I'm just guessing, but what you're seeing is probably ATi's subpar OpenGL performance compared to Nvidia cards. For reference, I've encountered situations where my Radeon 9600 XT was slower than my GeForce 4 Ti4200 in Quake 3, mostly when using low resolutions like 640x480 and older Catalyst 6.2 drivers.

If you want a Core 2 vs. Pentium 4 comparison, Phil has an interesting video on the subject.

Joseph_Joestar wrote on 2024-06-28, 06:37:

I'm just guessing, but what you're seeing is probably ATi's inferior OpenGL performance compared to Nvidia cards. For reference, I've encountered situations where my Radeon 9600 XT was slower than my GeForce 4 Ti4200 in Quake 3, mostly when using low resolutions like 640x480 and older Catalyst 6.2 drivers.

If you want a Core 2 vs. Pentium 4 comparison, Phil has an interesting video on the subject.

Maybe ati is slower than a comparable nvidia card. But here we have a newer ati vs older nvidia.
And benchmarks show that fx5500 cant go higher than 25fps despite the cpu power, while newer ati does 47fps. (doom3 medium)
I am sure if I find an even faster cpu for core2 it will go above 50fps, maybe even 60fps. Ati is not saturated yet.
I'll watch the video.

PlaneVuki wrote on 2024-06-28, 06:48:

Maybe ati is slower than a comparable nvidia card. But here we have a newer ati vs older nvidia.

All I'm saying is, if you want a fair comparison, you should leave the rest of the components the same and just exchange the CPUs. Obviously, I mean on a desktop machine.

Otherwise, a lot of other factors can influence the results, including ATi's underwhelming OpenGL performance compared to Nvidia.

Joseph_Joestar wrote on 2024-06-28, 06:58:

PlaneVuki wrote on 2024-06-28, 06:48:

Maybe ati is slower than a comparable nvidia card. But here we have a newer ati vs older nvidia.

All I'm saying is, if you want a fair comparison, you should leave the rest of the components the same and just exchange the CPUs. Obviously, I mean on a desktop machine.

Otherwise, a lot of other factors can influence the results, including ATi's underwhelming OpenGL performance compared to Nvidia.

Well I don't have hardware to do a "fair" comparison, this is all I have.
My aim is not comparing p4 vs core2 in general.
I just happen to have these PCs, and I want to compare them.
But even if components are not same we can make analysis right?
Like hd 2400 xt obviously faster than fx5500, as expected.
But it is also expected that core2, and even core2 based celeron, be faster than northwood.
How then p4 beats core2 and its celeron in some cases?

This is my question.

My 2 cents:
1. Drivers overhead. Later video cards need faster processors because of newer drivers. They demand more cpu power.
2. Mobile GPUs limit their power to be more energy efficient. The same applies to processors.
3. Check working temperatures of your laptop. It may be due to overheating.

PlaneVuki wrote on 2024-06-28, 07:15:

Well I don't have hardware to do a "fair" comparison, this is all I have. My aim is not comparing p4 vs core2 in general. I just […]
Show full quote

Well I don't have hardware to do a "fair" comparison, this is all I have.
My aim is not comparing p4 vs core2 in general.
I just happen to have these PCs, and I want to compare them.
But even if components are not same we can make analysis right?
Like hd 2400 xt obviously faster than fx5500, as expected.
But it is also expected that core2, and even core2 based celeron, be faster than northwood.
How then p4 beats core2 and its celeron in some cases?

This is my question.

If you're just benchmarking CPU power then games aren't a good idea. You should use tools like SuperPi.

Game FPS results can be affected by a lot of factors. ATI/nVidia aside, what graphics acceleration backend were the games using?

If using later DirectX (9.0+) then maybe the ATI card will perform better, but if it's earlier DirectX (8.0 or before) then the opposite may be true.

PlaneVuki wrote on 2024-06-28, 07:15:

[...] […]
Show full quote

[...]

