Dual core (E8600) vs quad core (Q9550) for .NET development

[Evgeny]

Yes, this is yet another dual vs quad core thread. I'm building a custom machine for work, which is programming in Visual Studio (ASP.NET particularly). The main tasks that need to be fast are compiling and debugging C# code, including lots of little compilations that ASP.NET does in the background.

C# compilations are pretty much single threaded. VS can in theory compile multiple projects at the same time, but I've never seen much of a performance benefit from this in practice. What makes me even consider quad core is the extra cache. As I understand, a single core can use all of the 12MB cache if necessary. Is that correct and will it make a significant difference - more significant than the extra clock speed of the E8600?

I'm planning to overclock this system, but not to the extreme and I'd imagine the dual core will still reach a higher clock speed.

 

[myocardia]

Originally posted by: Evgeny
As I understand, a single core can use all of the 12MB cache if necessary. Is that correct and will it make a significant difference - more significant than the extra clock speed of the E8600?

No, that's not how it works. With Intel's CPUs, each dual-core can use up to 100% of the shared L2 cache. Since all of Intel's quad-cores are two dual-cores, the most any one core of a quad can use would be half of the total L2 cache. So, in single-threaded apps, a Q9550 will definitely be slower than the E8600, and by quite alot in some instances. BTW, I would guess that the reason you weren't seeing a benefit from running multiple instances of the softwre you use is because you didn't have enough of either hard drives or RAM, if not both.

 

[Evgeny]

OK, then, E8600 it is. Thanks very much, myocardia.

 

[Idontcare]

http://weblogs.asp.net/scottgu...ance-with-VS-2005.aspx

As Myo was stating, compiling is a fairly disk-intensive nearly random read/write sequence for these type of projects.

Doing something to improve your hard-drives latency (get a 10k rpm raptor or even better get an Intel SSD) is likely to be the next best bottleneck to improve upon in your system.

 

[AmberClad]

In terms of memory usage, I don't think VS is particularly remarkable. Looking at my Task Manager, the devenv.exe process is standing at just slightly over 200 megs. It's a bit higher when debugging. That's not exactly lightweight, but it's nothing compared to CoH, which can chew up almost 2 full gigs of memory.

It does seem pretty disk-intensive though. Not so much on my desktop, but on my laptop (a dual core with a 5400RPM drive), I/O gets very sluggish while VS is compiling. I'd love to have a Velociraptor or Intel X25-M for my work laptop.

 

[BonzaiDuck]

Originally posted by: AmberClad
In terms of memory usage, I don't think VS is particularly remarkable. Looking at my Task Manager, the devenv.exe process is standing at just slightly over 200 megs. It's a bit higher when debugging. That's not exactly lightweight, but it's nothing compared to CoH, which can chew up almost 2 full gigs of memory.

It does seem pretty disk-intensive though. Not so much on my desktop, but on my laptop (a dual core with a 5400RPM drive), I/O gets very sluggish while VS is compiling. I'd love to have a Velociraptor or Intel X25-M for my work laptop.

I could try and say I "invented" the first laptop about the time of the Osborne-1. It was a battery-driven Sinclair-ZX81 or Timex-Sinclair-1000 using a Sony Watchman (2.5" diag. screen!), in my grandmother's overnight case stuffed with gray foam-rubber.

That was the last laptop I ever owned, and i've been "putting it off" because I'm not that mobile now.

The laptop drives have potential for those chained to their desktop systems, for building RAID arrays with so-so performance (but stellar compared to a single 5400-rpm drive.) I'd seen an external RAID box -- can't remember the manufacturer name (lucky I could even remember the Sinclair) -- which used them. They're compact; you can swap them out and store them easily. I contacted Western Digital a month ago to ask if you could just hook them up to a desktop system, and the answer was a simple "yes."

That would reduce the power-requirements for a desktop RAID5. I suppose -- if you could find that external box -- you could find a day-pack napsack to carry it in. THAT would speed up your laptop -- probably beef up your pecs and lats, too.

 

[soccerballtux]

So yeah, definitely the E8600, and be sure to overclock it. You can push those to 4.5Ghz if you want. That would get the processor out of the way.

 

[dmw16]

You would overclock a "work" machine? I wouldn't want t risk stability on a machine that is critical to what I am doing. A lockup before a save would drive me crazy.

