Intel’s Core Duo T2400 has a maximum operating frequency of 1.83GHz, 65nm process, 2MB of L2 cache, and 667MHz FSB; however, how does this dual-core component fare under Linux? Phoronix has taken a look at the Intel Core Duo T2400 in conjunction with the Lenovo ThinkPad T60, and has comparison results against the previous Pentium M 750 1.83GHz.
But I think the graphics benchmarks are both futile (X300 vs X1400?) and very interesting that you still can’t play Doom 3 & Quake 4 under linux with a decent framerate with a ati graphic card as recent as a X1400, which is a disappointment.
I’ve also been searching for some kind of benchmarks and comments on how good/bad intel embedded graphics work on linux (with xgl, etc.) Anyone want to share some pointers? It’s kinda hard to find a cheap, light laptop without intel graphics these days.
Anyone want to share some pointers?
I have Intel 855GME, forget about XGL. AIGLX is what you’ll need. My ancient integrated graphics just crawls while using XGL but on AIGLX I can get quite smooth animations and effects. I suppose that with Intel GMA900 or GMA950 it would be just as smooth as on my desktop computer running XGL on Nvidia GF6600.
I didn’t try to run many games on my notebook but suffice to say that I had to lower details to very low in Neverball
I have an i915GM, I say it could be better, on Windows it does about 4100 3dmarks in 3Dmark2001SE, on Linux it runs XGL smoothly, but if I start something like MPlayer in fullscreen or some 3D game like PlanetPenguin Racer, it fails miserably. It does not run Google Earth for Linux very well, and I didn’t try other 3D games (I am stuck with MAME
)
Edited 2006-07-02 07:49
same here with my i915 laptop chipsset. but realize your slow video performance and opengl performance is simply due to not having finished XGL drivers. The only thing currently accelerated in the i810 series of drivers is the basic compositing effects; everything else is CPU driven. Try using full screen video with the “indirect_fullscreen_bypass” or whatever that option is, and your video will speed up significantly in fullscreen mode.
I am interested in a Core 2 Duo laptop (why get a Core 1 Duo when 2 is right around the corner?) to speed up my Gentoo compile times. Other than that I hardly use my Pentium M 1.7GHz.
It’s interesting to see what looks like a speedup of 2.299 on a dual-core cpu (e.g. 20.10 fps vs. 46.20 fps). I had to remind myself that frame-per-second stats are based on wall time of a very specific stream of execution and not actual speedup of the entire process. These impressive results on average fps seem to make sense when considering two semi-realtime threads fighting for the cpu, especially knowing at least one thread is trying to interface to the video card. It seems the thread(s) interfacing to the video card were starved by more than 50% with with a single CPU and had more CPU dedication with the dual-core making it look like a speedup greater than the absolute max of 2.0.
Of the charts in the article, the frame-per-sec statistics are probably the most useful since it’s more or less “isolated” to a CPU-video relationship. Some of the other stats involve heavy disk i/o throwing far too many more variables into the test other than the CPU.
I don’t want to embarass you, but what were you smoking when you wrote that post???
First, it isn’t even comparing 1 core versus 2 core directly – its comparing a PM vs a Core Duo. Those are completely (well, slightly different) chips. For example, maybe that 2.299 speedup was greater than 2.0 because it liked the different caches the Core Duo has. Or different latencies, better SSE performance, etc….
Second, it is true many of the non-GPU tests involved disk i/o which makes them not so great. However, they at least had 2 hard drives in a similar performance range. The GPU tests say almost nothing about the core duo because of the completely different GPUs. The games they tested are going to be 99% GPU limited which means they were basically a test of the X300 vs X1400. They didn’t even have the same amount of memory – the X300 only had 64MBs, which is going to hurt it quite a bit in some of those games like Doom3 and Quake4.
[Edited for spelling]
Edited 2006-07-01 05:39
Not embarrassed. Point taken.
However, a multi-threaded soft-realtime app _will_ see significant speedup from having two cores vs. one. As you mention, though, there’s enough other hardware factors as well to consider as contributing to the overall speedup. Though, it would also be silly to simply dismiss how threads play with single vs dual core cpus.
Different video cards, different total RAM, different size HDD’s…
Seriously, was there a point to this comparison?
I ignored the gaming benchmarks because of the staggering difference between the video chipsets and hard drive speeds, but the compilation and compression benchmarks were interesting. I would have preferred seeing a Core Solo vs. Core Duo benchmark article though; that would have been the most accurate comparison.