Upgrading from Duo to Quad

[Pteradon]

Hi,

 

Im really enjoy DCS: A-10 but i like to have some more fps / smooth gameplay.

 

The best upgrade would be i7 system, but i would be to expensive for me right now.

Therefore i'm thinking to buy a quad core cpu, the Q9650.

But i'm not sure if this will preform any better in DCS: A-10 then the

Duo E8600 cpu (oc'ed to 4.0 Ghz)

 

Would it be a waste of money or will it run DCS:A-10 better then the cpu

im using now?

[EtherealN]

I'd say switching from e8600 to q9650 would be a waste. It would only increase your performance if you have a lot of background processes happening, otherwise it will reduce your performance unless you manage to overclock it to 4GHz as well - in which case it would be roughly equal to the e8500.

 

They are also relatively expensive. I would then much rather propose saving up for an i7 2500K (~250 dollars), an acceptable mobo (~150 to 180) and some DDR3 RAM (~100 dollars for an 8GB kit).

[EtherealN]

Something like this:

http://komplett.nl/Komplett/product/ZKB_01COM/08_MEMINT/productdetails/15491303/Corsair_XMS3_8GB_1333MHz_PC3_10600/CMX8GX3M4A1333C9/default.aspx

 

http://komplett.nl/Komplett/product/ZKB_01COM/15_MOBO/02_INTEL/productdetails/20078122/ASUS_P8P67_Pro_B3_Socket_1155_ATX/90_MIBE5A_G0EAY0KZ/default.aspx

 

http://komplett.nl/Komplett/product/ZKB_01COM/21_CPU/productdetails/20070433/Intel_Core_i5_2500K_3_3_Ghz_6MB_S1155_Unlocked/BX80623I52500K/default.aspx

 

There are cheaper alternatives, of course, that would still give you an improvement, but with the above you'd land at €408. Note that the Q processor costs pretty much just as much as the 2500K, but the 2500K is vastly superior in all respects.

[Rainmaker]

@Pteradon

 

Please read through this thread.

 

http://forums.eagle.ru/showthread.php?t=70983

 

Most of use that are on the E series Wolfdale chips are all in the same boat.

[celticcoho]

I975 chip, if you can find one, A Gem!!They moved so fast to the 980, they forgot how good the 975 is...Google it.. Cheers, I have one, Love it!!:thumbup: But the New Sandies are cheaper, NOT!!! You need new MOBO Too.. Then must consider your GPU's for MB...Stay Strong or go Sandies for whole new build. They are nice, If you can rebuild! memory to 2 instead of 3 too,,so, Upgrade or Rebuild...I am good with the I7!!!!

Edited April 7, 2011 by celticcoho

[EtherealN]

But the New Sandies are cheaper, NOT!!! You need new MOBO Too..

 

Dude, did you even read the original post? He's on a 775 system. His only chance of not switching mobo is to stay Core2, like a Q-series chip. (Which is a no-go IMO.)

 

So:

i7 975: 1049 dollars (I only found the extreme in stock anywhere)

mobo: ~200 dollars

ram: ~150 dollars

total: 1399 dollars

 

i5 2500K: 224 dollars

mobo: ~200 dolalrs

ram: ~150 dollars

total: 574 dollars

 

Some performance data from Techreport:

Hm. Not impressed.

 

Still not impressed.

 

Indeed, we have to go as low as the i7 950 to compete with the 2500K in price (and it's still more expensive), and as you can see: the 950 doesn't perform well in that comparison. And if we take something purposebuilt for just stressing the processor...

 

...they're still pretty much dead even. In spite of the 2500K not having HT, which that application will make use of, and which the 950 does have.

 

Well done. ;)

Edited April 7, 2011 by EtherealN

[Sinky]

Thought I'd share my experience with you. I recently unlocked my Phenom II X2 555 BE so it's now a quad core and I haven't seen any change in DCS. However in other games that take advantage of quad cores I've seen my fps almost double and I'm still running at the same clock speed as before.

