RBE v1.25 open beta test!

[BAGZZlash]

Hi everyone!

I (hopefully) just completed coding RBE v1.25. It contains fan support for the EMC2103 fan controller used on the Radeon 5970.
I suppose there might be not so many people out there using a 5970 by now, but if anyone who does would like to test the new fan controlling capabilities I'd be very grateful. Download your beta version here. If you test, please be sure to report back here and tell as exactly as possible what you did and what the outcome was.

Kind regards and thanks in advance

BAGZZlash​

 

[sparksltd]

Hey, I know you said this is updated for the 5970 but I tried on my 5870 for the sake of it and it stopped fan control via software like msi afterburner working.
Sorry if this is no use to you.

 

[Sparr]

-

 

[BAGZZlash]

Nobody? Seriously?

 

[Diverge]

If no one can test it out in the next couple days, I'll re-attach the fan to my 5970. Since it's watercooled, I'll scale the fan rpms based on max temp it reaches on water just for testing. So i'll make it reach 5016rmps at like 45C instead of the normal 95C.

Edit: hopefully the fans connector will be long enough to plugin without mounting the heatsink or I can't test it.

 

[damric]

Hi, I was wondering if you can do anything with the Sapphire Vapor-x 5750. I would love to get the voltage up.

 

[ml2517]

Nobody? Seriously?

I used the beta to program my 5970's successfully, unfortunately they are in a liquid cooled system so I can't help with the fan portion.

 

[BAGZZlash]

Hi, I was wondering if you can do anything with the Sapphire Vapor-x 5750. I would love to get the voltage up.

Revision history

v1.24
[...]
Unfortunately, 5670 and 5750 reference design cards can not be supported because their UP6201 chip does not support external programming.

If no one can test it out in the next couple days, I'll re-attach the fan to my 5970. Since it's watercooled, I'll scale the fan rpms based on max temp it reaches on water just for testing. So i'll make it reach 5016rmps at like 45C instead of the normal 95C.

Edit: hopefully the fans connector will be long enough to plugin without mounting the heatsink or I can't test it.

I'd love that!

 

[xFiniti64]

Please fix fan support for 5850.....

 

[Clement Chong]

I have a 5970.
What is the minimum and maximum RPM that I can input into the boxes (since it is not in percent).
And how is the RPM calculated from the BIOS (since non of the values listed in my BIOS is rounded nicely like XX00).

 

[Clement Chong]

Is there any reason why I cannot set a value that is between 4468-4637?
Any number higher than 4468 is rounded straight up to 4637.
I am trying to set it at 90% of 5016 which is 4514.

 

[BAGZZlash]

What is the minimum and maximum RPM that I can input into the boxes (since it is not in percent).

You can find out by hovering over the graph: Min. is 0 RPM, max. is 7,000 RPM.

And how is the RPM calculated from the BIOS (since non of the values listed in my BIOS is rounded nicely like XX00).

There is a look up table: Each byte means a certain RPM level.

Is there any reason why I cannot set a value that is between 4468-4637?
Any number higher than 4468 is rounded straight up to 4637.

The reason is this look up table. For each value the user enters an internal function determins the closest legal value. RBE will use this one, then.

Did you already flash a fan-modified BIOS? What are your results?

 

[Clement Chong]

You can find out by hovering over the graph: Min. is 0 RPM, max. is 7,000 RPM.

There is a look up table: Each byte means a certain RPM level.

The reason is this look up table. For each value the user enters an internal function determins the closest legal value. RBE will use this one, then.

Did you already flash a fan-modified BIOS? What are your results?

That's strange, ATI CCC deemed 2508 RPM to be 38% fan speed, going by that 100% should be around 6600 RPM.

Is the look up table linear? It seems for high RPM the values rounded off seems to be quite big (close to 200 RPM difference) compared to low RPM (around 30 RPM difference).

I modded the fan settings to this:
0-2643
65-2997
70-3321
75-3668
80-3964
85-4312
90-4637
95-5343
When the GPU is hovering at 85-86, the GPU-Z readings gave me only 4230 RPM, while the rest of the settings lower than 85C are correct according to what I set.

