Tweet
Switch Hitting Athlons/Abit K7TA Follow-up
I just want to confirm what others have been mentioning for the last couple of weeks. Many of the 1.2 GHz Athlons floating around have their L1 bridges intact. No need to have surgeon-like steady hands or the eyes of a hawk to unlock them.
My Abit KT7A genuinely didn't care which setting I used and seemed to consider both settings "spec." You don't even have to enable the "User Define" feature to manipulate the multiplier and bus frequency. There are two 1200 MHz settings and you can toggle back and forth from one to the other.
It's important to remember to change the DRAM Clock from HCLK+PCICLK back to the default Host CLK when switching CPU frequencies from 100 to 133 MHz. Most PC133 RAM will object to running at 166 MHz.
There are modest performance gains associated with the CPU running at 9x133 versus 12x100 with 133 MHz memory bus:
Content Creation Winstone 2001: 41.8->42.9
Business Winstone 2001: 45->46.4
Sysmark Rating: 218->223
3DMark 2000: 8866/648->9011/661
Q3Demo001/800x600/16-bit: 145.2->153.3
Sandra CPU Dhrystone/ALU: 3342->3360
Sandra CPU Whetstone/FPU: 1649->1642
Sandra ALU/RAM Bandwidth: 514->536
Sandra FPU/RAM Bandwidth: 587->580
Evolva Benchmark: 128.5->128.7
The KT7A has even more BIOS tweaks than my familiar and now retired KA7-100. I used the following:
Fast CPU Command Decode: Fast
CPU Drive Strength: 2 (Default)
Bank X/X DRAM Timing: Turbo
DRAM Bank Interleave: 4-way
MD Driving Strength: Hi
SDRAM Cycle Length: 2
For those of you not familiar with Abit's extensive menu of BIOS tweaks, these are fairly aggressive settings. Some of them will gleefully destabilize your system if you're not using high-grade CAS2 PC133 RAM. I messed around with most of these settings, and they can have a sizable impact on benchmark results.
Duron performance on the KT7A was quite good. At 800 MHz, the Duron held its own against a 750 MHz Thunderbird on a Abit KA7-100. Unlike the TBird on the KA-133 chipset-based board, I could overclock the Duron to 1008 MHz, and it put on quite a show.
The Evolva Benchmark appears to be a poor indicator of performance with this chipset. My PIII 933 MHz on an Asus CUSL2-C produces better Evolva numbers than the K7-1200/Abit KT7A, although the Athlon rig decisively cleans house in every other test. Possibly, Evolva isn't optimized well for 3DNow.
I used a Taisol heatsink with a 30 CFM Delta fan. The fan isn't the power hog (or as noisy) as Delta's 38 CFM model. It worked well plugged into the board with the Duron 800, but my system did a swan dive into an empty pool when I replaced the Duron with the Athlon.
I did this intentionally and Delta fan users should take note. It can be an serious issue, as the KT7A's shipping BIOS will not permit the board to boot without a CPU fan plugged into the motherboard (FAN1). The WZ_01 beta BIOS disables this feature by default and provides and option to re-enable it if you like. Of course, the system has to boot in order to flash the BIOS.
AMD Athlon 1.2 GHz
AMD Duron 800 MHz
Abit KT7A (WZ_01 Beta BIOS)
128 MB Crucial PC133 "7E" RAM
Soundblaster Live 1024
IBM ATA-100 15 GB
VIA 4in1 4.25a(1)
DirectX 7a
Paul
paulcs@flashcom.net
My Abit KT7A genuinely didn't care which setting I used and seemed to consider both settings "spec." You don't even have to enable the "User Define" feature to manipulate the multiplier and bus frequency. There are two 1200 MHz settings and you can toggle back and forth from one to the other.
It's important to remember to change the DRAM Clock from HCLK+PCICLK back to the default Host CLK when switching CPU frequencies from 100 to 133 MHz. Most PC133 RAM will object to running at 166 MHz.
There are modest performance gains associated with the CPU running at 9x133 versus 12x100 with 133 MHz memory bus:
Content Creation Winstone 2001: 41.8->42.9
Business Winstone 2001: 45->46.4
Sysmark Rating: 218->223
3DMark 2000: 8866/648->9011/661
Q3Demo001/800x600/16-bit: 145.2->153.3
Sandra CPU Dhrystone/ALU: 3342->3360
Sandra CPU Whetstone/FPU: 1649->1642
Sandra ALU/RAM Bandwidth: 514->536
Sandra FPU/RAM Bandwidth: 587->580
Evolva Benchmark: 128.5->128.7
The KT7A has even more BIOS tweaks than my familiar and now retired KA7-100. I used the following:
Fast CPU Command Decode: Fast
CPU Drive Strength: 2 (Default)
Bank X/X DRAM Timing: Turbo
DRAM Bank Interleave: 4-way
MD Driving Strength: Hi
SDRAM Cycle Length: 2
For those of you not familiar with Abit's extensive menu of BIOS tweaks, these are fairly aggressive settings. Some of them will gleefully destabilize your system if you're not using high-grade CAS2 PC133 RAM. I messed around with most of these settings, and they can have a sizable impact on benchmark results.
Duron performance on the KT7A was quite good. At 800 MHz, the Duron held its own against a 750 MHz Thunderbird on a Abit KA7-100. Unlike the TBird on the KA-133 chipset-based board, I could overclock the Duron to 1008 MHz, and it put on quite a show.
The Evolva Benchmark appears to be a poor indicator of performance with this chipset. My PIII 933 MHz on an Asus CUSL2-C produces better Evolva numbers than the K7-1200/Abit KT7A, although the Athlon rig decisively cleans house in every other test. Possibly, Evolva isn't optimized well for 3DNow.
I used a Taisol heatsink with a 30 CFM Delta fan. The fan isn't the power hog (or as noisy) as Delta's 38 CFM model. It worked well plugged into the board with the Duron 800, but my system did a swan dive into an empty pool when I replaced the Duron with the Athlon.
I did this intentionally and Delta fan users should take note. It can be an serious issue, as the KT7A's shipping BIOS will not permit the board to boot without a CPU fan plugged into the motherboard (FAN1). The WZ_01 beta BIOS disables this feature by default and provides and option to re-enable it if you like. Of course, the system has to boot in order to flash the BIOS.
AMD Athlon 1.2 GHz
AMD Duron 800 MHz
Abit KT7A (WZ_01 Beta BIOS)
128 MB Crucial PC133 "7E" RAM
Soundblaster Live 1024
IBM ATA-100 15 GB
VIA 4in1 4.25a(1)
DirectX 7a
Paul
paulcs@flashcom.net
Comment