Click for SMOKING FAST SATA 6G SSDs!
Review: PowerForce G3 ZIF 400MHz Upgrade
400/200/1MB Adjustable Speed ZIF Upgrade for AIO/Beige/B&W G3
Since the ZIF CPU Upgrade market is getting so crowded, consumers are reaping the benefits of lower prices from the intense competiton. Unlike the other form factors of G3 upgrades, ZIFs have proven basically immune to the issues that plagued many brands of older Mac upgrades (see the FAQ and CPU Reviews page articles for more details). Since the spectre of bus speed and compatibility problems/ROM issues are not present in the Apple G3 systems, this means the consumer can concentrate on basically three factors - price, adjustability and special software/features. Considering that often times one vendor's software works with another brand and ZIFs need no software trickery to patch compatibility issues or to disable onboard cache, the control software is almost a non-issue. (Before using software from another vendor check their licensing statement.)
For all the reasons above, in the ZIF market no one company offers any clear advantage of compatibility over the others and therefore cannot demand a premium over their competitor's products. Other than adjustability, most ZIF upgrades are almost as generic as RAM. There are exceptions such as those models with fast IBM Cache, but for the mainstream models it's a buyer's market.
Interestingly, many ZIF upgrades offered under several brand names are actually made by the same OEM as Apple uses I'm told. This is not the case with Powerlogix's design but I'm told Powerlogix does supply a fixed speed model to a vendor that sells it under their own brand name (one of the higher priced dealers historically). And since most all ZIF models of the same rated speed often use the same cache and basic design, performance (at the rated speed at least) is literally identical. The adjustable models can provide some added value for those who wish to overclock, in addition to being compatible with all Apple G3 models (Beige, AIO and B&W G3). With that said, I'll move on to the specifics of the review model.
PowerLogix provided their latest 400/200/1MB ZIF G3 upgrade for review which contained an IBM 400MHz CPU and 1MB of Gavlantec brand cache. The cache used on this model is identical to the previously reviewed Powerforce 366/183/1MB ZIF and XLR8 400/200/1MB ZIF (not the rare and expensive IBM cache used on the original PowerForce G3/400/266/1MB ZIF and PF 466/311/1MB ZIF models ).
Since this ZIF is adjustable, it offers the bonus of being able to be used in all Apple's G3 models and *may* run faster than the rated speed reliably for risk-takers who want to overclock the card. I had mixed results on this sample - it ran 450/179 reliably in my B&W G3 as long as the case was closed, but had errors at even 433MHz in the Beige G3 (finder errors, applications would not launch) even though the CPU temperature was reported as 10+ degress C cooler in the Beige G3 than the B&W G3. As noted at the site many times - I do not believe the reported temperatures are even close to accurate, as in the case of the beige G3 the reported 11C (when overclocked!) was laughable (below the room temperature). The software simply reads the reported value from the CPU, but it doesn't take a rocket scientist to know that these values are way off base. I highly encourage some company to do some lab testing to determine a factor or additive to apply to the reading so that it is at least somewhat accurate. If necessary have a table of factors to apply to different CPU models (voltage/composition), similar to the way that PPC Checker reads the PVR value and determines if the CPU is copper or aluminum based.
Although I never recommend you overclock your CPU, I won't insult your intelligence and tell you boogie man stories about the practice either. (I'm sure some of you may have heard scare tactics from a retailer of fixed speed upgrades, who prefers to sell products with no adjustments and therefore less potential tech support due to incorrect or too high a speed setting. What is interesting is their products are often higher priced if you shop around.) As noted by thousands of owners in my Rate Your G3 Upgrade database, overclocking is very common even with the stock Apple ZIFs (see the G3-ZONE overclocking page for details on Apple motherboard speed settings). I'd estimate that 90% of the entries from adjustable card owners indicate they are running them overclocked (either cache or CPU). Although I have run overclocked CPUs on both PCs and Macs for years, I cannot recommend you do so as there are risks involved. Overclocking information and results are provided for reference only.
Installing the Zero Insertion Force (ZIF) socket upgrade is accomplished in a few minutes. Since bus speed is fixed by the Apple jumper block (see the G3-ZONE for details), there is only one adjustment setting needed - System Bus to CPU ratio which determines CPU speed (bus speed times ratio = CPU speed). Select the proper ratio as shown in the PowerLogix manual (or addendum sheet for B&W G3 owners) according to your G3 card model and system. Different settings (bus speed to CPU speed ratio) are required due to the different bus speeds of the Apple G3 systems.
For example, the beige/AIO G3 has a 66Mhz system bus and the B&W G3 has a 100MHz bus. This means that B&W G3 CPU speeds must be a multiple of 50Mhz, (ratio incrementes are in .5) so the setting for the B&W G3 for this 400MHz ZIF would be a 4.0:1 Ratio (100MHz bus times 4 = 400MHz CPU speed). The 66MHz bus speed of the beige/AIO G3 allows increments 33MHz (66MHz times .5). To run this ZIF at 400MHz in a Beige/AIO G3, you set the ratio to 6:1 (400/66MHz bus). As with all adjustable upgrades, I experimented with higher CPU speed, which for this particular sample, were reliable at 50MHz over the rated speed in the B&W G3 as long as the cover was kept closed.