Well I don't have hardware to do a "fair" comparison, this is all I have.
My aim is not comparing p4 vs core2 in general.
I just happen to have these PCs, and I want to compare them.
But even if components are not same we can make analysis right?
Like hd 2400 xt obviously faster than fx5500, as expected.

How do you conclude that? If anything, the Quake3 scores show that in some things - OpenGL by the look of it - the HD2400XT is clearly slower than the FX5500. In any event they are two very different chips with different strengths and weaknesses.

But it is also expected that core2, and even core2 based celeron, be faster than northwood.
How then p4 beats core2 and its celeron in some cases?

This is my question.

To answer it you need more data. Apart from CPU, your two setups differ by GPU, GPU memory, motherboard chipset and connectivity, operating system, drivers, thermal constraints, type of RAM, amount of RAM and possibly also storage type/performance. All these factors influence both the results and the scaling results of your benchmarks.

With such platform differences, I would not consider these two systems at all suitable to answer the question you are asking, and if you want to try anyway, I'd focus on removing as many of the differences as possible:
- install the same OS on both systems for starters
- use the same storage medium on both
- if even remotely possible, get a Radeon HD2400Pro AGP for the desktop P4 system and try to tune core/memory clocks to same speed as on the laptop

That would still mean you are testing laptop (with thermal throttling) vs desktop, and PCIe vs AGP, and partly shared video memory vs totally dedicated, but at least GPU, software, drivers and storage would be level.

In terms of methodology, you're doing two good things on either platform by on the P4 exploring speed scaling of performance by testing with 100/133MHz FSB and testing cache/core scaling by taking the low-end Celeron and contrasting with the high-end C2D. But it's still two different things leading to an apples vs oranges comparison. At the very least, you could also explore FSB scaling on the laptop by thottling FSB down using SoftFSB or similar.

Finally, as others have suggested, games test a very wide range of system characteristics, which differ in your platform. If you're interested in CPU performance on its own, take a CPU-heavy benchmark instead, particularly given how different the platforms are that you are testing your CPUs on.

Don't want to do any of that? Fine, but the only conclusion you can then draw is that newer, known-faster CPUs on laptops aren't able to conclusively beat older CPUs on desktops, given rather different GPUs and a very different operating system. Can't make it more specific than that.

I have also quickly tested a pc with i3-6th gen (2ghz, 4 threads) + intel hd graphics.
This pc can run crysis at 23 fps on medium settings, and doom 3 medium at 69 fps, doom 3 high at 61 fps, beating the core2 in all as expected.
But it still gives 300 fps in quake3 low, comparable to p4.

My conclusion will be that quake3 is simply very old for benchmarking newer PCs. It doesn't always scale very well.
But games like doom3, css and crysis scale well, for now. I think doom3 is good opengl benchmark in this case. Can be used for PCs from about 2002.

Another conclusion for me is that northwood 2ghz is comparable to the single-core core 2 celeron at 2ghz (m550), at least in games optimized for p4.

Kruton 9000 wrote on 2024-06-28, 07:48:

2. Mobile GPUs limit their power to be more energy efficient. The same applies to processors.

This ^^

The T7250 is a 35W TDP part. It will inevitably throttle to some extend to meet that TDP.

PlaneVuki wrote on 2024-06-29, 05:56:

[...]

Another conclusion for me is that northwood 2ghz is comparable to the single-core core 2 celeron at 2ghz (m550), at least in games optimized for p4.

Too few caveats.

Northwood on a desktop board with GPU with dedicated memory is comparable to single core 2 Celeron on a laptop with thermal constraints with GPU with partly shared memory

Yeah, you can't directly compare laptop/mobile implementations with a desktop.

It's apples and oranges.

Maybe the 2400 is a little off pace, maybe drivers will improve it.
Some scores for it... https://www.notebookcheck.net/ATI-Mobility-Ra … 400.3773.0.html

Notice it's not all that quick on dx8, expand 3DM2001SE scores, it's below a 5700 mobile which is probably close to 5500 desktop.