 

[Diogenes2]

If it's stable it's stable .. Most 8600's seem to be doing 4g on default voltages ..

Running at default clock, is like pulling a couple of sparkplug wires off of a V8 ..

 

[BonzaiDuck]

Originally posted by: Diogenes2
If it's stable it's stable .. Most 8600's seem to be doing 4g on default voltages ..

Running at default clock, is like pulling a couple of sparkplug wires off of a V8 ..

"Hey, Babe! Take a walk on the wild side . . . I say . . . Hey, Babe! Take a walk on the wild side!" :laugh:

I used to be a software "professional." I can see the concerns. I swore -- all that time -- that over-clocking was asking for trouble. I bought plain-vanilla Intel boards.

But I'm also an economist who is well aware of Intel's market position, and of the reasons for the big "counterfeit Pentium II" debacle back in the late '90s.

You think they have a separate assembly-line for each processor-model? Maybe for various "sub-species." I think . . . . they test samples or sub-samples, and modify them if they pass certain specs. That means, for the remainder of the population not sampled but binned at lower-rated specs, there must be plenty that might have passed anyway.

Of course -- what high-end Intel processor runs at 4 Ghz? I'd say this: if it passes IntelBurnTest after going through a 24-hour-long PRIME95 gauntlet . . . . DOOO it! DOOO it!!

[retired person, held harmless for being cavalier and irresponsible. I approve this message.]

 

[soccerballtux]

Originally posted by: dmw16
You would overclock a "work" machine? I wouldn't want t risk stability on a machine that is critical to what I am doing. A lockup before a save would drive me crazy.

No, that's why you use the autosave function, and why you could Orthos test for 48 hours. Besides, it wouldn't lock up under 0% load, it would only be at 100% load that it is most likely to lock up. So, that would be when you're compiling. And so you use the "Save and Compile" button/keycombination, and you're good.

 

[GarionGoh]

Hi Evgeny,

I've also recently setup a 'work'station at home, primarily for .NET development. My line of tots is this: If I am going to do lots of development, it means I'll be doing lots of testing as well. And if I'm doing lots of testing, I may actually need more than 1 machine. At least 1 or 2 clients + 1 or 2 servers. I can imagine going with a SQL svr and a web svr, not to mention AD and such. So since I cannot have a 42U rack at home (even if I can afford it, my wife will almost kill me), the best way to do that is thru virtualization, running Virtual Servers or Virtual PCs, both of which are free from MS. In such a scenario, you would appreciate having 4 core.

Unless you have a huge website, the compile speed difference between the E8xxx and the Q9xxx is not going to be that significant. I would rather, if you have the spares, get that new Intel SSD or some SLC SSD, dun get the cheaper MLC SSD (read latest review)

 

[PlanetX]

Originally posted by: soccerballtux

You can push those to 4.5Ghz if you want.

OK, let's get rid of these silly myths. It's posts like this that build up too much hype for a CPU, then people get disappointed when their E8600 won't do 4.5GHz. Mine might do it, but it'd take crazy voltage to get there. 4.25GHz at 1.372 (load) is the best I can do without going over 1.4V. Some E8600's are dogs. Let's get this straight. Not all are created equal, and you'll find a number of week 20's that are worse than mine.

 

[Markfw]

Originally posted by: PlanetX

Originally posted by: soccerballtux

You can push those to 4.5Ghz if you want.

OK, let's get rid of these silly myths. It's posts like this that build up too much hype for a CPU, then people get disappointed when their E8600 won't do 4.5GHz. Mine might do it, but it'd take crazy voltage to get there. 4.25GHz at 1.372 (load) is the best I can do without going over 1.4V. Some E8600's are dogs. Let's get this straight. Not all are created equal, and you'll find a number of week 20's that are worse than mine.

Exactly. My E0 stepping E8400 won't go over 4.07 without a lot of voltage...

 

[AmberClad]

Originally posted by: dmw16
You would overclock a "work" machine? I wouldn't want t risk stability on a machine that is critical to what I am doing. A lockup before a save would drive me crazy.

You would certainly want to make sure it's highly stable, more so than for a typical gaming rig. Depending on what exactly it is that you're compiling and the type of work it's used for, I personally would want to make sure it's LinPack stable with no calculation errors, not merely OCCT/P95 stable.



