Building a PC, Part V: Upgrading

Last summer I posted a four part series on building your own PC:


My personal system is basically identical to that build, though it predates it by about six months. The only significant difference is the substitution of the Core 2 Duo E6600 CPU.

This is a companion discussion topic for the original blog entry at:

Very nice, shouldn’t a torture test be longer? I wouldn’t feel comfortable if my mechanic supped up my car and only test drove it around the block.

Quad core: I think you’re totally wrong on this issue. I don’t think any conscionable developer can, with honesty to their craft, use anything less in this day and age. It’s well known that CPUs are going to do far more scaling out, in terms of cores, than up, in terms of hertz, in the future.

If you describe yourself as a professional developer at all, you owe it to your customers and possible future employers to know as much as you can about making the software you write scale on these platforms of the future. That means making multicore as much a part of your world as possible.

By chasing hertz, you are part of the problem, not part of the solution.

FWIW, I run my Q6600 at 2.8GHz, after retiring my E6700 @ 3GHz. There is nothing much that happens on the desktop any more that really demands that much CPU power apart from gaming and media - particularly video encoding. However, media encoding is trivially parallelizable, so quad will beat dual there. Meanwhile, a Core 2 chip running at 2.8GHz, running with a reasonably recent graphics card (8800GTX in my case), can handle pretty much everything out there, especially when running at 1920x1200, a resolution where the graphics card is a bigger bottleneck than the CPU.

Would be nice to see a post about the things related to overclocking. What about heat levels? What cooling do you use? What kinds of fan does one have to invest in? How about noise levels?

It isn’t as easy as just clicking a button in the BIOS, right?

Adam – please read the first four links in the article. All is revealed there :slight_smile:

Also, this:

Building a Quiet PC

You don’t deserve a fast computer after that article.

AsusUpdate for Win64 bricks your motherboard in a very nice fashion

Yep, I read it has to do with updating the latest BIOS over a very old BIOS. Still unforgivable, since ASUS bricked my ASUS motherboard. I’ve grown a little gun shy of Windows BIOS updaters as a result, so I’ll be updating my BIOS the old fashioned way from this point on. At least newer BIOSes allow you to flash from the BIOS itself using a USB key which is pretty convenient.

IE was faster in at least some of the tests , but in none above? What gives?

Firefox 3 is much, much faster at JavaScript than Firefox 2.

Aren’t you liable to burn through your processor a lot faster [when overclocking]?

These are just hunks of silicon and metal; perhaps if you kept it for ten years you might see some ill effect, but even then, I doubt it. And who keeps the same CPU for ten years?

Taking a potshot at Scott Hanselman’s box? :confused:

No, not at all. I just want to dissuade people from the mindless 42 argument in favor of actually looking at some benchmarks. There are a few edge cases where 4 cores ARE faster, even at lower clock speeds. But not very many, and I’m unconvinced that most users spend much time at all encoding audio/video or rendering 3D scenes.

What about hardcore gaming (which I do) and running Apache and MySQL in the background? What about DVD encoding while streaming music and working on source code? How about virtualization for the developer that wants to test his cross-platform application?

I/O will be the largest bottleneck in the system by far, considering most CPUs are 99% idle all the time. People like to think that they multitask enough to replicate the behavior of a server under the load of thousands of individual users, but it’s pure fantasy IMO.

That said, if you spend all day encoding media or rendering 3D scenes, then you could certainly make a strong case that quad is a better choice for you. Make sure you have two spindles, though.

Jeff should have been eating humble pie shortly after he wrote his original article, but instead he’ll lean on dated, cherry-picked benchmarks in artificial situations.

Then link me some benchmarks that prove me wrong – common apps that benefit from having 4 cores over merely 2.

[Intel’s EDAT] On unoptimized threads that really need clocks, the CPU will be able to power down one core and overclock another automatically

This sounds perfect! I wonder how it “decides” cores are needed?

I generally agree that Dual core is sufficient for desktop, but there are cases where quad core is useful e.g. media work (photo/video editing, video encoding etc.). Compile times also speeded up using the -j4 switch.

That means making multicore as much a part of your world as possible.

Dual core is multicore. If you can’t scale to two cores worth a damn, throwing more in isn’t going to change the situation much.

At the point when quad and dual are the same price for the same speeds, then quad (on the desktop) will make sense.

