SMP. Worth it for gaming?

[james.miller]

well. You may or may not no my opinion on this. Regardless, i'd like to hear yours but please, back it up with some real evidence if you can.


Let's settle this once and for all

 

[Trotterwatch]

The moment this turns into a flamefest this gets shut too, so everyone try acting like an adult this time.

 

[james.miller]

yes indeed. sorry, i should have added that. All i want is some good hard evidence to prove either way. Let's keep it friendly guys

 

[Remote]

Afaik most games doesn't take advantage of the second cpu so it's could be a terrible waste of money, since you need an extra heatsink etc for the second cpu... There are however a couple of games that though but if I'm not mistaken the gain in speed / whatever you'll get isn't that major.. A couple of percent, it's no magical worker.. ;-)

So, perhaps a 64bit cpu could be worth taking into the discussion instead..

 

[james.miller]

definatly. That is exactly my view.


There are very few games that support SMP anyway. I posted this list in another thread:
http://www.geocities.com/a_noms/SMPgamelist.html

Games that support SMP

Alice (rumored to be multithreaded - uses QIII Engine)
Battle of Britain (rumored)
Falcon 4 (not very efficient handling SMP)
IL-2 Sturmovik (not very efficient handling of SMP)
Mig Alley (rumored)
Quake III Arena
Return to Castle Wolfenstein
Starsiege
Star Wars Jedi Knight II: Jedi Outcast
Star Trek: Elite Force
Total Club Manager 2003

the reply i got wasn't exactly the nicest of replies, something along the lines of "thats a bs list and its on fucking geocitys".

Anyway it seems to be pretty comprehensive. One thing that was said to me was that UT2k3 wasnt on that list. The reason i believe, is that UT2K3 only uses a second process to process the sound. It doesnt do anything else with SMP, therfore it isnt really a true SMP game if it only uses 5-10% of the second cpu....... That hardly makes SMP worthwhile.

I also agree that a 64bit is the way to go for gaming. I'd take an a64 3000+ over a HT p4 or an SMP system any day of the week. But there are others that disagree with me on these boards it would seem........

 

[Hacktarux]

I don't think you can get a much higher frame rate in games with SMP machines. But with a SMP machine, you can do several tasks in the background while still being able to play very CPU intensive games. I think it's very great to play some games while you're waiting untill you video has finished to encode or any other long task you have started.

On the other hand, if you have enough money to buy a SMP machine, you probably have enough to buy an 64 bits one. And knowing which one is better to buy right now is a good question. I think i'd prefer buying an Athlon64 because it's the technology that'll be used in the next few years, but maybe not right now as the socket will change in the next few months from what i've heard.

I'm planning to buy a new processor but i think i'll wait a few months. It's also hard to know how the price of the mobos/processors will evolve during these months, especially with new intel x86-64 processor. BTW, anyone remember when it's supposed to be out ?

 

[james.miller]

There's no doubt that the a64's are faster than anything intel has right now, but wether you buy a 64bit or an SMP system depends on what you do with your pc. If you are heavily into encoding then SMP would really be the logical choice.

however, for pure gaming alone the A64 is really a no-brainer.

 

[Kaoss626]

Having 2 SMP machines and having played numerous games , I'll toss in my opinions

Quake III - don't bother, SMP almost but not quite slowed down my machine. If I could get it to run in SMP without crashing

Alice - see Quake III (same game engine)

Return to Castle Wolfenstein - see Quake III (same game engine)

UT2k3 - dunno, I never knew this game had any SMP support even if it was minimal such as running sound as a second thread. I suppose I could install it again and fiddle around with it but I hardly think that it will make much of a difference

Straight up, IMHO game companies do not concider SMP "mainstream" enough to be troubled to make a game that really utilizes SMP properly. Maybe now with HyperThreading technology they will start to make multithreaded games, but for now I highly doubt it. More likely it will take a year or two before they do.

Is there any performance boost at all (for Unithreaded games)? Yes, but most people would concider it negligable. I can run a Unithreaded game and have it completely hog one of my processors and the second processor is still free to handle mundane OS tasks, handle sound, video, keyboard, USB and other drivers that are all needing a little process time as well. Its probably a 2% gain at most.

