Reviews and Daily News with a  Difference!
Mac Models/Mods  | CPU Upgrades  | Storage  | Video  | Audio/HT  | Apps/OS/Network/Misc  | Archives  | Search
News, Tips, Reviews or Questions to News at Xlr8YourMac.com
8/7/01 Tuesday's News: Story DetailReturn to News Page

Memory Bandwidth Tests with New Rev of 7450 CPU in G4/867:
Published: 8/7/2001
Updated: 8/8/2001 for Stream results from 2.0 rev 7450 for comparison

UPDATE: GaugePro rates seem to vary a lot depending on how it is run and if other apps are running in the background, etc. Later other 2.1 rev 7450 owners wrote they were not seeing any improvements over the previous reported (w/2.0 rev CPUs) rates of appx 145MB/sec or so. Stream test results however do seem to show some benefit from the 2.1 rev over the 2.0, based on limited sample testing. [I only have access to a rev 2.0 7450, not a 2.1 to run my own comparisons.] See Stream test results below.

Update The GaugePro results still show under 200MB/sec in tests with even a Dual G4 1GHz. For this benchmark, the 7400/7410 G4s show much higher rates. Also interesting to PB G4 owners - see this update to the PowerBook G4/667 article that notes the lower memory bandwidth of the 550/667 (7541 CPU based) models vs the previous 7410 CPU based PB G4s. (And tests in Logic Audio showing higher performance with the 7410 models vs 7451 models.)

I asked a reader that reported the new revision of the 7450 CPU in his G4/867 to run some memory bandwidth tests to see if they showed better results than the previously reported 2.0 revisions of the 7450 CPU (used in G4/733s and in some G4/867s). Note that not all G4/867s may have the new revision CPU, as an earlier report from one reader noted a rev 2.0 CPU (the same as the earlier G4/733s).

Andrew's 2.1 rev 7450 showed more than 100MB/sec higher GaugePro memory test rates than a 2.0 rev CPU, but still not that impressive when compared to results from 7400/7410 G4s - even my G4/400 Powerbook (100mhz bus) had rates near the 867mhz (133mhz bus) rev 2.1 system.

Of course memory bandwidth is far from a linear indicator of real world applications performance since most apps are not memory bandwidth limited. (The CPU, hard drive, video card, etc. are also factors, usually far more important factors than memory bandwidth - at least if you're comparing modern SDRAM based Macs.)
(This info is only of interest to benchmark fans and those that are curious as to the pure benchmark results with the later rev 7450 CPU.)

" Mike -- Attached are the window shots of the tests you requested. I rebooted between each test (can I just note that this machine boots faster than ANY Mac or Windows box I've ever used? Woohoo! ;) ).

System is a 867MHz G4 with CDRW, OEM SCSI card and 2 CL2 256MB DIMMs, running MacOS 9.2 (obviously).

As I noted in the G4 forum, Gauge Pro's memory bandwidth output varies wildly when I run it on this machine, especially right after launch. It swings up and down by 70MB/sec or more. After about a minute, it settles on a "final" figure and only fluctuates a little, although this figure can be pretty different each time I run the program (for example, when I ran it last night, it settled on 243.1MB/sec. This morning, it stopped almost 20MB/sec. higher!) Add my voice to the chorus of doubt that Gauge Pro is an effective measure of the PPC7450.
GaugePro 1.1, the same version I sent Andrew, reports only about 145.5MB/sec on a Digital Audio G4/733 (rev 2.0 CPU) and shows very little change (a few MB/sec max) usually.
[NOTE: Sometimes if I ran GaugePro when another program was running, then switched back to GaugePro I saw much higher scores (from 30 to 60MB/sec higher at times) than if run from a clean boot/only app running. The "usual" scores from a clean boot were 145MB/sec - but once while the QT player was open with a movie loaded, I saw as high as 204MB/sec rates. Not exactly the sort of thing that inspires confidence in using GaugePro for comparisons. I assume the scores below were run as requested, from a clean reboot/no other apps running. Later rev 2.1 CPU owners said they did not see any real improvement in memory bandwidth based on these types of tests.]
The 7450 CPU rates in this benchmark are much lower than even my 100mhz bus speed PB G4/400 which reports 239.2MB/sec., or any G4/AGP system that used 7400/7410 CPUs. A previous report from a G4/867 owner with a 2.0 CPU sent a screenshot showing only 144MB/sec results, but GaugePro doesn't seem to be the best benchmark for these tests. See below for Stream (cross-platform benchmark) results.-Mike