 

[BAGZZlash]

That's strange, ATI CCC deemed 2508 RPM to be 38% fan speed, going by that 100% should be around 6600 RPM.

I didn't say that 7,000 RPM is 100%. In fact, I don't know what 100% is. I just selected the control range within reasonable values. Good to hear that 7,000 RPM max should do the trick since 6,600 is 100%.

Is the look up table linear?

No, it isn't.

When the GPU is hovering at 85-86, the GPU-Z readings gave me only 4230 RPM, while the rest of the settings lower than 85C are correct according to what I set.

Seems okay. Readings might not be accurate all the time or the fan doesn't match the given value right all the time.

Thank you for this!

I would still like to wait for Diverge's results prior to releasing the final v1.25.

 

[Diverge]

Okay, seeing as I promised to do this today, i found my stock heatsink and was able to connect the fan connector up to my watercooled 5970



Stock lookup table: Stock @ idle: Stock @ load:

Modded lookup table: Modded @ idle: Moded @ load:

I thought my card would get to 45C under load..but it only got to 41-42C. But as you can see, it works It just takes a little fine tuning on the users part to come up with the lookup table they want to use.

edit: since my temp doesn't change much from idle to load, and doesn't show much fan changes, i decided to make one more lookup table.

New lookup table: Idle: Load: More load: Cooling:

It looks like the 5970 max fan speed is around 4500rpms, and looking at Clement Chong's data, my max speeds correlate with what he got. I think it's not too important if the % or the actual rpms's are exact.... it's more important that you can make your own fan curve, and i'm sure everything is close enough

 

[Clement Chong]

Seems okay. Readings might not be accurate all the time or the fan doesn't match the given value right all the time.

I don't believe readings should completely miss (as in 100%) the given value set, I checked with the log (the log is attached in my previous post), the set values fall in the range of the readings except for the 85C one, see below (GPU-Z readings is in the brackets):
65-2997 (2782-3141)
70-3321 (3187-3516)
75-3668 (3536-3878)
80-3964 (3893-4239)
85-4312 (4208-4253)

For the 85C section, not a single GPU-Z reading taken has ever reach 4300.
I am very sure the fan can achieve more than 4300 RPM cause I tested by manually setting the Fan Speed to 100% via Overdrive and the speed is around 4854 RPM (average of 30 readings).
30% - 1810
40% - 2600 (+790)
50% - 3209 (+609)
60% - 3671 (+462)
70% - 4065 (+394)
80% - 4375 (+310)
90% - 4635 (+260)
100% - 4854 (+219)

I am suggesting to take another look at the look up table and see whether there is any mistake in values that are above 4075 RPM (since that is the last value I have tested to be working correctly).

 

[BAGZZlash]

Okay, seeing as I promised to do this today, i found my stock heatsink and was able to connect the fan connector up to my watercooled 5970

[...]

It looks like the 5970 max fan speed is around 4500rpms, and looking at Clement Chong's data, my max speeds correlate with what he got. I think it's not too important if the % or the actual rpms's are exact.... it's more important that you can make your own fan curve, and i'm sure everything is close enough

I think so, too. And thank you again very much for your sophisticated testings!

I don't believe readings should completely miss (as in 100%) the given value set, I checked with the log (the log is attached in my previous post), the set values fall in the range of the readings except for the 85C one, see below (GPU-Z readings is in the brackets):

[...]

I am suggesting to take another look at the look up table and see whether there is any mistake in values that are above 4075 RPM (since that is the last value I have tested to be working correctly).

I checked the look up table again - perfect match.
Don't be too strict on this. You have to take into account the "ramp" effect, the "hysteresis" effect, the fact that entered RPMs are translated into certain "legal" values, inaccuratenesses during PWM programming and its implementation by the fan and inaccuratenesses during fan speed reading.
Thank you again for testing, RBE will go into release soon.