Sweeeeeeet, new CPU to play with!

[swaaye]

I guess you don't play CS:S, that game is CPU bottlenecked even on my Q6600@3.4ghz.

http://techreport.com/articles.x/14573/5

As is typical with older games, you'd be best off with a ultra clocked single core in that retro engine. In that way, you can be "CPU limited" in Quake 3, too.

(1024x768 & 8800GTX)

Sweet CRT, btw. I miss CRTs. I still use my 19" Samsung 950P and prefer it in various ways over my 24" LCD.

But, with regard to CPUs and games, I personally tend to think of situations like these: (I'm not a serious online FPS gamer though)
(1024x768 & 8800GTX)

Here's the first benchmark of a game where I've seen a quad core matter. It's sort of synthetic though, because it's not an in-game situation. The game devs set up a test to show you that their multi-threaded skinning or animation engine (or whatever) actually works. Another test of the game showed flat CPU results like the above games.
http://techreport.com/articles.x/14573/4
(1024x768 & 8800GTX)

But yes you can be "CPU limited" in any game, assuming you get rid of the GPU limitation or have a old CPU. Older games show this best. But it becomes rather academic unless you have special requirements.

 

[shiznit]

My main point was that even 3.4ghz Core 2 was still bottlenecking CS:S since you said that you couldn't tell the difference between a 1.8ghz Core 2 and a 3.6 in games yet the difference in CS:S would be double (therefore I concluded that you don't play CS:S), not that I have a quad core which be both know makes no difference in a singlethreaded games, beside the obvious benefits of Nvidia's multi-threaded driver.

However TF2 is a different story. I wouldn't expect the tech report to benchmark any cpu properly for gaming since they never do. Nobody does actually, CS:S is the most popular cpu-bottlenecked game on Earth, other than WoW maybe, yet not a single site tests it when reviewing a cpu. Average framerates mean NOTHING when your connection to the server is bound by the worst case MINIMUM framerate, and in other games such as COD4 it is also used as a clock which governs many things such as jump and nade throw length, never mind hit registration, yet very few cpu review show minimum framerates. When I read a CPU review and see gaming benchmarks showing either GPU limited games or average instead of minimum framerates I dismiss it as garbage immediately and close the tab. So for future reference their so called 'benchmarks' carry no weight with me at all.

TF2 IS multithreaded if you bother to enable it which they obviously did not. Type mat_queue_mode 2 in console and the cpu usage will look very different, I average 50-70% total with all 4 cores used and the framerate goes up by a good 20-30%.

 

[Albuquerque]

I dunno about all that crap, but the E7200 is looking damned fine in those benchies.

As for CPU utilization benchmark reviews somehow being incorrectly done? I'd really like to know how you think that. CPU utilization has to be the pretty much EASIEST thing to "review" on the face of this green and blue planet.

Turn graphics features to minimum.
Record CPU usage with Perfmon
Play game
Record results

If there's more to it than this, I'd really be interested to know...

 

[shiznit]

I dunno about all that crap, but the E7200 is looking damned fine in those benchies.

As for CPU utilization benchmark reviews somehow being incorrectly done? I'd really like to know how you think that. CPU utilization has to be the pretty much EASIEST thing to "review" on the face of this green and blue planet.

Turn graphics features to minimum.
Record CPU usage with Perfmon
Play game
Record results

If there's more to it than this, I'd really be interested to know...

no thats pretty much all there is to it, but you have to test relevant games known to be cpu bottlenecked and show minimum not average framerates, it's very easy as you said yet nobody does it. Nobody cares if Lost Planed can use all 4 cores or not, or how a single player game like Bioshock runs at low settings when most people will run at GPU bottlenecked settings since 40+ fps is acceptable for single player for most people. But there are 200,000+ daily Counter-Strike players who would like to know which cpu to buy to maintain 100fps minimum so their connection doesn't suffer, and there are many more WoW players who would like to know if a cpu can handle a 25man raid with each person running 50+ mods all parsing data and communicating with each other.

