Intel vs AMD for hosting games

[HmC_RaNgEr]

Heres the problem. My friend has 2 computers side by side networked on 1 meg SDSL. His computer is a Thunderbird 1100, and his other is PIII 650. Both have roughly same Vid card, sound HDD, and RAM. This is concerning hosting a game called Rogue Spear. On his PIII machine, he can host the limit of 16 people with no lag. On his Thunderbird he can only handle 10 then it lags. He is convinced that the Thunderbird doesn't have the capabilities to be a hosting machine, because he has read that most servers use Intel because it can do more, I don't know what he means by that. Can anyone help me, so I can help him, so I can have piece of mind.


Cheers,
Ranger

[daverme]

This will not help you one bit but your tale is very reminiscent of an experience I had several years and two jobs ago while working for a software development company. I guy working for me had an Everex Step PC with a 486 DX2/50 CPU and right beside it a NEC 486 DX/33. We figured the DX2/50 should be faster, right? WRONG! The guy did this several times and determined beyond a shadow of a doubt that the DX/33 could run complies of "C" code faster than the DX2/50 ... and we never figured out why. As far as I can recall, memory was not an issue. The slower bus speed of the DX2/50 may have been a factor but 25 vs 33 hardly seemed enough difference to account for the radical time difference. We didn't know enough at the time to consider hard drive differences (we didn't know there were any!)

[Gutter Ball]

That's odd. I noticed the opposite! If I host a UT game with my Tbird 800, it doesn't lag with 10 players. But on my P3500, after 6 it starts to lag! Maybe ram has something to do with it? My tbird has 256 and my p3500 only has 128.

[daverme]

The statement that about using Intel in servers because they can "do more" might refer to the fact that, until recently, only Intel CPUs could be used in multi-CPU machines and many servers have multiple CPUs.

The difference your friend is seeing may be due to the differing architectures between AMD and Intel processors. Even though they can execute the same instructions, I'm pretty sure they do it differently within the processor chips. This means that each brand of processor will have quirks that make it run certain apps better than others, which is why different benchmarks will yield different results. This is true of the P4's; they don't do all that well with many current apps because they are architected differently from the P-IIIs.

How much cache is on each machine and where is it and how does the app use memory? That's partly a rhetorical question but the point is that the app's use of memory can be a factor. If the T'bird has less cache, it might have to refresh the cache more often than the P-III and that will make a HUGE difference in performance. Also, what is the caching capability of each mobo? Is each board actually able to completely cache all the memory it has?

[wedor]

They have the same amount of cache, most Intel servers would use the Xeon which has more cache.

[HmC_RaNgEr]

Bump

[daveleau]

Many apps are optimized for a certain processor. For instance, SiSoft and 3D Mark are optimized for Intel (or at least used to be, haven'tchecked recent versions). It is possible that the server app in question was built witht he Intel architecture in mind.

that could acct for a difference but not as much of a difference.

I bet it has something to do with also with the age of the OS on each system. If one system has been roughed up over time and the Intel is a newer install, then that could account for some packet loss... What OS are you using on each?

[HmC_RaNgEr]

Thanks daveleau, he has Win ME running on both systems.


Cheers,
Ranger.