Review: PowerLogix PowerForce G3 250 in a PowerBase 180
PowerLogix PowerForce G3 250
in a PowerBase 180
Final Review
Pros: Incredible speed, easy to install and configure, great flexibility, price for high-end systems
Cons: Price for low-end systems, Be OS incompatible
Rating: None higher possible
Price: $2,199 (~$1,900 street)
November 20, 1997 (rev.1, 11/24/97)
On Friday (11/14), I installed a PowerLogix PowerForce G3 250 CPU daughtercard (PF) in my PowerComputing PowerBase 180. The PF replaced the original 603e/180 MHz daughter card, at least for the time being. I also removed the existing 512K motherboard L2 cache per the instructions. PowerLogix agreed to loan me this card to do some real world, hands-on and independent testing of one of their new PowerPC 750 (G3) cards in a heavily used but cheap, low-end PCI clone. Just, call me "Mr. Lucky" or maybe "Mr. Persistent." You see, I really didn't believe what I thought was "hype" of these CPU's and their cards, especially for low-end systems like the PowerBase. The MacBench-marks of high-end systems that were to use the 750 CPU, like PowerComputing's PowerTower Pro and Motorola's 6000 computers, looked great so I wasn't really too skeptical of the card's performance claims for high-end systems with a better motherboard and a faster system bus. A price of $5K more or less for those computers, however, would have been way out of my price ballpark anyway. My skepticism did extend to the difference between MacBench and real world use performance and that there is not a linear increase in performance with CPU speed. Thankfully, I was wrong on both counts--the G3 is something else and it makes a quantum jump in performance.
Installation and Setup Notes (11/17)
Installation and setup was a snap. I had absolutely no problems. The PowerBase case's cover is removed by unscrewing three knurled bolts in the back, sliding the cover back a little over a half inch and lifting it up from the front and off. The CPU daughtercard (the one with the big aluminum heatsink to the left of the fan when looking from the front) is held in place by two plastic clips that connect to the metal brace running front-to-back over the middle of the motherboard. The clips are removed by lightly squeezing them together and pulling them upward. Using a slight front-to-back rocking action firmly pull upward and work the daughtercard out of its slot. [Note: The daughtercard was more difficult to remove than I had expected, given the many times that I have pulled and inserted all kinds of cards in computers.] The instructions indicate that any L2 cache card already installed that is less than the size of the backside cache should be removed--any L2 cache (which becomes L3 when the PowerForce is in) equal to or larger than the backside cache can stay if that is physically possible but the instructions note that it is not likely to improve performance perceptibly. I have a 512K L2 cache card so I removed it. With the PowerForce in, there doesn't look like there is enough room for the L2 (L3) to stay anyway.
The PowerForce must be carefully aligned properly in the slot and then inserted with a slight rocking motion while pushing down firmly. Make sure that it goes all the way in. My first attempt, perhaps tempered by the fact that I was inserting a $2,000 item, resulted in it not seating all the way and not being able to boot the computer. A bit more rocking pressure seated it and my PowerBase booted when powered on. The bus speed came set at 50 MHz (switch A, position 4) and a 5:1 ratio (switch B, position 4) yielding a CPU speed of 250 MHz. The switches on the card that I am testing are more conveniently located (i.e. on the top or upper edge and facing up) than those shown in the documentation (on the side and facing to the left side).
Before closing the case, make sure that the computer boots. While the computer is on, copy the control panel for the backside cache from the diskette that comes with the PF card into your system folder's control panel folder and reboot the computer. Open the control panel called "G3 Cache Control v1.1.1" and the backside cache can be set a 1MB, 512K and 256K. I set it at 1 MB. The panel also allows the ratio of CPU to cache speed to be set from 1:1 to 3:1. I set it at the fastest, i.e. 1:1.