 

[swaaye]

So for future reference their so called 'benchmarks' carry no weight with me at all.

Well, that's nice. Did you notice that they did post min framerates for most of the results? Just not your game engine. You could ask them why they didn't.

TF2 IS multithreaded if you bother to enable it which they obviously did not. Type mat_queue_mode 2 in console and the cpu usage will look very different, I average 50-70% total with all 4 cores used and the framerate goes up by a good 20-30%.

Sounds like something that's not official enough for 99% of people to know about.

 

[shiznit]

There are ~10 million WoW players, but you don't see them posting benchmarks of the game, do you? They probably don't see it as a realistic way to test a brand new CPU. Maybe that's wrong and you should tell them so.

I don't know. Your concerns really don't affect me personally though because I haven't played CS since around 2000 whereas I do care about SupCom and Bioshock.

fair enough, to each his own. I added something about WoW but I guess it was too late for you to see it before you responded, WoW is almost as bad as CS:S with many mods running. But are you really going to run Bioshock with stripped graphics so that the CPU becomes a problem, I don't think so. In competitive mp shooters everyone does though.

 

[swaaye]

Ooops, and I just deleted the post you quoted lol.

My original point was that CPUs don't usually affect performance much when you are GPU limited. That is a fact with most recent games. In your situation, things are different. It's just that not everyone is in your position and some don't even know it exists.

I remember, back around 2000 or so, when people were playing old school CS with GeForce 2 GTS cards at 800x600 to get the fastest frame rate possible. That's like an 8800GTX with CS:S - way overspecced. It makes sense in that kind of play. But that's just not what single player games need, or even most multiplayer games.....

 

[AlphaWolf]

fair enough, to each his own. I added something about WoW but I guess it was too late for you to see it before you responded, WoW is almost as bad as CS:S with many mods running. But are you really going to run Bioshock with stripped graphics so that the CPU becomes a problem, I don't think so. In competitive mp shooters everyone does though.

WoW isn't a twitch game, your ability to react is entirely limited by global cooldowns (1.5 seconds). Running it at 100fps is pretty much pointless.

 

[shiznit]

WoW isn't a twitch game, your ability to react is entirely limited by global cooldowns (1.5 seconds). Running it at 100fps is pretty much pointless.

True, but it would be nice to get 60fps, or to see the difference in reviews between different cpus instead of Lost Planet. Anyway, I have derailed this thead enough, my bad. I'll post results of the X3350 when I get it later this week.

 

[shiznit]

Ok I got the X3550, and I managed to get it running Prime stable at 425x8 for 3.4ghz using 1.275 vcore in bios (1.257 idle and load thanks to vdroop compensation in Asus bios), 1.55 PLL, 1.49v Northbridge, 1.38 VTT (aka FSB termination), 0.63x CPU GTL, and 0.67x Norbridge GTL. I am trying for 450fsb but even with northbridge at 1.55v and 1.40v VTT it hangs every time at the Vista loading screen. Intel has said that more than 1.40v VTT will kill these 45nm chips and I'm afraid to push more volts to the X38 chipset as I have no active cooling on it and the heatsink is already very hot to the touch at 1.49v. The previous owner was running this cpu at 3.6ghz 24/7 crunching for XS with 1.267 load vcore and 1.36 VTT measured (I'm guessing tops 1.4 in bios) so it seems that my X38 is stubborn and wants more voltage for 450fsb. Later this week I'm going to take the heatsink off and replace the stock goo with some MX-2 so I can give it more juice without overheating it. Any suggestions? Sorry again for derailing the thread earlier.

 

[Albuquerque]

My X38 needs zero extra voltage for 450x8. In fact, it needed zero extra voltage for 575FSB x 6 on my old E6850, along with my four sticks of ram and dual 3870's in crossfire. I have a feeling it's something else limiting you, perhaps ram (timings, volts, who knows) or GTLREF needs additional tweaking.

Of course, you could always try aiming a generic fan at your NB heatsink to see if a bit of extra cooling will help. I strapped a 50mm ball bearing fan onto my Gigabyte's NB heatsink and it solved some crossfire stability problems under heavy gaming.

