Overclock.net › Forums › AMD › AMD CPUs › AMD Puma+ Performance Analysis - Memory on A8-6410
New Posts  All Forums:Forum Nav:

AMD Puma+ Performance Analysis - Memory on A8-6410 - Page 3

post #21 of 29
Quote:
Originally Posted by SoloCamo View Post

For cpu + discrete gpu, yea they are similar. However being that the gpu portion of apu is already starved for bandwidth 1866 will definitely perform better especially if you play AA or stick to higher res. I went from cas11 1600mhz to cas10 1866 as seen in the first post and I believe many of these laptops ship with crappy memory specs as well unfortunately.
With Kaveri 2133 CAS 9 pretty much = 2400 CAS 10, so I assume Beema performs similarly
post #22 of 29
The graphics performance on APUs depend on bandwidth, while the latency itself is irrelevant. Bandwidth = Frequency.
post #23 of 29
That was info I got from one of the Kaveri OC threads here back when I was initially reading up on it.

So you are saying 2400 11-12-12-34 is superior to 2133 9-11-10-24 for iGPU performance?
Edited by 7850K - 6/18/16 at 7:41am
post #24 of 29
Absolutely.
post #25 of 29
Hmm........... I actually have one of these laying around in a broken screen notebook that someone dropped off and never picked back up..

Gonna have to find the PLL definition file for it. I wouldn't mind taking a crack at OC'ing this little guy through the roof. what kind of turbo is it @ 2.4Ghz? I might try shooting for 2.5Ghz on all cores, then 3 if that takes well.

Info on that GPU portion is very interesting, depending on how hot it gets when doing such loads. I heard the memory controllers on these things were junk and couldn't get past 1600 @ CAS9 without locking up. Figured that was why they were heavily modified for going in the consoles. Guess not.
Edited by Uziman04 - 6/17/16 at 11:51pm
post #26 of 29
16h APUs are using the internal Pll. The controls are defined in the public datasheet. Basically it is just a divider applied on a reference (e.g. 100 / (25 / 25)).

Pretty pointless thou since the display and SATA will usually be lost at 102 - 104MHz BCLK.
post #27 of 29
Quote:
Originally Posted by The Stilt View Post

Absolutely.
I see, thanks for the clarification
post #28 of 29
Thread Starter 
Quote:
Originally Posted by The Stilt View Post

The graphics performance on APUs depend on bandwidth, while the latency itself is irrelevant. Bandwidth = Frequency.

While I agree that bandwidth is king, latency definitely does still matter on the cpu side of it. For an apu though I'd still take 2400mhz cas20 (obviously unrealistic) over 1600mhz cas7 any day.
The Struggle (4k)
(20 items)
 
File Server
(12 items)
 
Lenovo G50-45
(6 items)
 
CPUGraphicsRAMOS
Athlon II X2 250u Nvidia 6150SE  2gb DDR3 1066mhz Windows 10 Home 64 bit 
CPUGraphicsRAMHard Drive
AMD A8-6410 AMD R5 Crucial Ballistix 8GB DDR3L 1866 CAS10 Crucial BX100 250gb 
Optical DriveOS
DVD Windows 10 Home 
  hide details  
Reply
The Struggle (4k)
(20 items)
 
File Server
(12 items)
 
Lenovo G50-45
(6 items)
 
CPUGraphicsRAMOS
Athlon II X2 250u Nvidia 6150SE  2gb DDR3 1066mhz Windows 10 Home 64 bit 
CPUGraphicsRAMHard Drive
AMD A8-6410 AMD R5 Crucial Ballistix 8GB DDR3L 1866 CAS10 Crucial BX100 250gb 
Optical DriveOS
DVD Windows 10 Home 
  hide details  
Reply
post #29 of 29
Quote:
Originally Posted by The Stilt View Post

16h APUs are using the internal Pll. The controls are defined in the public datasheet. Basically it is just a divider applied on a reference (e.g. 100 / (25 / 25)).

Pretty pointless thou since the display and SATA will usually be lost at 102 - 104MHz BCLK.

Hmm, alright. Thanks, haven't gotten a chance to open it up again to inspect the board, have to also find a power adapter for it around here. Are you talking about the PCIe clock then? If so I might have a way around that... Got quite a few PIC16F1509's lying around for doing some reference inputs. Of course all this might be moot if the board is borked anyways. Gonna have to test it, will be back when done.
Quote:
While I agree that bandwidth is king, latency definitely does still matter on the cpu side of it. For an apu though I'd still take 2400mhz cas20 (obviously unrealistic) over 1600mhz cas7 any day.

Well, simply having 2400Mhz RAM your cycle time is ~.833 ns comparing that to 1600Mhz which is ~1.250 ns, of course that's on paper. Should really be using *real* latency for RAM nowadays since CAS latencies are a poor viewing of actual performance. As frequencies increase, true latencies decrease or remains about equal, meaning better frequencies deliver better performance (Usually, however, there are some outliers)

(CAS / Frequency (MHz)) × 1000 = X ns
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › AMD Puma+ Performance Analysis - Memory on A8-6410