P4 Northwood 2.53GHz + D850EMV2 Review - By Mark "Frugal" Bush Page 2 of 2
Synthetic Benchmarks
Now let's have a look at the chip. Well rather than looking at all the features of the chip we'll get straight down to the benchmarks, after all you dont care if it supports SSE2 or how many transistors it has (ok it has 55 million if it was bugging you:), you want to know how fast it will run your games .... right? Ok we'll get to that in a minute, but first let's have a look at some synthetic benchmarks using SiSoft Sandra and 3DMark 2001se.
CPU Arithmetic Benchmark
Looking at the Dhrystone scores you'll notice that the Athlon XP 2200+ edges past the P4 2.53GHz slightly. This is pretty impressive for AMD but don't read too much into it at this stage. The Whetstone tests on the other hand give the P4 a much bigger lead as you would expect considering the difference in clock speeds between the 2 chips.
CPU Multimedia Benchmark
In the integer tests the Athlon XP 2200+ once again edges past the P4 by a very small margin. However the P4 again recoups this in the floating point scores. This shows that the Athlon is a very powerful processor but Intels SSE2 can give the P4 quite an edge, considering more and more games are being released with SSE2 optimisation this can certainly have an impact with games.
Cache & Memory Benchmark
Here though is where the Pentium 4 really starts to stretch its legs, offering double the performance of an Athlon XP 2000 with DDR. The PC1066 RDRAM with its theoretical 4.2GHz of bandwidth really starts to make a difference. PC800 never really demonstrated any performance advantage over DDR but the dual pipelined PC 1066 RDRAM at 533MHz (1066MHz effective) really does start to shine. In the memory bandwidth tests the memory only managed around 3.2 GHz but I suspect that is because the D850EMV2 does not officially support PC1066 ram, mind you 3.2GHz is more than enough bandwidth for even the 2.8GHz and the impending 3GHz CPUs.
3DMark 2001se
800 x 600
15450 3D Marks
1024 x 768
14392 3D Marks
1280 x 1024
12227 3D Marks
1600 x 1200
10406 3D Marks
As you can see the 3D Mark scores are pretty impressive, even managing to go above 15000 at 800x600 and staying relatively close at 1024 x 768. Even at 1600 x 1200 we still manage to stay above 10,000. The Radeon 9700 is going to have a lot to do with these scores but 3DMark takes the complete system into account and so we are seeing the synergy between the memory, CPU and GPU. A good video card will show you some 3D love if you dont hold it back with a poor CPU.
Gaming Benchmarks
Now on to the only thing that really matters, actual performance in games.
Falcon 4
For a sim that was released in 1998 Falcon really is a CPU hog, we are
only just getting CPU's that can really handle F4. Even a 1.4GHz CPU can
be brought down to single figure frame rates over the forward line of
troops. I left the bubble and density sliders at the SP3 reccomended
settings of 3 and 6 respectively. The graphics are set to 1600x1200x32
All other graphic sliders were at maximum, all advanced options ticked. I
then flew a selection of campaign missions out of Seoul airbase (not
ideal for benchmarking but the only true in game test).
Day 1 11.00 OCA Strike
1600 x 1200 Min 38fps fps Max 76fps Average 66fps
Day 1 21.30 CAS
1600 x 1200 Min 39fps fps Max 76fps Average 68fps
Day 2 17.00 OCA Strike
1600 x 1200 Min 38fps fps Max 76fps Average 76fps
This system was able to handle Falcon 4 comfortably. The framerates drop
down to around the 40fps area on occasions, particularly on take off
over Seoul, and again over heavy activity. The real maximum fps in the
tests was 200fps but I ignored this figure as I do not believe this is
an accurate reflection of the actual frame rate. The frame rates stayed
around the maximum figure for most of the mission with only the
occasional drops to the minimum. We see that on day 2 where the enemy
activity is lighter and the "wall of Migs" is reduced, the fps are
slightly higher. I suspect that if I had used a dedicated benchmarking
TE the fps would have been lower but I wanted to see how this performed
in an actual campaign rather than in a benchmark TE.
Microsoft Flight Simulator 2002
For FS2002 I used the High global scenery setting at 1600x1200x32 then flew the "Along the Rockies" flight from the "Scenic Wonders".
Min 58.8 fps Max 64.6 fps Ave 60 fps
60 fps was pretty impressive so I decided to see if I could push the CPU a little harder, I increased the Terrain Mesh complexity to 90%, I also increased the autogen density to Dense and turned on Dynamic Scenery set at Dense. Even with these settings FS2002 was able to maintain around 40fps most of the time although it did drop into the low twenties on occasion.
CFS 2
CFS 2 was run at 1600x1200 32bit with all graphics options at their highest settings.
Min 83 Max 117 average 90
The framerates never dropped below 80. Again the system was more than
capable of running CFS2 without even breaking up a sweat.
Battlefield 1942
This new shooter from EA is known to be a system hog so it should give the system a good test. The settings used were 800x600x32 High detail settings. I also had 16x aniso enabled and 6x FSAA as I wanted eye candy. The testing was done in the main single player campaign.
Average 90 fps
The framerate was very smooth and rarely dropped below the 90fps. The Radeon has to take a lot of the credit for this but considering BF1942's reputation as a system hog the CPU certainly deserves some credit.
Unreal Tournament 2003
800x600
Fly By 188.351768
Botmatch 68.192522
1024x768
Fly By 179.157562
Botmatch 64.374580
1280x960
Fly By 154.033981
Botmatch 60.915012
1600x1200
Fly By 108.600288
Botmatch 62.667820
One thing you can see from these benchmarks is that the Fly By tests the fill rate of the video card, and so the fps are very high, but take a noticable hit as the resolution goes up. The Botmatch on the other hand, is CPU intensive, and as a result maintains a pretty steady 60fps through the increases in resolution. UT2003 is a pretty good benchmark because, despite the eye candy, UT 2003 is predominantly CPU limited rather than GPU limited. We can also see from Fly By results that the P4 and i850 chipset have more than enough available bandwidth to satisfy the Radeon 9700 Pro's needs.
Overclocking
As already mentioned, the Intel board does not allow for overclocking, so I was unable to test this aspect of the chip. A friend of mine is running a 2.53 GHz at slightly under 2.9GHz so there is definitely some scope there. Realistically though I have yet to find a game that pushes this CPU enough to warrant overclocking it.
Conclusion
The chip ran completely stable at all times, the low power requirement (it's rated at 57.8W) keeps it running exceptionally cool. Idle temperatures were around 36c and even under heavy load it never went over 50c. The performance of this chip would keep any gamer happy, currently the only thing that can touch it when combined with PC 1066 RDRAM is the 2.8 GHz P4 equipped with the same memory. It is suitably off of the bleeding edge to make it available at a good price, while offering more than enough raw horsepower to keep even the hungriest game well fed. It has often been said that you can't go wrong with an Intel. I'm not sure if that statement has always been true, but it is certainly true of the 2.53GHz Northwood, especially when coupled with the 850e chipset and PC 1066 Ram.
Previous page