 

Personally I don't think you will see a major difference in A-10C if you went for an equivalently clocked quad core but that's for you to decide since I'm no expert. Looking at what I've read around the forum, it's clock speed that really makes the difference in A-10C.

[Apples]

I have a core 2 quad 9650 overclocked, Its been my favorite processor so far.

[EtherealN]

Looking at what I've read around the forum, it's clock speed that really makes the difference in A-10C.

 

Not strictly true: what makes the difference is processing power per core. Newer architectures will do more on less clock cycles. (Otherwise one of the old 2005 P4's would have been better than the C2D's back when DCS:BS was first released.) The clock speed is just a clock, keeps the pace as the various parts of the processor chugs along. But a lot depends on what actually happens in each clock cycle - a typical modern processor will be able to perform several operations in each clock through simultaneously loading several execution units. A newer processor architecture will typically be able to either do more of these on each clock, or do them more efficiently through better branch prediction and so on, or be able to make use of "empty" units in a given clock cycle through inserting a second thread in there (Hyperthreading).

 

So the "higher clock is faster DCS" comparison only works within the same architecture. A 3GHz Nehalem will be faster than a 3GHz Core, and a 3GHz Sandy Bridge will be faster than both. (And Intel claims Ivy Bridge will be ~20% faster than Sandy Bridge on a per-clock comparison, but with no samples it's hard to say under which conditions that is true.)

 

(A more dramatic example of this difference can be achieved through underclocking a dual-core desktop processor and comparing it's performance to a dual-core Atom running at the same speed.)

[Sinky]

Not strictly true: what makes the difference is processing power per core. Newer architectures will do more on less clock cycles. (Otherwise one of the old 2005 P4's would have been better than the C2D's back when DCS:BS was first released.) The clock speed is just a clock, keeps the pace as the various parts of the processor chugs along. But a lot depends on what actually happens in each clock cycle - a typical modern processor will be able to perform several operations in each clock through simultaneously loading several execution units. A newer processor architecture will typically be able to either do more of these on each clock, or do them more efficiently through better branch prediction and so on, or be able to make use of "empty" units in a given clock cycle through inserting a second thread in there (Hyperthreading).

 

So the "higher clock is faster DCS" comparison only works within the same architecture. A 3GHz Nehalem will be faster than a 3GHz Core, and a 3GHz Sandy Bridge will be faster than both. (And Intel claims Ivy Bridge will be ~20% faster than Sandy Bridge on a per-clock comparison, but with no samples it's hard to say under which conditions that is true.)

 

(A more dramatic example of this difference can be achieved through underclocking a dual-core desktop processor and comparing it's performance to a dual-core Atom running at the same speed.)

 

I think I see what you mean. This is probably why I haven't seen a change in DCS performance from unlocking since it's the same architecture and the fact it only takes advantage of two cores.

[EtherealN]

Correct. The advantage of Nehalem or Sandy Bridge vis a vis Core (I've been Intel-centric on all my machines for the last 5-ish years so I'll use them as examples) isn't only the step up from dual-core model to quad-core (note: most C2Q's are actually two C2D's in the same chip, not true quadcores), it's that each of those cores does more computation per unit of time. This could be because it runs "faster" (higher clock), but for the last 5 years this hasn't quite been an option due to some thermodynamic and electrical particulars, or it can be because during each clock the processor does more.

 

You could for example imagine a processor that will only ever do one operation per clock cycle, and another that does two. As long as the code so allows, the second would equal the first in performance even if running at half speed, or double if running at the same speed.

 

It does get a whole bit more complex than that, of course - for example there is variability in how many actual operations are required for doing a specific thing. Fishing up an example from the Folding@Home FAQ as an example: on a CUDA execution core doing a square root requires one operation, while on a bog-standard x86 core it would require no less than 15 operations. (This might vary though: I don't know the numbers for actual current x86 implementations.) So if all you needed your processor to do was pull square roots, a 100 MHz CUDA core would be as fast as a 1.5GHz x86 - assuming the x86 has only one arithmetic unit.