Next comes the optimization (or acceleration) steps so shut down the computer. The computer must be off when adjusting the PF card's switches. This process can be rather time consuming but it can ensure that the computer operates at maximum performance. The manual describes the general steps so I won't repeat them. Here's what I did:
- I set the bus speed to 40 MHz (B=0) which is the rated speed for a PowerBase motherboard and the ratio so that the CPU ran at 320 MHz (A=A), the highest on the enclosed chart and tried to start--no go. I cranked the CPU back 2 steps to 280 MHz (A=8) the computer booted and seem to run fine so I shut down and moved up a notch (A=9) to 300 MHz and again it booted and seemed to run fine (Note: Later I discovered an occasional freeze occurred at this CPU setting so it is wise to use the next lower setting from the highest that appears to work during optimization.). Result = 40/280
- Next I bumped the bus up to 45 MHz (B=1) and the CPU to 315 (A=8)--freeze on startup. Down a CPU notch to 292.4 MHz (A=7) and the PowerBase booted and worked fine. Result = 45/292.4
- Up another notch on the bus speed to 47.5 MHz (B=2) and the CPU goes to 308.8 MHz at A=7 -- another starup freeze so I cranked the CPU down a notch to 285 MHz (A=6) and it worked. It was clear that my system and/or the card were not going to run the CPU above 300 MHz (Note: Again, I discovered later that 300MHz produced an occasional freeze at any bus speed). Result = 47.5/285
- Then I set the bus speed to 48.36 MHz (B=3) and left the A-switch at 6 (CPU=290.2 MHz) and it worked. Cranking the CPU up a notch to 314.3 (A=7), I got a startup freeze. Result = 48.36/290.2
- The bus speed at 50 MHz (B=4) and CPU at 300 MHz (A=6) ran but not in later tests without an occasional freeze. Result = 50/275
- A bus speed of 52.5 MHz or 55 MHz produced startup freezes down to at least A-switch = 2 (210 MHz) so I gave up on these bus speeds.
This whole process could be repeated with different backside cache settings. The manual states that by increasing the CPU/cache ratio from 1:1 to 1.5:1 or 2:1 (i.e. slowing down the cache relative to the CPU), the CPU speed might be able to be increased some. I will do this at a later date to document what happens in a PowerBase.
I decided to use the settings A=7 and B=1 (CPU=292.4 and bus=45 MHz) on advice that with a 1:1 backside cache the CPU speed was more important than bus speed. However, the close alternative is A=6, B=3 or 290.2/48.36 MHz, a 0.75% decrease in CPU speed for a 7.5% increase in bus speed, so it could be a little better [Note: According to MacBench run on 11/20 it is indeed slightly better. However, other tests {see Accelerate Your Mac} indicate that video or other hardware may be compromised by bus speeds higher than 45 MHz so I recommend the safer 45/292.4 combination].
All-Out Performance Results (11/15, 11/17 & 11/20)
The PF has a 250 MHz PowerPC 750 CPU that derived from the 603e line of PowerPC chips. It has 1 MB of the new "backside" cache on the CPU card itself and it allows one to adjust independently (1) the ratio between the CPU speed and the system bus speed, (2) the actual bus speed and (3) the ratio between the backside cache and CPU speed to achieve the best performance from one's system. I achieved a CPU speed of 300 MHz at 50 and 40 MHz bus speeds but it proved to not be completely stable after a couple of freezes working at both bus speeds. Table 1 lists the highest CPU speeds with a 1:1 backside cache ratio for each bus speed that was completely stable:
Table 1. -- PowerLogix PowerForce G3 250 MHz Daughtercard Performance*
System Bus Speed (MHz) |
Maximum CPU Speed (MHz) |
40.00 |
280.0 |
45.00 |
292.4 |
47.50 |
285.0 |
48.36 |
290.2 |
50.00 |
275.0 |
Table 2 lists the MacUser MacBench 4.0 test results for the best (highest CPU) settings resulting in absolute stability:
CPU/Bus Ratio |
Bus Speed (MHz) |
CPU Speed (MHz) |
Backside Cache/CPU Ratio (Speed) |
Processor Score |
FPU Score |
Disk Score |
6.5:1 |
44.98 |
292.4 |
1:1 (292.4) |
1,061 |
921 |
273 |
6.0:1 |
48.36 |
290.2 |
1:1 (290.2) |
1,083 |
928 |
282 |
5.5:1 |
50.00 |
275.0 |
1:1 (275.0) |
1,023 |
874 |
280 |
These results are phenomenal! A MacBench 4.0 CPU score of over 1,000 on my puny PowerBase? The scores absolutely blow away the very best performance on my 8500 with a 225 MHz 604e upgrade card (Processor=606, FPU=823) and those scores are better than double those of a stock 8500/120. Notice, however, that the disk speed is not great--this is more a measure of the hard disk performance than the card.