So why have a SMP game machine? For me it comes down to smoothness and the ability to do much more than just play a game. If a Uniprocessor machine was trying to play Quake III and burn a CD, forget it, you're going to burn a coaster. Quake III will completely dominate the processor and the burning software will never get the chance to refill the buffer for the CDR resulting in a buffer underrun. I can play Quake III, burn a CD, D/L 3 Linux ISOs, run a virus scan and run a distributed process client maintaining both processors at 100% load concurrently. For me with all that, not one hickup, not one error, and a flawless burn of that CD. For the few games that can handle Alt-Tab and not crash (UT for example) on a Uniprocessor machine hitting Alt-Tab will get you to the desktop in 10-15 seconds maybe longer, Dualprocessor machine 2-5 seconds. Task switching is simply smoother.

64 bit the way to go? Maybe. If game companies actually thought that it would be worth their time (time is money) either one SMP or 64 bit could both be a boost. As it is though 64 bit machines are steadily entering the market. I doubt you will see a single 64 bit game before you see 64 bit version of windows. If you want 64 bit right now you run linux and most game companies do not see linux as a viable platform for games. IMHO you are going to have to wait a year or two to see either implimented in games.

64 bit SMP - don't forget this, Its already here. But again, the game companies will have to see this as a viable option.

-edit wow... it took me so long to type this, that there were two replies just while I was typing

 

[james.miller]

great reply, and some first had experiance thrown in awell. nice

So again, we see that it all boils down to preferance. SMP doesnt do anything for gaming - but it does allow you to do other things at the same time.

Kaoss626 what you have to remember about AMD's 64bit cpus is that, although 64bit gamiing or even software in general is nowhere to be seen, they are still very much faster in 32bit opps that anything intel have got.

I've forgottenm the name of the forums, but i read a huge thread started by a guy called yuri. He had a midly overclock HT p4c (3.8ghz lol - just a little overclocked) and he also had an a64 3000+ based pc.

He did an experiment to see which was faster, and he used the same ATI r9800pro in each rig. With his a64 3000+ overclocked to 2.2ghz (200mhz over stock) , his p4c NEEDED to run at 3.8ghz just to keep up with the a64, and that in 32bit opp.

Now if that isnt impressive, i dont know what is lol.

 

[jollyrancher]

This sorta gets back to an earlier discussion about frame rates, but any good new system should be able to play any PC game at over 80 FPS, and since anything over that is overkill (eyes can't tell the difference) there's really no need to get something like that unless you're a professional and you're doing some hardcore computer work.

 

[Kaoss626]

clock for clock anything would smoke a P4..... well maybe not a cyrex... but they just smoke.

-edit (sorry had to do it) (giggling) I still remember stories about dual and quad processor cyrex powerd BBQs. edit-

seriously

yes A64s are faster clock for clock. So were K6-2s (VS the pentium 2) A lot of compairisons seem really to be apples to oranges. Thats why I say clock to clock. Clock to clock, my P IIIs would eat a P4 and have room left for more. As you said before it all comes down to what a user wants out of their machine. If they want the fastest rig out there right now, A64 is the way to go. Maybe that will change when Intel gets into the 64 bit market (but I doubt it).

SMP is like a pear compaired to those apples and oranges. And it does not come with out some drawbacks as well. Cost. A Mainboard costs 50-100% more to get a SMP board. And then you have to buy 2 processors to put in it, not just 1. HEAT. The main reason I went PIII vice Athlon MP. 2 40 watt processors Vs processors that put out 70-80 watts each. Power supply. Some power supplies do just fine on that uniprocessor machine, but will completely fail to maintain voltage or stability under the strain of keeping 2 processors happy.

I got into SMP for the reasons I gave in my last post, and a few others. 3D rendering was just one of them. On a single processor, 3D StudioMax took 5 minutes to render 1 frame. On a dually I could do 2 in the same time. on 2 duallies, I could do 4. (3D studioMax is one of the rare programs that would scale very well.) Tho I had a huge advantage, at the time I was working in a computer store and got my parts at wholesale $125 for the board vice $200. $200 each for the processors vice $300.

