Old discrete PCI card or newer onboard graphics: which is better?

A friend of mine has been building a PC out of some components he's been
given. It's a 1.7GHz P4 with 1GB of RAM running Windows XP SP2 - nothing
all that fancy, but as he's upgrading from a P2 it's a fairly sizeable
jump in speed!

Anyway, the system has an ASRock P4VM900-SATA2 motherboard, like so:

http://www.asrock.com/mb/overview.as...=P4VM900-SATA2

This has onboard "VIA Delta Chrome" graphics, a brand about which I know
more or less nothing other than that a current discrete card would
destroy it. The PC is going to be used largely for 2D work (especially
Photoshop) but some light gaming ability would be nice. "Light" doesn't
mean Far Cr; it means older games like Tomb Raider:Angel of Darkness.

My friend has an old but fully working PCI (conventional, not Express)
32MB Radeon 7000 graphics card from his old P2 available as a potential
alternative. He knows that onboard graphics tend to be very slow, and is
looking into upgrading to a half-decent discrete card in due course. For
the time being, though, should I advise him to use the ancient Radeon or
the modern-but-onboard VIA?

(Of course, "try it and see" is one answer, but I'm interested for myself
to know where the "tipping point" is at which a really old discrete card
like the 7000 becomes worse than much newer onboard graphics.)

On or about Wed, 09 Apr 2008 10:45:03 -0500 did David Buttery
<rabbiteer@gmail.com> dribble thusly:

>A friend of mine has been building a PC out of some components he's been
>given. It's a 1.7GHz P4 with 1GB of RAM running Windows XP SP2 - nothing
>all that fancy, but as he's upgrading from a P2 it's a fairly sizeable
>jump in speed!
>
>Anyway, the system has an ASRock P4VM900-SATA2 motherboard, like so:
>
>http://www.asrock.com/mb/overview.as...=P4VM900-SATA2
>
>This has onboard "VIA Delta Chrome" graphics, a brand about which I know
>more or less nothing other than that a current discrete card would
>destroy it. The PC is going to be used largely for 2D work (especially
>Photoshop) but some light gaming ability would be nice. "Light" doesn't
>mean Far Cr; it means older games like Tomb Raider:Angel of Darkness.
>
>My friend has an old but fully working PCI (conventional, not Express)
>32MB Radeon 7000 graphics card from his old P2 available as a potential
>alternative. He knows that onboard graphics tend to be very slow, and is
>looking into upgrading to a half-decent discrete card in due course. For
>the time being, though, should I advise him to use the ancient Radeon or
>the modern-but-onboard VIA?
>
>(Of course, "try it and see" is one answer, but I'm interested for myself
>to know where the "tipping point" is at which a really old discrete card
>like the 7000 becomes worse than much newer onboard graphics.)

I had never heard of the DeltaChrome chip, either. You piqued my interest
enough to have a look, and it seems that the Radeon 7000 would have an edge in
performance. Both are atrocious compared to modern cards.

The only benchmark figures I could find that included both are from here:

http://freestone-group.com/video-car...lts-v1002.html

The VIA DeltaChrome had a benchmark value of 35-38. The Radeon 7000 values
(can't tell which would cover the PCI card) are 42, 57, 59, and 11-73. The
latter would seem to cover a whole range of 7000-based cards.

I downloaded that program and ran the benchmark, and got a score of 2247 with
my 8800 GTX (stock speeds). I have a video transcoding in the background, but
set its priority to low, and that of the benchmark to high, so that figure
should be reasonably close to the max. The entire benchmark is a rotating
globe, and the figure would seem to be an average FPS value.

There's a newer version of the same program, which has an apparently more
detailed globe (on which all the continents are east-west reversed, oddly
enough), which scores 611 on my system, and 7 on the only reported Radeon 7000
score.

I think both the DeltaChrome and Radeon 7000 are going to be iffy even with
TR: AOD, which was pretty graphically demanding when it came out.
--
- Mike

Ignore the Python in me to send e-mail.