In assessing speed during general use, as opposed to graphics or scientific professional use, one of the useful benchmarks is how long it takes to launch an application. Table 3 shows a comparison of the PF with the PowerBase 180's original equipment:
Table 3. -- Launch Times in Seconds (mean of 3 non-consecutive tries)
| Application | 603e 180 MHz CPU | PowerForce G3 250* | Simulated G3 233** |
|---|---|---|---|
| Mac OS 8 Startup | 95 |
85 |
92 |
| Netscape Commun. 4.03 | 14 |
11 |
13 |
| MS Word 6.01 | 10 |
6 |
7 |
| CompuServe 3.02 | 9 |
7 |
7 |
| Adobe PhotoDeluxe 1.0 | 11 |
6 |
7 |
| Virtual PC - Windows 95 | 112 |
56 |
62 |
| cc:Mail 6.0 in VPC | 16 |
8 |
9 |
Launch times depend a lot on hard disk performance which is not great in the PowerBase. Still, the PF cut startup times up to 50%. It was hard to find useful everyday application operations that didn't result in trivial times like a second or less. Table 4 lists some application use benchmarks:
Table 4 -- Application Speeds in Seconds (mean of 3 non-consecutive tries)
| Application Operation | 603e 180 MHz CPU | PowerForce G3 250* | Simulated G3 233** |
|---|---|---|---|
Adobe PhotoDeluxe: - Rotate 130K image 90 deg |
9 |
3 |
5 |
Word 6.01: -spellcheck 9-page CV |
4 |
2 |
2 |
Virtual PC WordPerfect 6 - Open 9-page Word 2.x doc |
14 |
5 |
7 |
Virtual PC - Run Scandisk on C: drive |
35 |
14 |
16 |
Virtual PC - QuickTime movie sample |
24 |
22 |
23 |
Simulated PowerForce G3 233 Performance (11/20)
PowerLogix announced that it will release a PowerForce G3 233 with 512K backside cache and a CPU/cache speed ratio of 2:1 in a couple of weeks for $999. The backside cache size and the ratio between the CPU and cache can be adjusted on the PowerForce G3 250. Therefore, the PF 250's flexibility allows for a simulation of the characteristics of PowerLogix's upcoming cheaper G3 233 CPU daughtercard. In addition, Apple's new low-end PowerMac G3 233 DT has 512K of backside cache and it runs slower, i.e. at a 2:1 ratio of CPU speed to cache speed so it shares some of the same characteristics except most notably system bus speed (66 MHZ in the PowerMac). However, the system bus speed appears to have less effect on G3 MacBench scores than on other CPU's, presumably because of the large and fast backside cache.
The PowerForce's system bus runs at 48.36 MHz and the 250 MHz CPU runs up to 292.4 MHz or 17% faster. Using the same percentage acceleration, a 233 MHz CPU might run at 272.5 MHz. However, with a system bus speed of 48.36 MHz, the highest CPU setting not exceeding 272.5 is 266 MHz. I also chose a slower combination, 45/247.5 MHz, for MacBench comparison. The MacBench-marks of the simulated PowerForce G3 233 are shown in Table 5:
Table 5. -- MacBench 4.0 on Constrained PowerForce 250*
CPU/Bus Ratio |
System Bus Speed (MHz) |
CPU Speed (MHz) |
Backside Cache Ratio (Speed) |
Processor Score |
FPU Score |
Disk Score |
5.5:1 |
48.36 |
266.0 |
2:1 (133.00) |
829 |
842 |
278 |
5.5:1 |
45.00 |
247.5 |
2:1 (123.75) |
769 |
783 |
269 |
These results are better than for my 8500 with a 225 MHz CPU upgrade card and in the same ballpark as MacBench-marks (774-880) reported respectively for the Apple PowerMac G3 233 and the PowerLogix PowerForce G3 233.
I repeated the series of application launch and use tests for these settings--see the fourth column of Tables 3 and 4. Application launch and use times for the simulated PowerForce G3 233 are the same as or nearly the same as the optimized PowerForce 250 and much less in many cases than the stock 603e 180 MHz CPU. The launch time for VPC is very near that for my 8500 with a 225 MHz CPU card which does it in 60 seconds as compared with 62 for my PowerBase with the simulated G3 233 settings.
As a final check, I tried to boot with BeOS Preview Release 2--no go. I hope Be will support the G3 CPU in the near future.