Note: Although speeds shown here were 100% reliable during the product loan duration, I can't recommend overclocking to anyone else. The Powerlogix manual does not endorse or suggest overclocking the card, however their current FAQ Question 4 indicates they honor the warranty on their PowerForce G3 products regardless of speeds used. Based on last summer's G3 CPU Card Survey, the Current G3 Owner's Results and tens of thousands of mails I know that most owners of adjustable speed cards are risk takers and run them beyond the rated speeds.
For a guide to installing a ZIF upgrade, see the installation/docs page my PowerForce G3/400 ZIF review or download the PowerLogix Manual available in PDF (Acrobat) format. After installing the ZIF in your system, you set the white rotary switch on the corner of the module to the proper CPU speed indicated in the manual (currently B&W G3 settings are on an addendum sheet).
Speed (Ratio) Control Knob
This sets the bus speed multiplier (ratio) that determines the CPU speed. As mentioned previously, the advantage of adjustability is twofold; it allows the upgrade to work at the proper CPU speed in both beige G3/AIO and B&W G3s and it allows experimenting with higher CPU speeds (potentially risky, but often rewarding).
Cache Control Software:
Like every CPU upgrade I've seen except the XLR8 400ZIF (reviewed previously), this upgrade requires cache control software to enable and set the speed of the backside cache. The latest PowerLogix version at the time of the tests was v1.51. Powerlogix now has a software installer application that puts the control panel and Speedmeter utility on your hard drive without manual copying of files. I suggest you open the Cache control before restarting after the install and set it to the proper cache size and rated cache speed. Below is a picture of the v1.51 PowerForce Cache Control.
Cache Speed Selection:
As stated above this model is designed for the Beige G3, All-in-One G3 and the new Blue and White G3. There were no compatibility problems seen in any of the tests, including Retrospect backup software and Adaptec wide scsi cards (two sore points for many non-ZIF G3 Upgrades). OS X server ran fine but the backside cache was not enabled. (Of all the ZIF upgrades I've tested, only the XLR8 400 ZIF's cache was enabled without addon software.)
Since I used the Apple Studio LCD display, 1024x768 was the maximum resolution possible. I used thousands colors mode as it is my normal setting. Note the reference 1000/100% score in Macbench 5 is from a RagePro G3/300 at 1152x870, millions colors. Disk tests were not run since the size, fragmentation and free space make comparisons invalid and disk scores usually do not scale with CPU upgrades; normally the disk is the limit and already saturated (macbench test wise) with the stock CPU.
As usual, I used my standard working full extension sets (not trimmed). No Speed Doubler or Libmoto extensions active (see my Libmoto page for why I have not used that 'benchmark booster' for over a year). I do not believe in running non-essential 3rd party extensions as they are often a cause of compatibility issues (either now or later when OS versions are updated) and also many times they consume a lot of CPU cycles. SpeedDoubler can boost copy times, but in the last year I've stopped using it since many readers may not have it and to simplify my configurations. For a detailed list of B&W G3 and Beige G3 hardware used for this review, see the systems info at the bottom of this page.
Since this upgrade is most attractive to lower speed beige G3 owners I separated out the MacBench 5 scores from that system. Also included is an example of the RagePro graphics improvement.
Notes on Speeds in different Systems:
The following common/popular applications are primarily CPU/FPU speed bound and are used since I have a large series of comparison scores from dozens of systems/upgrades, as noted on my G3 Apps Test page. These results show the total time to complete renderings of sample scenes, create movies, etc.. As shown in the results, CPU speed (not bus speed or cache speed) is the primary performance factor, followed by Cache size and Cache/Bus speed to a much lower degree. All things being equal of course, a faster bus speed is better, although generally only applications like Photoshop show any benefit in these tests. If the application moves a lot of data over the bus, faster bus speeds should help, despite the 'isolation' the backside cache offers from slower main memory. For more details on the other upgrades/systems shown in the comparisons, see the CPU Card Reviews, Systems, B&W G3 and G3-ZONE areas of the site.
The following is a total time summary of the results from Photoshop 5 tests which were far more complete than just one Gaussian Blur filter - the results are the total time in seconds to perform 21 filter tests on a 10MB image. Photoshop was allocated 70MB of RAM to prevent disk activity (swap file use) from affecting the results. The top two results were tests I did to satisfy my own curiousity after replacing my B&W G3's mixed RAM with two matched 2CLK PC100 DIMMS from TransIntl.
For more details on PSBench5, see the Photoshop 5 Performance page, which lists the names of the 21 filter test used in the PSBench action and a link to the PSBench home page in addition to detailed system configurations and other test results.