Btw, I agree with those who find some people's expectations as far as E0 overclocks a bit unrealistic. It's not clear if those people actually own one, or if they're repeating what they've heard from others.

Can you get 4.5GHz+? Possibly. With safe voltages (<1.4v)? You'd have to get very lucky. The only possible example I've personally seen so far is this screenshot, but that's only 15 min stable and it's with P95, not OCCT or LinPack, which error much faster.

 

[taltamir]

A lockup on my GAMING machine would drive me even more crazy then a work machine... I game to relax and have fun, not to loose hours of work due to a lockup. (so I forgot to save, sue me)
Which is exactly why I made sure the OC is 100% stable.

You OCing it is no different then intel or AMD selling you a higher clocked chip if done correctly.

 

[Idontcare]

Originally posted by: taltamir
You OCing it is no different then intel or AMD selling you a higher clocked chip if done correctly.

That's actually a bad misnomer.

It could only be true for very specific and select situations where Intel or AMD intentionally create a speedbin which is populated by 100% higher speed-grade chips (but this is never the case in practice).

There will always be those chips which bin out for a particular speedbin which simply cannot function properly at higher speedbins, regardless of Vcore and cooling.

In our laymen terms we'd say such chips are "not TDP limited" or "not Vcore limited".

But it is far far from the reality of the physics that create speedbin distributions to claim that the reason you won't get an overclock comes down to the overclock "not being done properly".

Higher clocked chips from Intel or AMD does not mean they were overclocked lower-bin chips, nor does it mean that lower-bin chips are simply nothing more than higher-bin capable chips which have been intentionally down-binned.

This isn't to say this doesn't happen for some chips, of course it happens but it doesn't happen to 100% of the chips that populate a particular speedbin.

 

[Dadofamunky]

Originally posted by: PlanetX

Originally posted by: soccerballtux

You can push those to 4.5Ghz if you want.

OK, let's get rid of these silly myths. It's posts like this that build up too much hype for a CPU, then people get disappointed when their E8600 won't do 4.5GHz. Mine might do it, but it'd take crazy voltage to get there. 4.25GHz at 1.372 (load) is the best I can do without going over 1.4V. Some E8600's are dogs. Let's get this straight. Not all are created equal, and you'll find a number of week 20's that are worse than mine.

QFT. I think mine is (it won't do 4.3 GHz on stock volts. Grrr!). Still, I'm glad I plumped for it.

 

[Tempered81]

Originally posted by: Dadofamunky


it won't do 4.3 GHz on stock volts. Grrr!

this is NOT a bad thing. bad is when people cant get over 4.1 with > 1.3v

 

[Tempered81]

Originally posted by: AmberClad
The only possible example I've personally seen so far is this screenshot, but that's only 15 min stable and it's with P95, not OCCT or LinPack, which error much faster.

how about this
10 x 450 1.344v passed 10 hours prime just a few hours ago.
or this
10 x 450 1.368v passed 8 hours prime

Originally posted by: soccerballtux
So yeah, definitely the E8600, and be sure to overclock it. You can push those to 4.5Ghz if you want.

:thumbsup:

 

[soccerballtux]

Originally posted by: Dadofamunky

Originally posted by: PlanetX

Originally posted by: soccerballtux

You can push those to 4.5Ghz if you want.

OK, let's get rid of these silly myths. It's posts like this that build up too much hype for a CPU, then people get disappointed when their E8600 won't do 4.5GHz. Mine might do it, but it'd take crazy voltage to get there. 4.25GHz at 1.372 (load) is the best I can do without going over 1.4V. Some E8600's are dogs. Let's get this straight. Not all are created equal, and you'll find a number of week 20's that are worse than mine.

QFT. I think mine is (it won't do 4.3 GHz on stock volts. Grrr!). Still, I'm glad I plumped for it.

:/

Crazy voltage, like 1.4 something? 1.4 is the new 1.5. You can go a wee bit past. Get it so loaded it's 1.4 in CPU-z and see if you can hit 4.5Ghz.

Also, get better cooling? I just improved my e2180 overclock from 3.4Ghz to 3.52 (possibly 3.53) with a better fan and more voltage (1.512 in CPU-z).