Overclock.net › Forums › AMD › AMD CPUs › Analysis of Kaveri's Downclocking CPU to 3.0GHz when iGPU is under full load
New Posts  All Forums:Forum Nav:

Analysis of Kaveri's Downclocking CPU to 3.0GHz when iGPU is under full load - Page 2

post #11 of 38

The physics score is only one part of the test and yet you're harping on about it as if there's nothing else. First of all you tested a GPU limited scenario, in which case you would still have the same problem if the CPU was running at 10 gigglehertz. Now you no longer have the system so it's impossible for you to test things. But before you claim a drop in CPU speed has an affect on some random synthetic test I suggest you educate yourself on what it means to be either CPU or GPU limited.

 

Edit: What's actually unwelcome is your attitude towards the help you've been given. Instead of accepting that you're wrong you are the one who is clinging to that last straw and claiming it's the end all and be all but did you actually use the iGPU of that chip to play games or did you just run synthetic tests on it then declare it a failure? If so the failure is entirely on your part.

The girlfriend.
(15 items)
 
The Mistress
(13 items)
 
Media Server
(11 items)
 
CPUMotherboardGraphicsRAM
A8-6410 Lenovo Lancer 4B2 K16.3 R5 128 Shaders/M230 Hynix 8GB DDR3 1600 
Hard DriveHard DriveOSMonitor
Samsung 840 120 GB SSD Seagate Momentus 1TB 5400rmp Win 8.1 CMN1487 TN LED 14" 1366*768 
KeyboardPowerMouseMouse Pad
Lenovo AccuType 2900mAh/41Wh Elan Trackpad/Logitech M90 Super Flower 
Audio
AMD Avalon(Connexant) 
  hide details  
Reply
The girlfriend.
(15 items)
 
The Mistress
(13 items)
 
Media Server
(11 items)
 
CPUMotherboardGraphicsRAM
A8-6410 Lenovo Lancer 4B2 K16.3 R5 128 Shaders/M230 Hynix 8GB DDR3 1600 
Hard DriveHard DriveOSMonitor
Samsung 840 120 GB SSD Seagate Momentus 1TB 5400rmp Win 8.1 CMN1487 TN LED 14" 1366*768 
KeyboardPowerMouseMouse Pad
Lenovo AccuType 2900mAh/41Wh Elan Trackpad/Logitech M90 Super Flower 
Audio
AMD Avalon(Connexant) 
  hide details  
Reply
post #12 of 38
Thread Starter 
Quote:
The physics score is only one part of the test and yet you're harping on about it as if there's nothing else.

It is the one part of the test that does not meet the criteria for testing the hypothesis. I am not going on and on about it, i left it out. Others keep pointing to it over and over as proof of their point of view and saying see your are wrong.

#1: It does not implement the behavior that I am looking to investigate, namely when the CPU shows itself to be locked down to 3.0GHz in software monitoring. The Physics test does not show this. It clearly shows the CPU is allowed to increase its clocks during the run. It's not a valid test for understanding if the hypothesis is true or false. Specifically the "When Locked at 3.0GHz" part of the claim. If a scenario shows your clocks are going higher then we know you are getting the advantage of those clocks. The claim states that you gain those clocks even wen it shows you stay at 3.0GHz.

Quote:
What's actually unwelcome is your attitude towards the help you've been given.


Its clearly unhelpful when its full of failed logic and fallacy, all of which i have pointed out. The only evidence levied at my data has 1st been the physics test scores, which i have debunk as being apples to oranges comparison. I'm only looking at apples.

2nd That my mental capacity is too small to understand how PC's work, but providing no actual data or evidence.

And 3rd, that my position is one that states my data PROVES my point of view is absolute, and no further testing or investigation is required. I actually never said this, and instead have been patiently waiting for others to provide meaningful data to the discussion. Instead they choose to sling fallacies. All of these evidence are forms of fallacies which i'm happy to provide some literature on here. My position is always show me the data. I am still waiting.

Easiest way to debunk this would be to play a game and monitor CPU clocks and FPS. If the clocks stay at 3.0GHz the whole run record FPS with CPU locked at 3.0GHz in BIOS. Then do it again with CPU clocked at 3.7 to +4.0GHz in the BIOS. If bumping the clocks higher in the BIOS net you gains even when the software monitoring says you are always at 3.0GHz then i will concede that my result is probably limited to just the synthetic test i used. There is a reason i believe that's not the case though....

Quote:
did you actually use the iGPU of that chip to play games or did you just run synthetic tests on it then declare it a failure? If so the failure is entirely on your part.


Why yes, playing games is actually why we decided to do some synthetic testing. We noticed while monitoring temps and other info in games like Evolve, Skyrim, Assassins Creed, Torchlight, and others my son plays regularly, that his CPU was down clocking to 3.0GHz during the entire run of the game. I was aware of Stilts claim because i read a fair bit of the AMD posts regularly and have since 2006 when i joined this site.

What i didn't notice was any significant impact when i bumped CPU clocks to improve performance which was odd considering what i had heard. I was mainly using the eye test and a FPS counter in steam as well as other monitoring tools like Afterburner and HWinfo, so i decided to do a more scientific test that could provide me some legit and repeatable numbers to share instead of just throwing out my anecdotal evidence. I developed a test anyone with the proper hardware and software could replicate and expand on when they had time to provide more evidence in making a decision on the future of Hypothesis #2. I did not have time to do proper in game benchmarks as those are harder to replicate, require multiple runs, and can be biased depending on the level/map/settings/ect used.

