mhz just a theoretical limit? The kind we see from advertised network card and internet speeds, but never actually achieve?

The advertised speed from AMD is the speed which one core can reach at maximum, not all cores. Even when only considering one core, it's not certain you will actually achieve the advertised clock. There are differences on how high the CPU(s) clock depending on the motherboard aswell, not all are the same here.

Link to comment
Share on other sites

New AGESA in 2 weeks

 

https://www.tomshardware.com/news/amd-ryzen-3000-boost-frequency-fix-bios-agesa-1.0.0.3abba-longevity-reliability,40368.html

 

25-50 more MHz planned... mine is still 200+MHz below spec on single core load. Maybe now if I’m lucky I will only be 150mhz slower than advertised.

Ryzen 7 5800X3D / Asus Crosshair VI Hero X370 / Corsair H110i / Sapphire Nitro+ 6800XT / 32Gb G.Skill TridentZ 3200 / Samsung 980 Pro M.2 / Virpil Warbrd base + VFX and TM grips / Virpil CM3 Throttle / Saitek Pro Combat pedals / Reverb G2

Link to comment
Share on other sites

New AGESA in 2 weeks

 

https://www.tomshardware.com/news/amd-ryzen-3000-boost-frequency-fix-bios-agesa-1.0.0.3abba-longevity-reliability,40368.html

 

25-50 more MHz planned... mine is still 200+MHz below spec on single core load. Maybe now if I’m lucky I will only be 150mhz slower than advertised.

 

Do you have PBO turned on? My 3900x get up to 4540, but I use PBO. There is also evidence of clock variance between motherboards that isn't the CPU itself being the source of the problem. A 200mhz deficit is a lot, and leads me to think its an Asus issue with the x370 boards.

 

https://www.guru3d.com/news-story/ryzen-7-3800x-show-different-boost-clocks-with-different-motherboards.html

Link to comment
Share on other sites

Do you have PBO turned on? My 3900x get up to 4540, but I use PBO. There is also evidence of clock variance between motherboards that isn't the CPU itself being the source of the problem. A 200mhz deficit is a lot, and leads me to think its an Asus issue with the x370 boards.

 

https://www.guru3d.com/news-story/ryzen-7-3800x-show-different-boost-clocks-with-different-motherboards.html

 

I’m wondering about the board myself. PBO is on (though PBO is the all core boost, PB2 is the single core boost)

 

I’ve tried several combinations of BIOS settings...

Ryzen 7 5800X3D / Asus Crosshair VI Hero X370 / Corsair H110i / Sapphire Nitro+ 6800XT / 32Gb G.Skill TridentZ 3200 / Samsung 980 Pro M.2 / Virpil Warbrd base + VFX and TM grips / Virpil CM3 Throttle / Saitek Pro Combat pedals / Reverb G2

Link to comment
Share on other sites

I have an AMD ryzen 1700x (manually overclocked)

When I initially purchased this processor, motherboard, and heatsink-fan combo I did not know specifics on how AMD processors overclocked so it was a big learning curve.

 

In a nut shell without getting into it too much the TurboBoost technology of AMD's processors isn't very good for benchmarking numbers.

 

The first core is the only core that gets TurboBoost (automagical overclock) and the frequency that it will hit is very much a function of Temperature and Voltage. If your particular system at the instant moment of time the TurboBoost AI wants to kick in spikes too much in temperature then it won't get that "advertised TurboBoost" number. If there isn't enough voltage being supplied then it won't get that advertised number.

 

There are a lot of factors that determine how much the TurboBoost will boost. Additionally it will only kick in for temporary amounts of time. The core0 load has to reach a certain threshold and it will only engage the TurboBoost temporarily (sometimes for less than 1 second) to get the processes processed faster.

 

If you're concerned about the automagic TurboBoost technology not being good enough then I would recommend manually overclocking the processor or purchasing an Intel processor. Just realize that TurboBoost technology in AMD or Intel processors isn't a guarantee.

 

They do not advertise otherwise: source - https://www.amd.com/en/products/cpu/amd-ryzen-9-3900x

go to bottom, mouse over the Max Boost Clock info.

 

The real world performance of the TurboBoost is very good, the benchmarking world for it is completely different because of its instantaneous on/off behavior to better grind the process queue on just the 1 core 1 thread at a time. If you're running a heavy benchmark that is utilizing all of your cores at 100% cpu usage the HEAT from that will prevent the TurboBoost from kicking in unless you have some really amazing cooler.

 

TLDR; AMD/Intel Automatic overclock technology is specifically designed for high single core/single thread usage games. A manual overclock will yield better results for everything else.

Link to comment
Share on other sites

I guess people are more concerned some chips not even reaching the advertised clock speed on one core, it differs from motherboard to motherboard. In my case the first core, which is almost as designed for DCS, could hit peak 4.5 or 4.6 and the others 4.1 or 4.2 or whatever as long as it is stable. 4.7 Ryzen 3000 is like ~5.1/5.2 9900K.

Link to comment
Share on other sites

......... 4.7 Ryzen 3000 is like ~5.1/5.2 9900K.

 

I don't like to assume things usually, but here I do assume that is achieved by the very large cache compared to the Intel chips. Keep it close to compute pays as it seems.