I'm just throwing in my experiance with SMP itself. Your are right about the A64s being faster. But again one must compair the apples and oranges equaly (which makes the A64s look even faster). There are a lot of delusions about SMP and what it can and cannot do. I went into SMP with my eyes open, and I wasn't dissapointed. But a lot of people think that SMP can be a big help and in reality it does not help them do what they want of it. Case in point a guy came into the computer store and wanted his game machine upgraded. He had told a major computer chain starting with a "G" that he wanted the fastest thing he could buy money not an option they sold him a dual Xeon for $5000 instead of a ordinary desktop box for $1500. He just wasn't seeing the benfits of SMP and for what he did I doubt he ever would.

 

[Kaoss626]

LOL again someone posts while I'm typing

jollyrancher... LOL 80 FPS? maybe these days. I recall when maintaining 30 FPS in glQuake was neigh on impossible and the average game struggled to maintain more than 20 FPS on the "average" machine

 

[Hacktarux]

Well it's not really compared clock to clock, otherwise intel performances would be very ridiculous.

AMD AthlonXP 3200 is clocked at 2GHz and is usually compared to the P4 3.2GHz....

 

[james.miller]

i wasnt talking clock for clock at all. Im talking overall, regardless of clockspeeds. what im saying is, the a64 at 200mhz over stock was running neck and neck with a p4c overclocked to 3.8ghz in almost every game he tried.


Regardless of clockspeed, that is a hell of a feat. An 64 which can be bought for around £160 running with a 3.8ghz p4. Intel dont even make 3.8ghz p4's. the fastest they have is a 3.2c, which costs nearly £300.......

but this is a little off topic (i think :blink: )......

Well it's not really compared clock to clock, otherwise intel performances would be very ridiculous.

AMD AthlonXP 3200 is clocked at 2GHz and is usually compared to the P4 3.2GHz....

the barton 3200+ sadly doesnt compare at all to 3.2p4, the a64 3200+ on the other hand trounces it.

 

[Kaoss626]

Oh yea... I do remember one major gaming advantage for SMP.

I could run a dedicated server for UT on my machine. Not a big challenge for a Uniprocessor machine. When I wanted to play some UT with my buddies who were logged into the Dedicated server, I'd just start up UT and play. For a uniprocessor machine, trying to run the game and the dedicated server would have smoke coming out of its ears. My machine would handle this flawlessly.

 

[sheik124]

/me rethinks the possibilities of dual FX-51

 

[AlphaWolf]

Ok, here are my own personal tests and conclusions:

Downstairs, I have a p4 that supports hyperthreading (not I nor my brother are not much for intel, but this motherboard/cpu was extremely cheap.) I was doing my own personal tests in comparing games with HT disabled vs HT enabled.

I tested several games (not q3 however, I honestly hate that game,) to include the 4 I mentioned in the other thread (wanted to see the overall speedup due to the fact that SMP should ideally give a neglegible performance increase to the system abroad so long as the kernel supports it, so I even tested non SMP games as well.)

Now, I don't have actual numbers, but having HT enabled vs not enabled gave a performance increase comparable to that of conservatively overclocking a video card. Overall it was noticeable, and a worthwhile thing to have. RTCW saw an overall performance that was better than the rest.

The game that really piqued my interest however was UT2k3. The overall framerate wasn't the big improvement, the big improvement was that there were never any noticeable framerate drops with HT enabled, whereas with it disabled it happened frequently when there was a lot of action on the screen. Prior to doing this, I had no idea that UT2k3 had any multithreading, but after a search I found that it does.

Now, granted these aren't end all to be all speed improvements, I think SMP gaming stands as much a chance of becoming as common as 64-bit gaming. As it currently stands today, SMP is already better for gaming than 64-bit, because at least there are games that support it. Albeit SMP runs a tad (probably less than .01%) slower on single processor systems, this is the total opposite with a single hyperthreading CPU.

In terms of the future of gaming: if AMD chips supported hyperthreading (thus everybody would have it) then SMP would be nice in the fact that if you wanted to have more performance, you could just add more processors (assuming the motherboards of the future could work that way, which isn't far fetched really.)