 

This all makes it even harder to predict comparative performance between different architectures, especially when comparing between AMD and Intel - the AMD product might be better at one type of calculation, the Intel at another, and which one comes ahead would end up depending heavily on what kind of program one is running.

 

This is all a very hard balancing act that they're doing when they design the chips - they can't just cram them full of arithmetic units and make all of them able to do all types of operations in a single instruction: the units would become huge, and since a lot of processing has data dependencies (can't do operation 2 until you know the result of operation 1) they might end up with a lot of die area just sucking power and doing nothing at any given time. (That, incidentally, is largely what Hyperthreading is about - it's letting the system use such "free" units. Which is also why HT can sometimes actually decrease performance: if your intensive program really needs all of it, but the OS schedules some other stuff to the other "logical" core caused by the HT implementation, things slow down.)

 

And gods, now you guys got me all nerding out again. I'm off for some coffee. :P

 

One note though on a sideways advantage you specifically might get from that unlocking though: you don't need to worry as much about background processes. Say you want to FRAPS - on two cores the FRAPS encoding would have to share cores with your game. On quad-core, it doesn't. Same with Windows, antivirus or whatever else is running. A pretty cool implementation I've seen of this is Starcraft 2 commentators playing the game/replay, FRAPSing it, and encoding the previous run in Vegas all at the same time as they are livestreaming. Having those extra cores won't help your Starcraft much - but you can be reasonably certain that general OS/background tasks won't impede your performance. In some cases, this in itself can be a huge convenience - I personally usually don't care about shutting down iTunes, web browsers and all of those things when playing, since it usually doesn't matter. Same for running streaming TV on the other screen - it can sit in the other two cores and just be happy. Non-trivial things for me, and I love having that capability, but that's a personal preference/habit thing.

 

...gods I wrote an essay! :D

Edited April 7, 2011 by EtherealN
fixing some minor errors

[Sinky]

This all makes it even harder to predict comparative performance between different architectures, especially when comparing between AMD and Intel - the AMD product might be better at one type of calculation, the Intel at another, and which one comes ahead would end up depending heavily on what kind of program one is running.

 

You've jogged my memory. I've experienced this when I upgraded from an old AMD Athlon 64 X2 6400+ 3.2 to my current AMD Phenom II X2 3.2 BE before it was unlocked. I got a small boost from doing that, it would have been more noticeable I think if my old 8800GT wasn't bottlenecking it at the time.

 

As for the rest of your post, I'm lost :smilewink:.

[beers]

Thank you Professor EtherealN!! You are an excellent nerd, wonderfully related information!

[Pteradon]

Thank you all guys. :thumbup:

 

@EtherealN - Thank you for your help and the in-depth explanation about the difference of the cpu's. :thumbsup:

[Pyroflash]

Ethereal, I am coming to hate you now. Every time I read something about the Sandy Bridge chips, it makes me have E-Peen envy over the fact that the Ivy Bridge chips aren't out yet... Just you people wait though, I will be sitting nice and pretty with my new $1000 processor and its triple channel RAM, while you are still stuck with ol' Sandy :)

 

P.S. Unless you upgrade again; which in that case it doesn't matter because I will be having too much fun with my new processor to care.

 

P.P.S. On another note, I can't give you any rep for your article, even though anyone who writes a :tldr: worthy essay probably deserves it.

Edited April 8, 2011 by Pyroflash

[EtherealN]

Yeah, I'm torn about ivy and the 1366 replacement architectures. Up to 40 lanes of PCIe directly from the CPU and so many other nice boosts. Remains to be seen if they'll be worth the upgrade when coming from a 2600K, but I do expect them to be tempting. I guess in the end it'll mainly be a question about which pricetags they end up with, but they'll definitely be a nice candidate to replace your 975. :)