Gigabyte Aorus X570S Master - Ryzen 5900X - Gskill 64GB 3200/CL14@3600/CL14 - Asus 1080ti EK-waterblock - 4x Samsung 980Pro 1TB - 1x Samsung 870 Evo 1TB - 1x SanDisc 120GB SSD - Heatkiller IV - MoRa3-360LT@9x120mm Noctua F12 - Corsair AXi-1200 - TiR5-Pro - Warthog Hotas - Saitek Combat Pedals - Asus PG278Q 27" QHD Gsync 144Hz - Corsair K70 RGB Pro - Win11 Pro/Linux - Phanteks Evolv-X 

Link to comment
Share on other sites

  • 1 month later...
  • ED Team

I just tried that new AGESA 1.0.0.4 at my X570 with Ryzen 3900X and I get worse performance across CPU-Z, HW info, AIDA etc. than I had with AGESA 1.0.0.3ABBA. I have worse single and multicore performance and maximum boost is now just 4525 MHz. Does anyone else tried new AGESA 1.0.0.4 with Ryzen 3900X?

AMD Ryzen 9 3900X, GeForce RTX 2080Ti, 32 GB DRAM, HOTAS TM Warthog, FSSB R3 Lighting, MFG Crosswind, Win 10 Pro

Link to comment
Share on other sites

What software did you use to test the boost clock? I dont have much installed yet on my system, but i do have DCS up and running, and i just did a quick free flight in DCS and monitored behavior with the MSI afterburnner overlay and my cpu boosted to 4525 mhz and I saw it boost to at least 4600 once or twice, but not consistently - of note, is that cpu usage was all very low when it was doing this. Not a god test, as I think a single core cinebench test would be ideal to test this, but i did see it get to 3.6. I'm using AGESA 1.0.0.3 bios on an x570 gigabyte elite mobo. I may be wrong, but should the cpu be begged at its maximal boost, but only when it's demanded? I just hit the power button yesterday and might not update the bios to the 1.0.0.4 as i want to leave well enough alone. Again, this was a minor look see to see what my silicon was doing.


Edited by DocSigma

Ryzen9 5800X3D, Gigabyte Aorus X570 Elite, 32Gb Gskill Trident DDR4 3600 CL16, Samsung 990 Pr0 1Tb Nvme Gen4, Evo860 1Tb 2.5 SSD and Team 1Tb 2.5 SSD, MSI Suprim X RTX4090 , Corsair h115i Platinum AIO, NZXT H710i case, Seasonic Focus 850W psu, Gigabyte Aorus AD27QHD Gsync 1ms IPS 2k monitor 144Mhz, Track ir4, VKB Gunfighter Ultimate w/extension, Virpil T50 CM3 Throttle, Saitek terrible pedals, RiftS

 

Link to comment
Share on other sites

  • ED Team

I found the solution on the internet.:) So in AMD Ryzen Master application https://www.amd.com/en/technologies/ryzen-master I set 93 EDC in Profile 1 under PBO settings. Everything else I left to AUTO. Now it works even better than previous BIOS. It looks like Gigabyte did something opposite for EDC settings than it was in previous versions.

https://valid.x86.fr/8b0qw9

AMD Ryzen 9 3900X, GeForce RTX 2080Ti, 32 GB DRAM, HOTAS TM Warthog, FSSB R3 Lighting, MFG Crosswind, Win 10 Pro

Link to comment
Share on other sites

  • ED Team
What software did you use to test the boost clock? I dont have much installed yet on my system, but i do have DCS up and running, and i just did a quick free flight in DCS and monitored behavior with the MSI afterburnner overlay and my cpu boosted to 4525 mhz and I saw it boost to at least 4600 once or twice, but not consistently - of note, is that cpu usage was all very low when it was doing this. Not a god test, as I think a single core cinebench test would be ideal to test this, but i did see it get to 3.6. I'm using AGESA 1.0.0.3 bios on an x570 gigabyte elite mobo. I may be wrong, but should the cpu be begged at its maximal boost, but only when it's demanded? I just hit the power button yesterday and might not update the bios to the 1.0.0.4 as i want to leave well enough alone. Again, this was a minor look see to see what my silicon was doing.

 

For monitor the frequency you can use HW info https://www.hwinfo.com/download/. Right now I use 6.13.3970 beta version. For some simple single and multicore test you can try CPU-Z https://www.cpuid.com/softwares/cpu-z.html or cinebench.

AMD Ryzen 9 3900X, GeForce RTX 2080Ti, 32 GB DRAM, HOTAS TM Warthog, FSSB R3 Lighting, MFG Crosswind, Win 10 Pro

Link to comment
Share on other sites

On the 1.0.0.3 Aorus Elite board, it looks like my cpu behaves similar to yours, at it should.

https://valid.x86.fr/bg5fk1

Ryzen9 5800X3D, Gigabyte Aorus X570 Elite, 32Gb Gskill Trident DDR4 3600 CL16, Samsung 990 Pr0 1Tb Nvme Gen4, Evo860 1Tb 2.5 SSD and Team 1Tb 2.5 SSD, MSI Suprim X RTX4090 , Corsair h115i Platinum AIO, NZXT H710i case, Seasonic Focus 850W psu, Gigabyte Aorus AD27QHD Gsync 1ms IPS 2k monitor 144Mhz, Track ir4, VKB Gunfighter Ultimate w/extension, Virpil T50 CM3 Throttle, Saitek terrible pedals, RiftS

 

Link to comment
Share on other sites

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...