My results may indeed turn out to be limited to the synthetic, if they are i will be the first to admit it. Show me the data that provides evidence of this. My thought is that it will largely be case by case. I've made that claim in other parts of the Forum. In games that more closely resemble the stresses of GFX test 1 , 2, and the combined test; you will see no gains from CPU clocks above 3.0GHz without using the BIOS fix or software fixes. In games that more closely resemble the load of the physics test you should see measurable gains from higher CPU clocks.
post #13 of 38

As you didn't test any games you don't understand what GPU limited means in this case so it's pointless to claim that the down clocking of the CPU has an effect on gaming performance.

 

I've played games on my laptop's APU just for the fun of it. It has 128 shaders running in single channel memory at 850MHz. While I have a dGPU I have forced games to use the iGPU just for the fun of it (League of Legends is one for example). When the iGPU is fully loaded the CPU section will clock down to 2GHz or even 1.8 (from its maximum 2.4). I have, as an experiment, clocked the CPU down even to 1GHz and found that in a lot of games there is absolutely no difference (maybe 1 or 2 fps) simply because the iGPU is too weak to make full use of four cores running at 1.8 to 2GHz. When I switch to the dGPU the situation changes and I am CPU limited as the M230 is several times more powerful than the iGPU of this A8-6410.

 

That doesn't mean the CPU or the iGPU are bad, it simply means it's not a gaming chip as it's only 15W with a very weak iGPU and AMD made it clear from the start this is not a gaming chip.

 

If it's still not clear that CPU speed doesn't have any effect in GPU limited scenario's I suggest you somehow do your own testing to understand.

The girlfriend.
(15 items)
 
The Mistress
(13 items)
 
Media Server
(11 items)
 
CPUMotherboardGraphicsRAM
A8-6410 Lenovo Lancer 4B2 K16.3 R5 128 Shaders/M230 Hynix 8GB DDR3 1600 
Hard DriveHard DriveOSMonitor
Samsung 840 120 GB SSD Seagate Momentus 1TB 5400rmp Win 8.1 CMN1487 TN LED 14" 1366*768 
KeyboardPowerMouseMouse Pad
Lenovo AccuType 2900mAh/41Wh Elan Trackpad/Logitech M90 Super Flower 
Audio
AMD Avalon(Connexant) 
  hide details  
Reply
The girlfriend.
(15 items)
 
The Mistress
(13 items)
 
Media Server
(11 items)
 
CPUMotherboardGraphicsRAM
A8-6410 Lenovo Lancer 4B2 K16.3 R5 128 Shaders/M230 Hynix 8GB DDR3 1600 
Hard DriveHard DriveOSMonitor
Samsung 840 120 GB SSD Seagate Momentus 1TB 5400rmp Win 8.1 CMN1487 TN LED 14" 1366*768 
KeyboardPowerMouseMouse Pad
Lenovo AccuType 2900mAh/41Wh Elan Trackpad/Logitech M90 Super Flower 
Audio
AMD Avalon(Connexant) 
  hide details  
Reply
post #14 of 38
Quote:
Originally Posted by Liranan View Post

As you didn't test any games you don't understand what GPU limited means

Thank you! That's the entire problem.
post #15 of 38
I run my 7870K@3.5GHz (downvolted massively), all pstates fixed at that frequency, through amdmsrtweaker. GPU is stock @866MHz and RAM works at 2400MHz dual channel, dual sided. If any tests are needed I'll be happy to help.
post #16 of 38
Thread Starter 
Please do two runs of 3Dmark Skydiver, One with your CPU cores locked at 3.0GHz in bios, and one with your unlocked CPU settings at 3.5GHz.

Report your FPS in the tests mentioned above. Thanks.
post #17 of 38
Quote:
Originally Posted by gapottberg View Post

Please do two runs of 3Dmark Skydiver, One with your CPU cores locked at 3.0GHz in bios, and one with your unlocked CPU settings at 3.5GHz.

Report your FPS in the tests mentioned above. Thanks.

Hmm, is Skydiver affected by CPU speeds in general?
post #18 of 38
Thread Starter 
That is the debate. According to 3Dmark only the physics and combined tests put significant loads on the CPU.

However the physiscs test while showing gains from higher CPU clocks on Kaveri...does not push the GPU hard enough for it to maintain its self imposed p state limit.

If you run the tests and provide the data you will likely show small yo no gai. In FPS under Gfx tests 1 and 2 but perhaps a significant gain in the combined test due to being abke to maintain your higher CLU clocks through the run, instead of being limited to lower p states.

It would also be useful data if you ran a third at stock setting without your fix.

Then you would have 3 data points to compare. One with the p state limit, one locked at 3.0 and one locked at 3.5 without the limit.
post #19 of 38
*sigh*

Why are you using 3DMark Skydiver? Try something like Team Fortress 2, or . . . hell run Dirty Bomb, it is heavily CPU-dependent, but it will push the iGPU as well.
post #20 of 38
Downloading Dirty Bomb, is there an integrated benchmark available?
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
  • Analysis of Kaveri's Downclocking CPU to 3.0GHz when iGPU is under full load
Overclock.net › Forums › AMD › AMD CPUs › Analysis of Kaveri's Downclocking CPU to 3.0GHz when iGPU is under full load