Mediator vs Prometheus

[downix]

@Karlos

Thanks.

Now reminder, even if the Promethius is found slower than the Mediator, the Promethius is a lot cheaper than the other solution.

[Ilwrath]

@downix

Now reminder, even if the Promethius is found slower than the Mediator, the Promethius is a lot cheaper than the other solution.

Certainly.  Though it is good to have a numerical idea of the real-world performance difference (or lack thereof) between the two solutions.  I still suspect both systems will post similiar numbers.  What are the rest of the specs for your test system, Downix?

If the prommy turns in a solid performance, it would, indeed, make it a VERY tempting goodie.  (I have an extra Voodoo3 and RTL8139 sitting here...)

[downix]

@Ilwrath

I have 2 setups that will be used on each card

System 1:
68030 28Mhz w FPU
16MB fast RAM
1GB HD
OS 3.1

System 2:
68060-50 & 604e 233
64MB Fast RAM
9GB HD
OS 3.9

The logic here is simple, don't have a driver that better utilizes the CPU to "edge out" the competition.  Also, if the cards utilize non-DMA modes, you'll clearly see it thanks to the weak CPU on System 1.  By having both cards on both a high-end and a low-end, and comparing them side-by-side, you can get a better idea where the CPU loads are found, and where better driver utilization is.

[Karlos]

@downix

Hmm :-) *watches with intrest*

[mboehmer_e3b]

Half-true. You forgot, you can "stack up" PCI providing you have multiple PCI busses. (example being a northbridge with two PCI busses, I've seen some very odd embedded machines

Right, downix. I was speaking about one bus segment, but splitting up the bus into several segments is possible (with additional PCI-BCI bridges).
Anyhow, you can use this for speeding up data transfers between PCI cards on each segment, but for the processor itself it won't make much difference, as still only one is available The data rate on one segment will still stay the same.

using both for the same peripheral for added performance) It is this ability that was extended for PCI-Express (not PCI-X, don't mix the two) for their multiple-bus model.

You may also use two Zorro busses

[downix]

@mboehmer_e3b

Slight difference, you can bridge to get more PCI, plus there are north/southbridges with multiple PCI bus channels off of them.  Last time I checked, noone made a double Zorro3 controller.  (which would be interesting to see, actually)

[Terminills]

@nate

Just shipped them today... only like 3 months late mind you ... but hey it's on it's way now...

[downix]

Terminills wrote:
@nate

Just shipped them today... only like 3 months late mind you ... but hey it's on it's way now...

Thanks pal.

[Morley]

@Nightshade737

Actually, the Mediator doesn't plug directly into the daughterboard slot. It consists of two different pieces:

1. New daughterboard with inactive PCI slots, replaces the original one.
2. Zorro -> PCI bridge that interfaces the inactive PCI slots on the new daughterboard to the Zorro bus.

So the mediator interfaces it's PCI bus through the Zorro-bus, just like the Prometheus. No speed gain there, just a much neater and prettier solution.
The G-Rex, on the other hand, should be fastest solution, as it is connected to the "PCI"(?) slot on the Cyberstorm MK3 boards. But I believe the G-Rex is no longer being produced.

-Edit-
Oh, and if the Mediator eventually interfaced directly on the daughterboard, it wouldn't make any difference, as the daughterboard-connector is just "plain Zorro". The daughterboard is nothing but a riser-card.

[NightShade737]

Morley: Yeah, I worked out the inactive and bridge thing, but thanks for clarifying it. Also thanks for clearing up the other info though.

Hmm, wouldn't it be useful if someone made new G-Rex boards...  *cough*Schoenfield*cough*Power Computing*cough*.

Still waiting for the comparison of the Promethus and the Mediator though...

[T_Bone]

NightShade737 wrote:
Morley: Yeah, I worked out the inactive and bridge thing, but thanks for clarifying it. Also thanks for clearing up the other info though.

Hmm, wouldn't it be useful if someone made new G-Rex boards...  *cough*Schoenfield*cough*Power Computing*cough*.

Still waiting for the comparison of the Promethus and the Mediator though...

Well, there's the unscientific approach...

GOOGLEFIGHT!
mediator PCI ( 12 100 results) versus prometheus PCI ( 5 240 results)

The winner is: mediator PCI

damn.

rematch!!

 [troll]
mediator trojan ( 21 400 results) versus prometheus trojan ( 8 480 results)

The winner is:    mediator trojan
 [/troll]

[Morley]

@Nightshade737

Well, Vesalia still has the A1200, A4000D and A4000T versions of the G-Rex listed on their web-site. Looks like the A4000T and A1200 versions are in stock, as they have prices tagged onto them, but the 4000D version has "call" instead of the price tagged onto it...doh.

Lucky me has a 4000T, just needs some money :-D

How is driver support for the G-Rex going, btw? Anybody know?

[NightShade737]

Does the G-Rex use the same drivers as the Promethus? (OpenPCI) As the drivers are the only thing I am worried about...

[Morley]

@Nightshade737

Yes, OpenPCI is supported by the G-Rex, (or visa-versa? :-) ), but there aren't many drivers available. GFXcard support looks good though.
When it comes to drivers, it looks like Mediator has the best "coverage", but...I don't like RDB killing code in my drivers, no matter the reason behind it.

Check out the german G-Rex homepage. There are some forums there, in german sadly, which answers a lot of questions regarding the G-Rex.

[NightShade737]

At least the advantage of OpenPCI is that G-Rex and Promethus both benefit from it, so if drivers are made for the one, they will work on the other. At the moment the G-Rex seems like a better option for me... but on the other hand, it would be nice to see comparisons of all 3 in the same machine...

[NightShade737]

So, how are those benchmarks coming along? Would be even better if you could find a G-Rex to try too as that seems to be the most appealing option right now.