A8-3850 Has Ineffective BClk Multiplier

Fail?

"Empty Overclocking" is a term we just made up, to describe unreal overclocking headroom that does not translate into any performance improvements, with AMD's A8-3850 APU. This chip has an upwards unlocked base clock (BClk) multipler. Setting it above the factory default will increase clock speed (at least the clock speed that's reported to you), but that "increased" clock speed will not translate to any performance improvements at all.

This means that the multiplier is ineffective in driving the clock above its maximum default value. So the next time you see screenshots screaming something like "6.00 GHz" on air with the base clock at its default 100 MHz, don't be fooled, trust only those overclocking feats in which the multiplier is set at the maximum default (29.0x) or lesser, and in which the overclocker has increased the base clock among other things.

It seems like AMD is aware of the issue, and forewarned reviewers about it. Apparently a glitch in the BIOS code allows the users to "set" higher multiplier values than the chip can respond to, even as the chip doesn't run at those values. Utilities like CPU-Z read those BIOS-set values and display the effective clock speed, even as the actual clock speed doesn't budge. AMD recommends only the base clock increase method for overclocking. As always, AMD warned that overclocked chips are not covered by product warranties. Perhaps future BIOS updates by motherboard vendors will fix this bug.


17

United Kingdom El Gappo says:

That's nothing new, same when people were doing fake 7ghz runs. You could set locked athlons to whatever you wanted lol

United States chew* says:

These findings are inaccurate to a certain extent.

I am working on a way to get it implemented so it works ;)

24x http://chew.ln2cooling.com/qdig-files/converted-images/Simon%20stuff/lrg_simon%20llano%202.jpg

49x http://chew.ln2cooling.com/qdig-files/converted-images/Simon%20stuff/lrg_simon%20llano%2010.jpg

49X!!!?? yes there is a way. 47X is not the limit.

United Kingdom borandi says:

Hmm that's about right, a CPU score of 15.9k at 4.9G. I can get 12.3k at 3.7G, and it's 8.3k at 2.9G.

Canada The Smith says:

Interesting. Chew, does it need any hardware modification or a simple BIOS update?

United States chew* says:

The Smith said: Interesting. Chew, does it need any hardware modification or a simple BIOS update?


Need to figure out how to add the register to bios or I need someone good with making simple programs like k10 stat.....but the person must be internally in the NDA loop.

It's in the works either way. When possible it will be public, however it's not much use for the 24/7 guys.

On avg chips topped out at 3.7 with water/air.

United States sin0822 says:

Chew is probably right as he always is, but i noticed that i can do that too increase the multiplier as much as I want but with no CPU performance increase.

I guess its possible though.

United States chew* says:

sin0822 said: Chew is probably right as he always is, but i noticed that i can do that too increase the multiplier as much as I want but with no CPU performance increase.

I guess its possible though.


As much as you want to 47X.

That is not a fix when the chips are capable of up to 5200, the fix i'm referring to will allow to well beyond what current cpu's are capable of.

United States sin0822 says:

I am not that savvy on AMD CPUs, why limited at 5.2?

United States chew* says:

sin0822 said: I am not that savvy on AMD CPUs, why limited at 5.2?


Just an average number seen when testing chips for max clocks cold.......the issue is voltage scaling is almost nonexistent, these chips scale with cold far more than voltage.

You can cram 1.7 on air/water and see no gains over 3.7.

Run on single stage and do 4.2 at stock volts.

My ln2 pot's reaction to the chips under load was rather violent considering the clocks/voltage.

Czech Republic OBR says:

chew* said: These findings are inaccurate to a certain extent. I am working on a way to get it implemented so it works ;) 24x http://chew.ln2cooling.com/qdig-files/converted-images/Simon%20stuff/lrg_simon%20llano%202.jpg 49x http://chew.ln2cooling.com/qdig-files/converted-images/Simon%20stuff/lrg_simon%20llano%2010.jpg 49X!!!?? yes there is a way. 47X is not the limit.
its irrelevant, because there is not a BCLK freq. and multi visible ...

Romania matose says:

chew* said: These findings are inaccurate to a certain extent.

I am working on a way to get it implemented so it works ;)

24x http://chew.ln2cooling.com/qdig-files/converted-images/Simon%20stuff/lrg_simon%20llano%202.jpg

49x http://chew.ln2cooling.com/qdig-files/converted-images/Simon%20stuff/lrg_simon%20llano%2010.jpg

49X!!!?? yes there is a way. 47X is not the limit.


What MB were you on? The GPU score is very low considering the 1100MHz frequency...
The GPU multiplier is also not working on boards that I used so far

United States chew* says:

OBR said: its irrelevant, because there is not a BCLK freq. and multi visible ...


Actually it's quite relevant when i post it. I actually have credibility and my name is not oftenly paired with the word salt ;)

@ matose.

Yes it very very possible that the gpu settings were not applying however for this testing the P score was not the target. It's also quite possible that the score is low due to the memory and the ht/nb not being pushed.

Ht had quite an impact on prior AMD IGP platforms.......

How far the cpu's can go was. It was the A75 UD4 something.

United Kingdom El Gappo says:

I was going to say "You are irrelevant." But that works as well. :)

Netherlands rsnubje says:

you should update your cpu-z chew ;)

United States chew* says:

rsnubje said: you should update your cpu-z chew ;)


I can and will but you need to realize when the testing was done that was what was available ;)

Indonesia xqrfjr says:

AMD need some power

Belgium HybridChiller says:

AMD fail, Intel all the way

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)