To answer the question posed by the subject: yes, SMP is definitely worth it. Having dual CPUs on the other hand, maybe not because the cost still outweighs the benefits quite a bit, but SMP is definitely a good thing.

 

[james.miller]

RTCW uses the q3 engine. Other people have said there is no difference. Could you possibly do some benchmarks?

BTW, SMP and HT are not the same thing, just in case some people wernt sure.

edit:

HERE'S SOME SOLID EVIDENCE
http://www.vr-zone.com/reviews/Intel/P43060/page8.htm

That shows no benefit from HT whatsoever, which is totally the opposite to what alphawolf has been saying. It also shows that HT Actually DROPS the performance slightly, which is exactly what i told alphawolf in PM, but aparently i was "wrong".

So at this point, is SMP really worth it? does it make a difference in games? (a positive difference lol)

Which would you recommend for a gaming-only system? single cpu, or SMP/HT?

 

[AlphaWolf]

BTW, SMP and HT are not the same thing, just in case some people wernt sure.

That isn't accurate...that is just the same as saying SMP and dual processors are not the same thing. One complements the other, and when one isn't present, the other is useless (well, nearly useless, in a dual CPU or HT environment, a single application would gain no benefits, but background threads would.)

HERE'S SOME SOLID EVIDENCE
http://www.vr-zone.com/reviews/Intel/P43060/page8.htm

That shows no benefit from HT whatsoever, which is totally the opposite to what alphawolf has been saying. It also shows that HT Actually DROPS the performance slightly, which is exactly what i told alphawolf in PM, but aparently i was "wrong".

Well, it looks like their tests weren't thorough enough (probably why that site is still a bit of a noname.) According to this page:

Here, in the normal mode, there is no difference in the scores between that of a HT-enabled CPU and non-HT-enabled one. But the difference was noticeable in the high and max quality modes.

So there you have it from a source other than myself.

I run all of my games in their highest possible graphics options, all the time, so in this case that would explain why my results differ from that of your source. Also, if you read my earlier post, note that I said in UT2k3 there werent any differences in framerates, other than the fact that during high loads, the FPS drops weren't noticeable - ever, whereas they were noticeable without HT. Your source of information probably didn't ever look at this, they probably just ran a simple timedemo, which wouldn't tell you anything of that nature.

Oh, and by the way, if you want something more than a noname computer hardware reviewing source, read what thomshardware has to say, which again answers the question posed by the subject of the thread.

 

[james.miller]

toms hardware is a bs site, and thats a fact. Nothing from toms is accurate.just check out their reviews of nvidia cards, or their recent LN2 overclock of their p4 where they quickly removed the performance charts because the fx-51 was STILL giving it a beating, or their video of the cpu's when the heatsinks are taken off.

VR-ZONE may be less known to YOU, but their results are far more accurate than toms Tom's hardware is BS.

where was that quote taken from, because it isnt on the games benchmark page...

I run all of my games in their highest possible graphics options, all the time, so in this case that would explain why my results differ from that of your source. Also, if you read my earlier post, note that I said in UT2k3 there werent any differences in framerates, other than the fact that during high loads, the FPS drops weren't noticeable - ever, whereas they were noticeable without HT

If the minimum frame-rate was higher, then the average frame rate would also be higher wouldnt it.

A little SMP/HT lesson.

SMP stands for Symetrical Multi Processing. SMP is where you have two physical cpu's and the operating system distributes tasks between them

HT stands for HyperThreading. Similar to SMP, except there is one one physical cpu, and one "logical" cpu. Tasks are shared in much the same way, but is more widley supported by the various operating systems (SMP is only support by winxp pro and win200 AFAIK). Also, HT cpu's aren't as efficiant as a true SMP alternative because it is still only 1 cpu at the end of the day.

....so yes, they are different.

AlphaWolf you have still yet to provide any evidence to prove your opinion, and untill you do it i just that - an opinion, nothing more.

meanwhile, here's some evidence to prove you wrong:

http://www.hardware-unlimited.com/reviews/pentium4_3ghz/index13.shtml
Ut2k3 - no difference, or worse[/i] performance with HT enabled with games.

read through the pages - its quite interesting. more soon......