 

[shiznit]

Auto setting for nothbridge is deceiving, you would think that you are not giving the X38 any extra voltage but what it's actually doing is over volting by itself, I wouldn't be surprised if Auto is unnecessarily pushing 1.6v on your northbridge, most people can reduce it below Auto values and maintain stability. If your bios has a list of detected voltages (usually in hardware monitor but I'm not familiar with Gigabyte bios) you can see this for your self. When I get the cooling situation sorted out I will give it 1.6v or more and try to hit 450fsb.

 

[Albuquerque]

I'm not sure what you're talking about, but if you're responding to me, please direct me to where I said ANYTHING is set to "auto".

Here's a hint: Nothing in my BIOS settings is set to automatic, not even SATA drive detection. MCH voltage is stock. MCH GTLREF threshold is stock. MCH VTT (termination) voltage is stock. CPU voltage is stock. CPU VTT (termination) voltage is stock. CPU GTLREF threshold is not stock. RAM voltage is not stock either. But RAM termination voltage is stock.

Hope that clears that up.

 

[shiznit]

I'm not sure what you're talking about, but if you're responding to me, please direct me to where I said ANYTHING is set to "auto".

Here's a hint: Nothing in my BIOS settings is set to automatic, not even SATA drive detection. MCH voltage is stock. MCH GTLREF threshold is stock. MCH VTT (termination) voltage is stock. CPU voltage is stock. CPU VTT (termination) voltage is stock. CPU GTLREF threshold is not stock. RAM voltage is not stock either. But RAM termination voltage is stock.

Hope that clears that up.

i definitely misunderstood you. guess my X38 is crappy, even 1.55 isn't enough for 450.

 

[Albuquerque]

Could also be the chip I suppose; I've seen people on G35's do 480FSB on an X3350. We all know it's luck of the draw afterall, but these puppies definitely have a pronounced "wall" at a certain point, no questions asked.

Might try fiddling with the GTLREF some more, both CPU and MCH if you have the option. CPU GTLREF made a significant difference in the stability of my chip in relation to voltage -- without adjustment, it needed about ~1.33v (stock = 1.2375) to be stable at 3.6Ghz. After tweaking though, it came down to stock volts.

 

[Tim Murray]

Well, still no processor! Goddamn FedEx and goddamn MicroCenter have conspired against me to lose my package and refuse to send me a new one until the FedEx trace is complete. Except FedEx didn't even attempt to deliver the CPU in the first place (lost in Ohio, I'm in Pittsburgh), so I don't know why I have to wait for them...

argh. aaaaaaaargh.

 

[Albuquerque]

That does suck; especially having to wait when it never even made it to your state...

 

[Tim Murray]

That does suck; especially having to wait when it never even made it to your state...

If Newegg had a Q9450 or an X3350 in stock, I would have done a chargeback already. I'm going to call again in a minute (after they said they would call me back immediately when I called this morning and what they also said when I called yesterday...), and man SOMETHING better happen.

 

[Geo]

I've got a Q9300 running at 3.0 GHz (i.e. 400 fsb) in since last night, with 8GB of 1066 DDR2 to go with it. So far so good.

And I took out ALL my PATA/floppy cables, my floppy drive, my PATA DVD burner (have a SATA DVD burner), my X-Fi (sucks azz in x64), and a massive Zalman heatsink/fan, which these 45nm Intel don't really need for a reasonable OC.

You could play tennis in my case now with the 8800GTX as the net. . .

Edit: And, no, I didn't push the OC. I'll take 20% stable at low temps and be happy every time. Tho if I get bored later I may push a little harder without juicing. I like cool and quiet.

 

[Tim Murray]

Canceled my order with Micro Center (NOW they don't have any in stock, either, so who the fuck KNOWS when they would ship me a new one) and just ordered a retail Q9300. The extra cache won't do me too much good anymore (wanted it for a paper I'm working on, except the due date will be here before I would have the processor), and I can overclock the snot out of it.