I love your blog Jeff, the parts about coding and software issues at least. But when it comes to hardware advice and tutorials, I really couldn’t give a fu*k. It’s not exactly “coding” horror any more.

Am I the only geek on the planet who cares about software, but not hardware?

Next week on coding horror: how to build a table to put your PC on.

Just wondering if you have updated power consumption values for the 3GHz vs. 4GHz at idle.

"Am I the only geek on the planet who cares about software, but not hardware?

Next week on coding horror: how to build a table to put your PC on."

I think I just lost all of the remaining respect I had for you Jeff. Not that you care, I know you hear it a lot from many longterm disappointed readers – and I thought about just quietly removing your blog from my iGoogle with a sad shake of my head, but I thought I’d at least first make mention that I thought you were a bit more reasonable than this.

Good luck with your new business venture. You’re a smart person and I’m sure you’ll be successful, but if you take any advice away from your blog, let it be humility.

Then link me some benchmarks that prove me wrong – common apps that benefit from having 4 cores over merely 2.

Your opinion is one of “enthusiast”: You park your PC in the Burger King parking lot and compare 0-benchmark speeds.

Here in the real world, most people have a small set of software pain points – things that actually take enough time, or are prevalent enough, that they would even notice a difference between a Centrino 1.5Ghz system and a Core 2 2.4Ghz (not just in “single tasks”, but in nuisances throughout the day, like the PC becoming sluggish because it is doing {X}). Most of us don’t sit running benchmarks all day, or downgrade to low resolution so we can demonstrate a hypothetical advantage of clock rate.

There’s a remarkably high correlation between those “pain points” and applications that already support highly parallelized operations. Doing a project build is highly scalable across cores. Encoding video – especially now with high def video cameras – is a huge pain point for a lot of people, and quad cores (and of course moreso octuple-cores and dual octuple-cores, and…) absolutely dominate.

But keep on running Sunspider and tweaking your settings. For actual everyday use some guy with a stock, el-cheapo Q6600 is likely running circles around you and your polemics.

CoreTemp is now digitally signed, so it works beautifully under Vista and XP in all cases:

I just sent the developer $5 via PayPal as a reward, I am very pleased that I can run this great little program easily under x64 now!

Note that C1E state is DISABLED by default. C1E allows the chip to reduce voltage as well as multiplier at idle, but can lead to instability when overclocking so it is typically disabled by overclockers. I am surprised that it’s off by default, though.

Also, I am still using the same giant Scythe Infinity tower CPU cooler that I linked in the post.

Dual Prime95 torture load temps (E8500 @ 4.0 GHz, 1.368v)

Core #0 - 75c
Core #1 - 75c

Idle temps:

Core #0 - 49c
Core #1 - 49c

TJMax is reported as 105c which I assume is the internal throttle temperature.

@RWW - yes, you probably are!

Hardware is irrevocably tied to software and vice versa. If you don’t care about hardware, try running VS2008 on an Athlon 1200 with 128mb RAM.

I had the exact same case ! And there are more reports on the internet.

AsusUpdate for Win64 bricks your motherboard in a very nice fashion:

Erasing bios: 100%
Flashing bios: 100%
Verifying bios: 0%

And then it’s game over. (= RMA with 5-6 weeks delay before your board is returned)

For actual everyday use some guy with a stock, el-cheapo Q6600 is likely running circles around you and your polemics

That’s an interesting position, but I still believe that something akin to JavaScript performance in webmail is far more relevant to mainstream users than encoding high definition video.

But as I said, if you really do encode high-definition video all day, then certainly you want 4 cores. There are edge conditions where 4 cores will win by a large margin – but I want people to understand those edge conditions and where their work fits relative to those conditions, not just blindly assume 42. When that happens, Intel’s marketing team wins. :frowning:

I miss overclocking, I’ve had nothing but a notebook in about 2 years. One day!

I really don’t wan tto look back in two years and realised I missed that was OC’ing the CoreX**** series as they seem to be giving great returns, maybe no Celeron 300A, but close enough!

I bought a Q6600 for cheap on eBay and I sold my former E6420 for a very good price. So the “upgrade” cost me about 50 € (79$).

Worth it for the price, I think. Do I really need the 4 cores? I don’t think so. But now I can brag I have a Quad Core! :slight_smile: