Overclock.net banner
101 - 120 of 158 Posts

·
Registered
Joined
·
1,765 Posts
I want to test but it takes too much time in the end for all variables.

Might test 1333-1600FCLK and see if I can maximize CO to -20 & -30 something for the cores later today.

I don't think we need to go too low on FLCK before we maximize the available settings for CO.
Would be good to find a "floor" for it as we already were touching the ceiling for it.
 
  • Rep+
Reactions: TimeDrapery

·
Registered
Joined
·
1,765 Posts
Hmm... Only using 1600FCLK. 1/2 UCLK and 4266MEM.

All-core -15 CO is still not working for me. Get blue screen on Windows loading.
Indicating cores can't mange. Didn't help as much as I thought to lower FCLK, everything else similar.
 
  • Rep+
Reactions: TimeDrapery

·
Registered
Joined
·
804 Posts
Didn't help as much as I thought to lower FCLK, everything else similar.
I'd say, not a big surprise.
These RAM-guys are discussing about CO degradtion from 3200Mhz to 3800Mhz RAM speed. This discussion is about X-series, but something similar should apply to our AUPs too, probably in a bit higher RAM/FCLK-speeds
Come on...
It shouldn't be taken seriously, MC speed (FCLK) and CCX's are not just in different voltage domains, but even in physically different chips.
 
  • Rep+
Reactions: TimeDrapery

·
Registered
Joined
·
1,765 Posts
Really?...
Yeah...

Core 0 can't do -10 CO @ 1600FCLK after testing individual cores.
Didn't scale with lower FCLK that much as I thought it would. Barely did anything.
Going 1333FCLK didn't help either to have Core 0 be stable with -10 CO.

Gonna test more but it looks bad. Doesn't allow lower CO as much as I thought it would on my sample.

So doesn't scale below 2000FCLK? Or doesn't improve results thus far? Bummer. Would have loved to have it do -20 or something but it doesn't even do -10.
 
  • Rep+
Reactions: TimeDrapery

·
Registered
Joined
·
804 Posts
So doesn't scale below 2000FCLK?
I'd rather say it shouldn't scale with fclk decreased, but certainly may scale down with FCLK going upwards.
For example, too much stress for the L3 (since it twice as low) as possible reason, or some power-gating come into effect
 

·
Registered
Joined
·
1,332 Posts
Discussion Starter · #106 ·
Yeah...

Core 0 can't do -10 CO @ 1600FCLK after testing individual cores.
Didn't scale with lower FCLK that much as I thought it would. Barely did anything.
Going 1333FCLK didn't help either to have Core 0 be stable with -10 CO.

Gonna test more but it looks bad. Doesn't allow lower CO as much as I thought it would on my sample.

So doesn't scale below 2000FCLK? Or doesn't improve results thus far? Bummer. Would have loved to have it do -20 or something but it doesn't even do -10.
the 5600g and 5700g tops out at 4650 and 4850 and limited tdp.

i think you just need to set +200 autoOC and give them a 2x-4x scalar to get that peaks. even with poor CO offsets
 

·
Registered
Joined
·
36 Posts
Hmm... Only using 1600FCLK. 1/2 UCLK and 4266MEM.

All-core -15 CO is still not working for me. Get blue screen on Windows loading.
Indicating cores can't mange. Didn't help as much as I thought to lower FCLK, everything else similar.
Interesting. I noised that my worst cores gets more effected of high FCLK/MEM speed. So, you could try one of your best cores and see if it makes a difference. The problem is that it takes too much time to eliminate these kind of issues per core.

But maybe it's not the FCLK? maybe it's other Uncore-issues or something else? We have to eliminate them one by one, become this CO is a magic-box and nobody knows what's exactly happening in there.
 

·
Registered
Joined
·
36 Posts
Come on...
It shouldn't be taken seriously, MC speed (FCLK) and CCX's are not just in different voltage domains, but even in physically different chips.
These APUs are all-in-one laptop-chips that AMD has redesigned for desktops. So, what other "physically different chips " are you talking about?
 

·
Registered
Joined
·
36 Posts
the 5600g and 5700g tops out at 4650 and 4850 and limited tdp.

i think you just need to set +200 autoOC and give them a 2x-4x scalar to get that peaks. even with poor CO offsets
That's a good point. This is the way to get the highest speed.

But a good CO would reduce the voltage/heat/noise, and it may even eliminate some limitings factors too.. So, a good CO is absolutely desirable. It's worth investigating the optimization of CO on these relatively new and "unexplored" APUs.
 

·
Registered
Joined
·
804 Posts

·
Registered
Joined
·
36 Posts
You were.
Please don't cut my lines in half to score a cheap point. I said "This discussion is about X-series, but something similar should apply to our AUPs too, probably in a bit higher RAM/FCLK-speeds". Besides you know that we are talking about APUs here.

These APUs have no other "physically different chips". So, what are you talking about?
 

·
Registered
Joined
·
804 Posts
Are you fkn dumb or you prefer a wordplay instead of tech discussion?
I clearly pointed out that my answer was in context of your reference to X model discussion and hence was related to it. What else you can't understand?
You'd better learn to communicate before posting in a tech forum.
 