On Thu, 10 Apr 2008 02:37:13 -0400, Mike Ruskai wrote:

> I had never heard of the DeltaChrome chip, either. You piqued my
> interest enough to have a look, and it seems that the Radeon 7000 would
> have an edge in performance. Both are atrocious compared to modern
> cards.
>
> The only benchmark figures I could find that included both are from
> here:
>
> http://freestone-group.com/video-car...hmark-results-
v1002.html
>
> The VIA DeltaChrome had a benchmark value of 35-38. The Radeon 7000
> values (can't tell which would cover the PCI card) are 42, 57, 59, and
> 11-73. The latter would seem to cover a whole range of 7000-based
> cards.
>
> I downloaded that program and ran the benchmark, and got a score of 2247
> with my 8800 GTX (stock speeds). I have a video transcoding in the
> background, but set its priority to low, and that of the benchmark to
> high, so that figure should be reasonably close to the max. The entire
> benchmark is a rotating globe, and the figure would seem to be an
> average FPS value.
>
> There's a newer version of the same program, which has an apparently
> more detailed globe (on which all the continents are east-west reversed,
> oddly enough), which scores 611 on my system, and 7 on the only reported
> Radeon 7000 score.
>
> I think both the DeltaChrome and Radeon 7000 are going to be iffy even
> with TR: AOD, which was pretty graphically demanding when it came out.

Interesting; thanks. I ran the newer benchmark on my own PC (I have an
8600GT) and got a score in the 140s. That's at the low end of the results
for that card, but considering that I wasn't after a very accurate result
and so left a few background programs running that didn't give me any
cause for concern.

A shame the DeltaChrome chip turns out to be so slow; I was rather hoping
it would be in the same league as something like the integrated Intel 945
or GeForce 6100 chips, but sadly not. My friend's PC has inherited a
pretty breathless PSU, you see (10A on the +12V rail...) so an upgrade to
a remotely modern graphics card is also going to necessitate some extra
expenditure on a new power supply.

--
Bewdley, Worcs. ~90m asl.

David Buttery wrote:

>A friend of mine has been building a PC out of some components he's been
>given. It's a 1.7GHz P4 with 1GB of RAM running Windows XP SP2 - nothing
>all that fancy, but as he's upgrading from a P2 it's a fairly sizeable
>jump in speed!

Well, not all that much, really. The early < GHz P4's were not good
performers.

Seriously, I'd say "why bother" unless it's absolutely free.

On Fri, 11 Apr 2008 14:02:10 -0500, chrisv wrote:

> David Buttery wrote:
>
>>A friend of mine has been building a PC out of some components he's been
>>given. It's a 1.7GHz P4 with 1GB of RAM running Windows XP SP2 - nothing
>>all that fancy, but as he's upgrading from a P2 it's a fairly sizeable
>>jump in speed!
>
> Well, not all that much, really. The early < GHz P4's were not good
> performers.
>
> Seriously, I'd say "why bother" unless it's absolutely free.

It *was* absolutely free. <g>

Actually I made a mistake in my initial post: it's a 1.8GHz processor,
not a 1.7GHz, though still a Williamette. And I don't have any quarrel
with saying that the early P4s were pretty disappointing chips.

But I still think you're being rather harsh in your assessment. I know
benchmarks are to be taken with several bushels of salt, and in any case
I couldn't find one that had both chips in, but:

http://www.cpubenchmark.net/low_end_cpus.html

has a bar for the AMD-K6, with an average frequency of 478MHz. They
produce a score of 53. Doing some back-of-the-envelope maths, picking a
figure out of the air that the K6 was about 1.5 times slower clock-for-
clock than the P2, gives us a hypothetical benchmark result for the
P2-400 of 400/478*53*1.5=66.5 or thereabouts.

The P4-1.8 bar on the same chart gives a score of 246, around 3.7 times
faster than the P2. That might not be the speed increase that the passing
years and increasing clock speeds suggest it ought to have been, but it's
certainly one that's worth having, especially if you can get it for free!

--
Bewdley, Worcs. ~90m asl.

Hmmm, I didn't take account of the fact that some 1.8GHz chips were
Northwoods. But the 1.5GHz, which was only produced as a Williamette, has
a score of 198, just under three times faster than the P2-400. Sure,
synthetic benchmark and all that, but the only other thing I have to go
on is my friend's reports of how it feels subjectively - and that's
"quite a lot faster".

--
Bewdley, Worcs. ~90m asl.

On or about Fri, 11 Apr 2008 19:33:01 -0500 did David Buttery
<rabbiteer@gmail.com> dribble thusly:

>Hmmm, I didn't take account of the fact that some 1.8GHz chips were
>Northwoods. But the 1.5GHz, which was only produced as a Williamette, has
>a score of 198, just under three times faster than the P2-400. Sure,
>synthetic benchmark and all that, but the only other thing I have to go
>on is my friend's reports of how it feels subjectively - and that's
>"quite a lot faster".