Overclocking Concerns? (11/24)
Some may question the wisdom of overclocking a CPU and system bus. Historically, overclocking really didn't get one much in actual performance but it could result in a lot of grief in system stability, mostly with the 680xx CPUs and the early PowerPC CPUs. My perspective changed when I installed a PowerLogix PowerBoost Pro 225 in my 8500/120--it has been running optimally overclocked for nearly 4 months without any problems--granted, it hasn't been years. The results of the PowerForce G3 250 in my PowerBase above did not contain speed benchmarks without overclocking the system bus (40 MHz) and the CPU (250 Mhz). Since the CPU speed is determined by a ratio to the bus speed (e.g. 6:1 or 6.5:1), it can't be set at 250 with the system bus at 40. Therefore, I've included the results with the CPU slightly underclocked (240 Mhz) and slightly overclocked (260 Mhz). Here are the results:
| CPU** | CPU/Bus Ratio | System Bus (MHz) | CPU Speed (MHz) | CPU Score | Floating Point Score | Disk Score | Launch VPC-W95 (Sec) |
|---|---|---|---|---|---|---|---|
| 603e/180 | 4.5:1 |
40.00 |
180.0 |
329 |
343 |
222 |
112 |
| PF G3/250 | 6.0:1 |
40.00 |
240.0 |
887 |
767 |
277 |
62 |
| PF G3/250 | 6.5:1 |
40.00 |
260.0 |
970 |
829 |
277 |
61 |
| PF G3/250+ | 6.0:1 |
48.36 |
290.2 |
1083 |
928 |
282 |
56 |
**603e/180 MHz = stock PowerBase CPU with 512K L2 cache, Mac OS 8, MathLib and Virtual Memory "on". PF G3/250 = PowerLogix PowerForce G3 250 MHz, 1 MB backside cache, 1:1 CPU/cache ratio, Mac OS 8, MathLib and Virtual Memory "on". (+) optimized for maximum MacBench CPU score.
These results show that those concerned about overclocking can still get great performance from the PowerForce G3 250 without overclocking. In comparison, MacWEEK tested the new Apple PowerMac G3's with 66 MHz system bus speeds and obtained CPU MacBench-marks of 785 for the 233 MHz model and 895 for the 266 MHz model (Floating Point scores were not published). The better performance by the PowerLogix PowerForce G3 250 is due the 1 MB of backside cache (vs 512K) running at a 1:1 CPU/cache speed ratio (vs 2:1) and certainly not my discontinued PowerComputing PowerBase!
Conclusions (11/20)
The PowerForce G3 250 is one oustanding piece of hardware. The results for my PowerBase are pleasantly surprising. If you have a high-end computer with lots of RAM and the live-or-die need for speed with calculation-intensive jobs, then this card or its faster sibling, the PowerForce G3 275, is for you. For ~$1,900-2,700 your computer can be as fast as the never-to-be PowerTower Pro G3 and Motorola 6000 which were planned to be sold in the neighborhood of $5,000+.
However, this review is principally targetted at low-end computer systems like my PowerBase 180. It is truly amazing that the PF 250 MacBenches nearly as fast in my PowerBase as in other tests with higher-end computers. I used the PF 250 set at its best CPU speed (292.4 MHz, bus=45 MHz, cache=1 MB, CPU/cache ratio=1:1) for three days. My PowerBase was blazingly fast and I experienced not one crash nor freeze during intensive use of the applications that I frequently use. With the PowerForce 250, one can step up their PowerBase to be among the fastest desktops in existence.
Cooling the PF 250 down to about the level of the announced PowerLogix PowerForce G3 233 (CPU=266.0 MHz, bus=48.36 MHz, cache=512K, CPU/cache ratio=2:1 or CPU=247.5 MHz, bus=45.00 MHz, cache=512K, CPU/cache ratio=2:1), resulted in performance marks only marginally less than for the maximized PF 250. I used these settings for another three days of intensives use--again no crashes nor freezes. My PowerBase seems to perform feel-wise a little bit better than my 8500/225 and, of course, a lot better than with the 603e 180 MHz CPU.
Clearly, the PowerLogix PowerForce G3 250 in a PowerBase is a performance alternative to any new high performance computer. But better than that, at under $1,000, the announced PowerLogix PowerForce G3 233 daughter card in an exisitng PowerBase appears to be an excellent price/performance alternative to any new shipping computer.
For more information contact PowerLogix.
[Back to Macs Only!]
Saw something? Send a tip or a correction
The archive ran on reader tips. What did you see, where, and do you want the credit? Something wrong on this page? Say so and it gets fixed.