Time to complete rendering two test scenes from the Bryce 2 'Sample Scenes' folder, Kai subfolder (I used these standard scenes to allow other Bryce2 owners to compare their results with mine). Default 16980k memory allocated to Bryce2. Anti-aliasing was enabled for all tests.
Infini-D 4.01 Rendering Test:
Time to render the "Chapter 7 Finished" scene. Options were set to:
After Effects 3.1 Rendering:
10MB movie rendering with an explosion special effect:
Game Performance:Tests were done in Rave Quake and Unreal. Games are one area where, after about 300Mhz at least, little benefit is seen from throwing more CPU speed into your machine. If you're well equipped with RAM (128MB or more) and have a G3 CPU already, a faster video card is a far better investment for improved game performance.
As shown in the above table of results (and based on reader comments), the Opengl drivers seem to have cost a few FPS in Rave Quake compared to the previous Rage128 1.0 driver update.
(Unreal tests prior to this review used B&W G3 used Rage128 driver update 1.0 as noted. The Beige G3 (RagePro) used the ATI universal driver update v3.3 found on most recent RAVE game CDs and available at ATI's web site. Unreal Beta v220b2 was used for all tests.)
Game Performance Conclusions:
As I've commented before in other fast ZIF Upgrade reviews, I'm beginning to think Rave or the MacOS is a bottleneck at some point, as 3D Games do not seem to scale with CPU speed as on the PC, even with good graphics cards (Voodoo2 or Rage128) and plenty of RAM. This was most clearly shown in my review of the PF 466 ZIF running at 550MHz with a Voodoo2 and Rage128 card - Unreal and Quake performance seemed to improve very little over the same configuration with a 300MHz G3 CPU. Lesson? Gamer's would be far better off buying a 3Dfx Voodoo2 or Rage128 card if your primary concern is game performance (assuming you have plenty of RAM already). As noted in my graphics cards reviews the Game Wizard (Voodoo2) was by far the best game performance addon you can buy - even with a 120MHz 604E (my slowest system) it still provided good Quake (3Dfx) performance. And if you read the IX3D chip based card reviews you know they don't run Rave games well (or at all in some cases) due to a lack of RAVE support in thousands color mode and (I believe) a lack of hardware 3D supported features. I'll have a review of the ATI Rage128 retail card as soon as they are available (mid May I'm told).
Granted a 603e or sub-200MHz 604 would see a benefit from a G3 CPU upgrade in games, but owners of most G3 CPUs should not expect visible speed improvements by just buying a faster CPU based on what I've seen in 3D intensive games. For those without a 3dfx card, using onboard video or a video card that lacks good 3D acceleration (the ATI RagePro is really the lowest end chip for 3D games now), I strongly suggest you look at a new video card if improved game performance is your primary concern. A good video card can add hardware 3D acceleration and features that lesser graphics chips can't deliver no matter how hard they are 'fed' with data. See the FAQ's 3Dfx & and Video card topics, the graphics card reviews page and/or check out my 3Dfx forums.
BTW - setting the desktop to 640x480 mode (vs 1024x768) increased framerate by 1.3fps at 640x480 mode with the RagePro/Beige G3 (less work for the display chip). Note the Voodoo1 and Voodoo2 cards have their own video RAM independent of the 2D display. (See the FAQ 3Dfx topic area for more details on the various models of 3Dfx graphics cards.)
Review ZIF Specs:
Note: 450/179 was only reliable with the B&W G3 case closed where the internal fans provide airflow over the heatsink. Although Speedmeter 1.3 reported 31C, the heatsink was felt much hotter and finder errors were proof the CPU was overheating.
For a closeup (zoomed) look at this upgrade card click here. Unlike their first ZIF design, there are no clearance issues with the ATI Rage128 graphics card in the B&W G3.
As expected, performance was good (on par with other 400MHz G3s in the same machine type) in every category except Games, which showed little to no benefit. For owners looking to run OS X Server right now, the XLR8 400Z might be a better choice as it required no software to enable the cache, at least in my B&W G3, however it would not run at the 450MHz speed this Powerlogix model did (your results may vary on each model however).
For owners of older Beige G3s or for B&W G3 300 owners this upgrade is still attractive now that prices have fell dramatically on 400MHz G3 upgrades, however for a limited time at least, the $299 OWC 366MHz ZIF may be the best value ZIF on the market. That's the unvarnished truth and I'd be less than honest to tell you otherwise.
Watch for limited time special offers at vendor sites and in the main site news and as usual, G3 upgrade prices change almost daily it seems so the price listed here will be obsolete very soon I expect.
Test Systems Used:
System Hardware Summary:
Thanks to PowerLogix for supplying the sample for review.
Back to WWW.XLR8YOURMAC.COM
Copyright © Mike, 1999.
No part of this sites content or images are to be reproduced or
distributed without written permission.
Users of the web site must read and are bound by the terms and conditions of use.