| Click here to select a new forum. |
| Why does a IIfx seem faster than an 040 accelerated IIci? |
Posted by: olePigeon on 2022-09-30 11:12:10
Until my IIci died, that was my main machine. I had a 40MHz 6040 installed in it. However, after using my IIfx, it seems a LOT snappier. I'm sure the raw CPU score will be better on the IIci, but for whatever reason the IIfx is just more responsive.
Or am I imagining this? |
Posted by: ArmorAlley on 2022-09-30 11:57:53
I'm on for helping with a few simple tests, say how long it takes for a specific CD with a lot of files & custom icons to be displayed or for the amount of time it takes to copy said CD to a hard-drive.
If you can pick a CD from the Macintosh Garden, let me know how fast your CD-ROM is and what kind of hard disk you have, I can match it in my IIfx.
The bus-speed & memory is the other big difference after the CPU. The IIfx has a 40MHz bus while that of the IIci is 25MHz.
Nonetheless, I am on for a comparison. |
Posted by: olePigeon on 2022-09-30 12:39:50
My IIci is currently dead. I'm using the IIfx now. |
Posted by: Byrd on 2022-09-30 14:00:13
The IIfx was Apple’s final last gasp for the ‘030, squeezing as much optimisation as they could muster. An accelerated 040 machine can benchmark much faster, but it’s like bolting on a fat exhaust system onto a car that can only handle a smaller part. You also have to consider a tightly honed balance with disk I/O, memory and other buses not just CPU speed. |
Posted by: MrFahrenheit on 2022-09-30 14:31:54
Until my IIci died, that was my main machine. I had a 40MHz 6040 installed in it. However, after using my IIfx, it seems a LOT snappier. I'm sure the raw CPU score will be better on the IIci, but for whatever reason the IIfx is just more responsive.
Or am I imagining this? Could some of the difference be the video card used? The IIci has integrated video and the IIfx would have used a NUBUS one. Or did you use the same NUBUS card in your IIci as your IIfx? |
Posted by: Trash80toHP_Mini on 2022-09-30 16:02:35
HRMMM?????
25MHz vs. 40Mhz System Bus = 160% overall throughput speed increase
? MHz 68040 vs. 40MHz 68030
? MHz subset of 68882 on 040 die vs. full 68882 @40MHz
DMA I/O Co-processors (pretty much only under A/UX though?)
MUUUCH FASTER IIfx memory on 40MHz system bus (interleaved vs. not?)
Dunno, I think I'd want at least a 33MHz Quadra to compete depending on code base/applications in use? Wouldn't bet on a 40MHz 68040 racehorse hitched to a 25MHz Conestoga Wagon unless there'd be a LOT of Cache on board and Code/Applications amenable to such support. 😉 |
Posted by: Phipli on 2022-09-30 16:41:37
25MHz vs. 40Mhz System Bus = 160% overall throughput speed increase *60% speed increase 😛 |
Posted by: olePigeon on 2022-09-30 18:53:53
@MrFahrenheit Exact same cards. I just moved all my IIci stuff over to my IIfx.
@Trash80toHP_Mini Those might something to do with it. 🙂 |
Posted by: Trash80toHP_Mini on 2022-09-30 19:34:02
*60% speed increase 😛 Indeed. 😉 Poorly stated, was a bit rushed when home for lunch with the power out. What's left of Ian is straggling just about straight through here on the way to Virginia. Dad's fine in Florida. Rain's relenting, power's good now, prayers are heading south. |
Posted by: joshc on 2022-10-01 05:18:37
IIfx has fancy ASICs for various things including faster RAM access I think so that may play into it too? |
Posted by: MrFahrenheit on 2022-10-01 06:14:57
I know from my own experience, on both an LC and a IIsi, that a 68040 upgrade is extremely limited.
I can’t remember the exact multiplier, but let’s say a 68040 is something like 3-4x a 68030 at the same clock speed.
In my experience the speed up using a 25mhz 68040 upgrade on an LC resulted in about 2x performance from the stock 16mhz 68020.
On the IIsi, using a 40mhz 68040 upgrade, it was about 2-2.5x the IIsi stock speed. That was with integrated video, though. The 68040 being twice as fast already (40mhz vs 20mhz), plus an estimated 3x performance increase for the same clock, should see 6x 20mhz performance, but it’s no where near that.
So it seems the whole rest of the system has some inefficiencies which result in the upgrade not performing to the full CPU potential.
Which would then make sense that a 68040 upgraded IIci would feel about 2x faster than a stock IIci, or about 50mhz, but in some ways that “50mhz” would have spikes and drops which the stock 40mhz IIfx would “smooth out”.
It’s the reason I don’t really care to have CPU upgrade cards in any of my vintage Mac’s, and instead just have a faster one if I want faster speed. |
Posted by: ArmorAlley on 2022-10-01 06:36:30
My IIci is currently dead. I'm using the IIfx now. Until my IIci died,.. You even started with this statement.
Well, once you either get a working IIci board or get your broken board fixed, I would love to test the hypothesis. |
Posted by: Trash80toHP_Mini on 2022-10-01 09:54:24
The best way to isolate a bus speed bump for benchmarking that I can think of offhand might be to move a G3/L2 back and forth between 6400/6500 boards? That's a 25% increase.
Is there a similar test to be done on 68K? Overclocking wouldn't seem to provide enough differential. Maybe underclocking something like the IIsi's 20MHz to 10MHz would be feasible? That should give a clear picture if SCSI remains functional. ISTR the controller being on its own clock, so might be worth a look? If it's derived from system clock you'd probably need to noodle out an interesting bodge. 😛 |
Posted by: Phipli on 2022-10-01 10:13:47
It’s the reason I don’t really care to have CPU upgrade cards in any of my vintage Mac’s, and instead just have a faster one if I want faster speed. The good ones have onboard RAM and copy the ROM into that RAM, that removes a lot of the slowdowns. That sort of thing had died out by the mid 90s. |
Posted by: jeremywork on 2022-11-19 02:37:31
I haven't tested an 040 in my IIci yet but I have tested it with two 50MHz 030 cards against the IIfx overclocked to 50MHz and can also report the IIfx seems to have a snap to it that the IIci and even the earlier Quadras don't quite match. Vs the IIci/Diimo, the 50MHz IIfx trades blows back and forth, only pulling ahead by 10-15% in a few tests. The most dramatic disparity I can see is in Apple Personal Diagnostics' memory test which favors the IIfx by about 25%, so I'm betting the difference is owed primarily to the fx's memory architecture. Those coprosessing routines will help during specific loads (LocalTalk transfers, etc) to great effect, but in my experience aren't likely to be felt in general navigation. The performance of the IIci's onboard video is very hard to beat with any Nubus interface, especially when the host is driven by a fast accelerator card. @trag had this to say a while back:
(https://68kmla.org/bb/index.php?threads/iifx-or-quadra-950-fight.23009/)
I don't know how much real world difference it makes, but the IIfx doesn't just have a faster bus speed, it does things with RAM Writes which cause them to retire in two cycles instead of six cycles. However, if there's a bunch of writes in a row, it doesn't speed things up any, except for the last one.
This is why the IIfx has the funny memory with sixty-four pins. It's needed to support the write-buffering feature. |
| 1 |