Sapphire Radeon 9600XT 128Mb AGP: Advanced Overclocking

The graphprocessor on this card is fed with 1.3V power supply. The stabiliser is simple fairly as there are two International Rectifier 7413 (IRF7413 – PDF datasheet, 224Kb) N-channel field-effect transistors driven by a Chip Advanced Tech 7522 (CAT7522 – PDF datasheet, 134Kb) PWM controller located on the back side of the card. By the way, these IRF7413 are fine transistors capable of max. 11mΩ at Vds=10V and Id=7.3A.

Actually, the controller is a CAT7522G. Nothing really serious, a Pb-free version of a regular CAT7522. The stabiliser's input voltage is +12V smoothed by one TEAPO 470µF/16V liquid electrolytic capacitor. Output voltage is passed through a 3.3µH choke and smoothed by two liquid electrolytic capacitors of low impedance, a TEAPO 1000µF/6.3V and a Nichicon HC 470µF/10V. Not sure about the TEAPO one, but this Nichicon HC is good enough to deliver max. 30mΩ impedance at 20°C and 1230mArms peak current at 105°C given 100KHz operating frequency in both cases. There is no reason to admit that this unlabelled Taiwanese one by TEAPO is any better than its Japanese colleague by Nichicon. Most likely, it's less expensive and worse considerably. In general, it's supposed to be as good as of OST RLP series, so here come figures of max. 78mΩ impedance and 700mArms peak current. By the way, the stabiliser operates at about 250kHz (240kHz calculated accordingly to a 51kΩ resistor hooked up to the controller's RT pin, 260kHz measured actually). No doubt that such a pair of capacitors would do the job just fine in case of linear stabiliser, but pulsed-width designs impose much higher demands on output capacitors. That's difficult to measure how much current this RV360 consumes under heavy load, but supposed to be about 6A. If we want to keep voltage delta at +/- 5% (+/- 65mV), then 130 ÷ 6 = 22mΩ total output capacitors' impedance requirement must be met. Our pair can do exactly 22mΩ at 100kHz, but don't forget to add about 30% to account for increased losses on reactive inductance at 250kHz. In a matter of fact, these two capacitors don't fit the design really well. Peak current is another question which affects capacitors' lifespan. Of course, this card will be retired probably before they blow up or dry out, but who knows. However, with increased voltage and current due to upcoming advanced overclocking they may fail much sooner. Anyway, the author has replaced both of them with a pair of excellent liquid electrolytic Rubycon MBZ 2200µF/6.3V capacitors. Each of them delivers max. 13mΩ impedance and 2550mArms peak current, so they shouldn't die until the next century arrives. Just a joke, but you should have got the idea. In addition, they offer 3 times more capacitance combined than the previous pair, so the graphprocessor must feel really happy with this upgrade. Just in case, a 470µF/16V capacitor on the input has been replaced with a 1000µF/16V one of the same manufacturer. There are also two 22µF tantalum capacitors (C323 and C324) installed near the primary output electrolytes and a 10µF multilayer ceramic capacitor (MLCC) situated behind the graphprocessor on the back side of the card (C24). All of them are intended to reduce high-frequency voltage losses, and that's good. Another 10µF MLCC wouldn't hurt for sure, so it has been soldered on the top of the existing one. Having finished with the major components, let's get straight to voltmodding of this RV360.

CAT7522 uses a resistor divider to set output voltage, and there is a FB pin for this purpose. The resistors are located on the other side of the card near the upper mounting hole.

R356 of 3.0kΩ is a so-called lower resistor because it's placed between earth and the middle point, and R353 of 1.9kΩ is a so-called upper resistor because it's located beteween output voltage and the middle point. CAT7522 tries to maintain 0.8V in the middle point. Hence, output voltage is set at (1.9 ÷ 3.0) × 0.8 + 0.8 = 1.31V. You have to decide how much voltage you want and install the resistors accordingly. For instance, any two identical resistors from 1kΩ to 5kΩ range would give you 1.6V. A 2.7kΩ upper resitor with the existing 3.0kΩ lower resistor would deliver 1.52V. A 2.2kΩ upper resistor and a 2.7kΩ lower one — 1.45V. All power to you, don't be afraid of experimenting, just do your soldering job carefully.