Let me know if there's anything else you would like me to test! I'm happy to contribute.
Andrew "

The screenshots from GaugePro 1.1 below compare a prev. 2.0 rev 7450 G4/867mhz vs Andrews later rev (assumed to be 2.1 rev from the PVR reported by PPCChecker). Although they are still not that impressive compared to 500Mhz G4/7400 or 7410 CPUs with external 1/2 CPU speed L2 cache and no L3 cache, where the 867's have onchip full speed 256KB cache and 2MB DDR L3 caches running at 1/2 the CPU speed (effective DDR rate). Later 2.1 CPU owner GaugePro tests showed lower rates than Andrew's below, so this may be another case of GaugePro rates fluctuating.

[Note - one G4/867 owner sent a note that his 2.1 rev CPU still has low GaugePro rates as shown in this screenshot. Later more 2.1 7450 CPU owners sent similar (low) results, so the above image may just be another case where GaugePro just creeps up or has a lot of varation in some cases.]

As shown since the first G4/7450s were sold, GaugePro does not properly ID the 7450's CPU type or CPU speed. (Newer Tech folded before the 7450 G4s were released so GaugePro wasn't updated.) I don't think this invalidates the memory rates shown however, since other memory benchmarks in the past have shown much lower rates with the 2.0 7450 CPUs also.

Stream Results: There's no G4 optimized version of Stream, but results from the 601 and 604 optimized versions as well as the "unroll" version are below. Results of 2.1 rev 7450 are on top, rev 2.1 7450 below. (Same g4/867 new Quicksilver systems otherwise.)

(Thanks to Danny for the rev 2.0 G4/867 Stream results.)

ThroughPut 1.5: This is Rene Trost's graphics card pure benchmark.

    (Rene's comments)
    " ThroughPut is a little application, written in PPC assembler, that tests how much data your Mac can push through the PCI or AGP port to feed your graphics card with data."

(See his website for details and download links.) There is no performance difference in this test between the two CPU revisions.

In my G4/733 Digital Audio System w/GeForce3 review I posted a graph of Throughput results (as well as real world tests) that also compared the GF3 vs Radeon, G4/733 vs G4/500 DP, etc.

CPU Revison Checking: As noted here previously, you can check the revision of your PowerPC by using our PPC Checker utility. (It has been updated to correctly ID the rev 2.1 7450 CPUs.)

Return to Main News Page

Click for iPhone Accessories!

(Ad/Sale items)

= UPGRADES by Mac =
Upgrades just for

= Used Macs =
(Click for List)

SSDs from under $50!
Fast SSDs for Most Macs/PCs

= ThunderBolt =
Drives & More

Up to 6 TB HDs!
Hitachi, WD, Seagate, Samsung HDs

= 2.5in HDs & SSDs =
Notebook Hard Drives and DIY drive/case kit bundles

Lifetime warranty RAM Upgrades!

Internal and External Superdrives/Blu-Ray drives

Graphics cards, TV tuners, DVR, adapters and more.

Apps, Utilities, OS, VM, Games and more

WiFi and Bluetooth Devices/Adapters/More

= Repair Service =
for iPhone, iPad, Macs

= iPad/iPhone/iPod =
Accessories, Cases, Repairs & More

NuGuard iPhone Case *Extreme* Drop Tested!

XLR8YourMac T-Shirts


Help Fight Hunger

= back to www.XLR8YOURMAC.com =

= Other Site Topic Areas =
Mac Mods/Upgrades | CPU Upgrades | Storage | Video | Audio/HT | Misc/Software | Search | Recent

Copyright © , 1997-2014. All Rights Reserved
All brand or product names mentioned are properties of their respective companies.

Legal: Site Privacy and terms/conditions of use.