All P4's were garbage chips, but they were also clocked up so high that they
had no choice but to be faster than any older Intel chip, including the PII.
There's no question that a 1.8GHz P4 is quite a lot faster than a 400MHz PII.
Probably 3.5+ times as fast.
--
- Mike

Ignore the Python in me to send e-mail.

On Fri, 11 Apr 2008 23:08:04 -0400, Mike Ruskai wrote:

> All P4's were garbage chips, but they were also clocked up so high that
> they had no choice but to be faster than any older Intel chip, including
> the PII. There's no question that a 1.8GHz P4 is quite a lot faster than
> a 400MHz PII. Probably 3.5+ times as fast.

Well, there you go: my earlier estimate of 3.7x seems to have been more
or less right. And *all* P4s "garbage"? Right from the earliest
Williamette to whatever it is they're just phasing out now? I suppose it
depends on your definition of "garbage". For most people, if it'll do
what they want at acceptable speed, it's not garbage. Maybe not a best
buy, but that's not the same thing!

In a last-ditch attempt to stay even vaguely on-topic for this group,
I'll mention that my old P3-533 with a GeForce 2MX was surprisingly
capable (for its age) at playing games. I suppose it must be that
reviewers tend to assume that people won't look at a game unless it has
all the eye candy turned on, and so benchmark them that way. I mostly
play racing games, and in 90% of cases turning things down to medium
results in better gameplay as there are fewer visual distractions to the
business of actually racing!

--
Bewdley, Worcs. ~90m asl.

On or about Sat, 12 Apr 2008 15:18:25 -0500 did David Buttery
<rabbiteer@gmail.com> dribble thusly:

>On Fri, 11 Apr 2008 23:08:04 -0400, Mike Ruskai wrote:
>
>> All P4's were garbage chips, but they were also clocked up so high that
>> they had no choice but to be faster than any older Intel chip, including
>> the PII. There's no question that a 1.8GHz P4 is quite a lot faster than
>> a 400MHz PII. Probably 3.5+ times as fast.
>
>Well, there you go: my earlier estimate of 3.7x seems to have been more
>or less right. And *all* P4s "garbage"? Right from the earliest
>Williamette to whatever it is they're just phasing out now? I suppose it
>depends on your definition of "garbage". For most people, if it'll do
>what they want at acceptable speed, it's not garbage. Maybe not a best
>buy, but that's not the same thing!

Well, let's call it a spectrum of trashiness. To a nerdist who cares about
design, they are all garbage. Every single one was a huge step back on
performance per clock rate than Intel's own P3. In other words, if you were
to take a P3 and any P4, and run them at the same clock rate, the P4 would
lose.

To an end-user, there's still an element of trashiness, because their
ignorance of the atrocious design decisions made by Intel for the P4 series
did not protect them from the large power consumption and thermal footprint
thereof.

Only when a group in Intel went back to the drawing board for a new mobile
chip (an application for which the power-hungry P4 line is worse than
unsuitable) did the company get a design that was a genuine improvement, which
had good performance at much lower clock speeds.

>In a last-ditch attempt to stay even vaguely on-topic for this group,
>I'll mention that my old P3-533 with a GeForce 2MX was surprisingly
>capable (for its age) at playing games. I suppose it must be that
>reviewers tend to assume that people won't look at a game unless it has
>all the eye candy turned on, and so benchmark them that way. I mostly
>play racing games, and in 90% of cases turning things down to medium
>results in better gameplay as there are fewer visual distractions to the
>business of actually racing!

Most benchmarks are definitely guilty of using multiple games as a means of
testing the performance of a range of hardware.

Some do have measurements for lower resolutions, even with lower detail
settings in the specific game. But no doubt for reasons of feasibility, they
almost invariably test with only modern systems, so that the lower settings
show a ridiculously large numbers that cannot be extrapolated for older rigs.

That's why we're left with the task of tracking down old benchmarks and trying
to find a numerical connection between old hardware and new hardware.
--
- Mike

Ignore the Python in me to send e-mail.

Mike Ruskai wrote:

>All P4's were garbage chips,

The Northwood was a decent chip, in it's time...