Does CPU Architecture really Change Performance?

[mallenwho]

I have two computers. Both computers have the same 560ti graphics card, both run an SSD boot drive, both have 8gb memory and both are in the same case with similar thermals.

 

The difference is the CPU (and motherboard but that shouldn't affect performance)

 

One system runs an overclocked Sandybridge 2600K (hyperthreading off) while the other runs a Haswell 4770 (hyperthreading on).

Two generations appart from each other, while both being 3.4gHz i7 processors

 

However the computer with the Sandybridge processor is severely underpowered it appears compared to the Haswell computer.

 

For example, in Space Engineers both can run a single player world just fine (sandybridge running at 90-100% CPU usage while haswell running at 60-80%) but when multiplayer is introduced, specifically other playes moving around in ships, the workload increases (Sandybridge sitting at 100% the whole time with about 2 FPS while Haswell sits at 90-100%)

 

Similar performance issues can be noticed in late game Game Dev Tycoon (a large studio with all sorts of events happening). Frame rate drops to about 2 FPS on Sandybridge (don't have it on Haswell) and because in-game clicking is tied to frame rate makes the game close to unplayable. Re-loading the game helps in the short term. Not sure why it does that.

 

Late game Prison Architect (with 500 prisoners and appropriate staff to match) on Sandybridge runs at 4FPS. Similar large game on Haswell runs at about 80% usage.

 

Simcity 2013 (any city size over 40K people) will run at 4FPS on Sandybridge while smooth on Haswell up to about 120K

 

All of those games are multithreaded workloads. The Haswell system also runs Windows 8 vs Win. 7 on Sandybridge but i don't think that should explain the performance gap.

 

My question is this:

 

Does CPU architecture explain this change or is another variable bottlenecking my Sandybridge system?

 

in some of these games (SC 2013, Space engineers) GPU also sits at 100% usage, however it is the same for both systems with the same model of graphics card (560ti Asus twin frozer cooler)

 

also, I imagine upgrading to haswell or Devil's Canyon would match performance but would buying a xeon 6 core processor or waiting til Q3 for Intel's new 6 core thing improve more so? is it worth the extra money in this instance to upgrade beyond gaming hardware into enthusiast/extreme?

 

Thank you. =)

 

Any assistance

[Nineshadow]

Of course.

Why do you think they are different architectures?

[jmaster299]

There is no reason to disable HT. Even if the games you play can't use it, Windows can. By leaving it on it can shift Windows processes that support it off the 0 or 1 cores that the games use most often. Also, games dropping frames is usually caused by running out of VRAM. Reaching the processing power limit of a GPU doesn't do anything but cap your max FPS, but maxing out the frame buffer will cause your frame rate to drop to nothing.

[Kyuubixchidori]

you should not be getting that low of FPS. something else is wrong. 

 

my I5-3570K only stutters when we crash very large ships together. normal flying with others players is fine, even on max settings. 

 

even my brother with a 6 year old pentium and  gt 460, it wasn't the best thing in the world, but we could fly around and it would be OKAY, but way more 2 fps.  I think something else is wrong with that computer. 

[mallenwho]

I doubt VRAM is the issue with space engineers. The frame rate goes up again when my friends stop moving on their ships. It could explain games like simcity and game dev tycoon.

I can't think of other problems that could be caused. Perhaps networking with other computer in space engineers but I doubt that would cause the same sort of issues I've been having.

For a game like space engineers (specifically space engineers) would it be able to utilise 6 cores if I have it 6?

[jmaster299]

I doubt VRAM is the issue with space engineers. The frame rate goes up again when my friends stop moving on their ships. It could explain games like simcity and game dev tycoon.

I can't think of other problems that could be caused. Perhaps networking with other computer in space engineers but I doubt that would cause the same sort of issues I've been having.

For a game like space engineers (specifically space engineers) would it be able to utilise 6 cores if I have it 6?

What's the OS? Windows 8 has a horrible memory leak issue that affects users at random. Meaning it does not affect everyone.

[Kyuubixchidori]

I doubt VRAM is the issue with space engineers. The frame rate goes up again when my friends stop moving on their ships. It could explain games like simcity and game dev tycoon.

I can't think of other problems that could be caused. Perhaps networking with other computer in space engineers but I doubt that would cause the same sort of issues I've been having.

For a game like space engineers (specifically space engineers) would it be able to utilise 6 cores if I have it 6?

For space engineers, I doubt it. but the other cores will be left for background programs and windows, allowing you to have more power for space engineers. If you want ill do some testing on space engineers for you. it seems like a GPU bound game though. I have a I5-3570K, and a HD 7950. hosting a small game with me and my brother, with V-sync off, it maxes out my GPU. I'll see what the cpu is at if you want. 

 

 

edit: so I5 3570K, HD 7950 OCd, and windows 8.1. max settings, largest world, hosting for 3 people, all of us derping around in the spawn ships, uses 20-25% CPU. gpu is at 35-40%, never broke 41%.

 

v-sync off, same map and everything, my gpu maxes out, and while my gpu usage doubles, my cpu usage only goes up to 40%.

 

note, the cpu usage is just the game, nothing else. as i have chrome open, and a feed the beast minecraft server running

[patrickjp93]

Each new architecture implements new instruction sets which tend to bundle older, simpler instructions together in less total clock cycles (but sometimes they're completely unique), increasing instruction per cycle count and thus increasing efficiency in the right situations. Sometimes new architectures manage to lower clock counts for older instructions too. But, of late, the main goal is upping clock speed (since applications don't need more than 4 cores unless you're on a workstation), lowering power consumption, integrating better onboard graphics solutions (Intel rattling sabres against Nvidia and AMD both for cheaper low to mid-range graphics solutions puts more pressure on AMD on two fronts), and adding cache which helps in any application to speed up the process (which tends to produce more heat).

 

Bottom line, unless you're AMD which went after a less efficient but higher clock rate architecture in the FX Series (to chase the 5 GHz barrier), each new architecture means more performance than the last in some way.

[Catmanistheman]

Well if you wanted to test this further try turning hyper threading off on the 4770.