·
Registered
Joined
·
36 Posts
Are you fkn dumb or you prefer a wordplay instead of tech discussion?
I clearly pointed out that my answer was in context of your reference to X model discussion and hence was related to it. What else you can't understand?
You'd better learn to communicate before posting in a tech forum.
This is not a civilized way to discuss technical disputes . I have reported your post and leave it to moderators.
 
  • Rep+
Reactions: TimeDrapery

·
Registered
Joined
·
36 Posts
Work in progress ...

5700G @ stock, with CO per core.
Gigabyte B550 Vision D, BIOS F14C.
G.Skill TridentZ Neo DDR4 RGB 3600Mhz CL16-16-16-36 2x8GB, @3800 CL16-16-16-36 1.42v(BIOS sett), FCLK @1900.
Dropped the FCLK to 1900 for 4 days ago, and no idle/light-load reeboot yet.
Precision Boost Overdrive is sett to disabled, for now. No overclocking yet.

First OCCT benchmarks.It seams to be beating everything out there in single-thread bunching:love:

Font Screenshot Technology Electronic device Software



Font Screenshot Slope Software Technology



Font Line Screenshot Slope Technology
 
  • Rep+
Reactions: TimeDrapery

·
Registered
Joined
·
36 Posts
Stability test, for the same settings as above.
5700G with CO per core. Negative -24, -24. -12, -24, -14, -22, -12. -22, for Now
Precision Boost Overdrive is sett to disabled, for now. The CPU was keeping about 55'c under the whole stability test.

Gigabyte B550 Vision D, BIOS F14c.
G.Skill TridentZ Neo DDR4 RGB 3600Mhz CL16-16-16-36 2x8GB, @3800 CL16-16-16-36 1.42v(BIOS sett)

First OCCT stability test for CPU has passed, for 1 hour. ;)

Slope Font Screenshot Rectangle Software
 

·
Registered
Joined
·
1 Posts
Font Screenshot Technology Software Electric blue


Would I be completely nuts for dailying this? I'm using a 5600G, and this overclock is tm5 and occt stable at 1.5v VDIMM. I'm fairly new to overclocking (i've been tinkering with my ram settings for about a month), and I can't seem to find any concrete information on safe vSoc voltages for the G series. Articles like these ones linked below have used 1.35v - 1.37v to stabilise iGPU overclocks, and i've seen earlier replies about vSoc up to 1.6v (not suggesting this is safe, just that there are examples) to stabilise the infinity fabric clock. But at the same time, the prevailing information, as I understand it, seems to be 1.2v limit for earlier 5000 series. My Patriot B-die requires 1.33v vSoc at 4400mhz and i'm worried about damaging the IMC over time.

 

·
Registered
Joined
·
1,765 Posts
Would I be completely nuts for dailying this? I'm using a 5600G, and this overclock is tm5 and occt stable at 1.5v VDIMM. I'm fairly new to overclocking (i've been tinkering with my ram settings for about a month), and I can't seem to find any concrete information on safe vSoc voltages for the G series. Articles like these ones linked below have used 1.35v - 1.37v to stabilise iGPU overclocks, and i've seen earlier replies about vSoc up to 1.6v (not suggesting this is safe, just that there are examples) to stabilise the infinity fabric clock. But at the same time, the prevailing information, as I understand it, seems to be 1.2v limit for earlier 5000 series. My Patriot B-die requires 1.33v vSoc at 4400mhz and i'm worried about damaging the IMC over time.
Should be safe.
I used up-to 1.450V for SoC when I pushed it for a while running 2366FCLK though 1.400V was only really necessary.
iGPU voltage is derived and is a offset from vSoC. Doesn't need to be that high. Actaully you don't want the iGPU to guzzle voltage for unnecessary wattage if it for some reason defaults to vSoC which I found it out to do at times. (check it's not using more than ~1.150V @ stock)
It should remain @ stock even if you fiddle with the vSoC voltage.
iGPU doesn't need that much voltage at stock speeds even if you OC MEM & FCLK.

It's mainly FCLK that requires the vSoC voltages, depending if you have your iGPU enabled it can shift ~0.050V more or less for stability requirements.

CLDO_VDDP seems more than necessary though. I need only 800-900mv for 4200-4266Mts speeds.
 

·
Registered
Joined
·
26 Posts
Should be safe.
I used up-to 1.450V for SoC when I pushed it for a while running 2366FCLK though 1.400V was only really necessary.
iGPU voltage is derived and is a offset from vSoC. Doesn't need to be that high. Actaully you don't want the iGPU to guzzle voltage for unnecessary wattage if it for some reason defaults to vSoC which I found it out to do at times. (check it's not using more than ~1.150V @ stock)
It should remain @ stock even if you fiddle with the vSoC voltage.
iGPU doesn't need that much voltage at stock speeds even if you OC MEM & FCLK.

It's mainly FCLK that requires the vSoC voltages, depending if you have your iGPU enabled it can shift ~0.050V more or less for stability requirements.

CLDO_VDDP seems more than necessary though. I need only 800-900mv for 4200-4266Mts speeds.
I noticed the GFX Voltage follows whatever the SOC Voltage is on my Gigabyte motherboard. That was not the case on my ASRock board.
 

·
Registered
Joined
·
804 Posts

·
Registered
Joined
·
26 Posts
Exactly so on mine.
On stock settings, the GB board switches between SOC and GFX voltages accordingly. However, the moment I set the SOC manually, the GFX voltage just follows SOC. Weird.
 
101 - 120 of 158 Posts
Top