Ah, yes, well, it is not the latest one out there but it is a start.

Something to avoid: John Deere PSUs. I had 2 of them give up the ghost after one boot. After powering down, the system would never boot again. :D

Deer is notorious. What's more, a lot of PSU vendors sell Deer products under their own name. You can end up with one of these fire hazards without even knowing it.

A shopping list consisting of "any Corsair, any Seasonic, any Antec except 'Basiq'" is not a bad starting point. Usually, the lowest-priced PSU of acceptable quality in any power range will be one of these three, anyway.

I know you're feeling the limitations of your setup, but I think you would be better off just waiting. There are so many components that don't mesh well with what you have and an entirely new rig with modest upgrades wouldn't cost you much more. If anything, you could get those monitors (as you were already told, the CPU wouldn't be worth much) or do like I did and get a hybrid hard drive like the Momentus. I'm on a late 2006 laptop and I'm fairly impressed with it's performance right now, though I'm not doing much cutting edge gaming. My processor may be obsolete, but I rarely feel it.

Well thank you everyone for your thoughts on this. I tried to look a dept stores last night for a generic PC, but there was nothing in the price range that looked right.


I actually went a route I didn't think I'd be able to do. I'm using my video card which for the most part is quite a decent card that still holds up today, my PSU, DVD Rom and HDD from my current PC and I went on newegg today. and with that 200-300 got some parts to build me a quad core machine, that will make me happy and last me a couple years I think. I just hope my PC building skills have improved... I get nervous about putting one together.

http://www.newegg.com/Product/Product.aspx?Item=N82E16811147121
http://www.newegg.com/Product/Product.aspx?Item=N82E16813138345
http://www.newegg.com/Product/Product.aspx?Item=N82E16819103996
http://www.newegg.com/Product/Product.aspx?Item=N82E16820104231
http://www.newegg.com/Product/Product.aspx?Item=N82E16835100007

$250 after discounts (the mobo/CPU comes in a combo deal)

Add to that my

700w PSU
500GB SATA HDD
One of my spare CD/DVD Roms.
1GB Gigabyte Radeon 4850 HD
Windows 7 Ultimate 64bit

I think I'll have a decent enough machine for awhile.

Nice idea and I recommend you follow that route. Some might say the FX series are bad, and they'd be right but in you situation the cpu is the best you could buy. However, from the reviews I've read, the cpu does run a little hot. will the current case you have provide sufficient cooling for the new cpu? Is it roomy enough to provide good airflow and circulation?

Oh, and when you do take your GFX card out, give it a good clean with an air can, in fact give your entire case a good clean. Can and does make a difference!

Yeah I would invest for a longer term if I had somehow gotten into an i5 or something but this is cheap and while the FX isn't the best CPU quad there is, it's leagues better than what I have.

Oh did you see the case I got? It's the top link. looks to be pretty cool (temp wise) and roomy to me.


As long as I don't screw it up in building, this will be the best PC I've owned before.

Good approach.

The rap against the newer AMD CPUs is that they do not reach their full potential under Windows 7. They still run fine, just not so fast as they could under a smarter OS. The problem will be corrected in Windows 8. So I would consider upgrading to Windows 8 when it goes GA (presumably later this year).

Since that is a microATX motherboard, you do not really need a new ATX case for it. It should fit fine in your existing case.

The approach is good. The CPU is crap though. It's weaker than a Phenom II in the same price range, it needs too much power and cooling for its rather meager performance, and its general architecture has shown to be a failure in many regards. Currently even dual core CPUs can handle threads more efficiently than this architecture (though that may change with a Windows update that ought to increase performance by about 10% by distributing threads better). The marketing also borders on fraud since you don't really get 4 full cores - important resources are shared, and are one of the reasons for the abysmal performance of this CPU. Any Sandy Bridge CPU runs circles around the FX-4xxx for a not much higher cost. The FX architecture is so spectacularly bad that AMD didn't even send the lower-end chips to the review sites. When pitched against each other, the i3-2xxx beats the FX-6xxx in almost every single application, professional image rendering with certain programs being one of the very few exceptions. And your CPU is considerably weaker than an FX-6xxx.

I didn't see an answer to the question what you're wanting to do with your machine, but if it's anything CPU-dependent, then you'll want to replace that CPU as soon as you can. Personally, I'd seriously consider returning it.

Sorry if that sounds harsh. But I've read a lot about exactly these CPUs in the past (our second machine is due for an update), and I really wouldn't recommend them to anybody.

Re FX-4100

Personally, I like the new way of thinking in the Bulldozer architecture. It wasn't very well received, though; quite a lot of the "cons" have been blown totally out of proportion. Like the way they've implemented the FPU, or more specifically, that you "only" have half as many 256bit wide FPU's as you have cores.

These 256bit FPU's are capable of splitting into two 128bit FPU's each, so you do have as many 128bit FPU's as you have ALU's etc, but that apparently isn't good enough for some. Thing is, the only use for 256bit FPU's is the added support of AVX (Advanced Vector Extensions). The first AVX capable processor was first shipped one year ago, and you need SP1 for Win7 to even be able to use it. So, IMO, crying over the lack of a full set of 256bit wide FPU's is quite an overreaction.

Another "problem" is that most OS'es today can't utilize this architecture optimally. I don't see why we'd have to wait for Win8 for this kind of support, though. The way I see it, the Bulldozer architecture is a vast improvement over the Intel Hyperthreading architecture, rather than a doubling of cores. Viewed as such - and there already is support for HT - it shouldn't be very hard to extend this support to the Bulldozer line.

Trying to get back on topic; your FX-4100 isn't comparable to an Intel quad, but that isn't necessarily a problem. It's cheaper (I assume - my budget for a new rig where I live would be more like $1600+, sans monitors - I paid more than $600 just for my i7 CPU (and it doesn't even have a single 256bit FPU/AVX support), so I don't know if $110 is cheap or pricey) but it'll give you most of what you'll actually be needing. There simply aren't that many programs out there that efficiently utilize multiple cores, at least not the kind of programs you'll likely be using.

</rant>
Correction: That's true only if the first thread is currently utilizing AVX. Any other scenario and the 256bit wide FPU will split nicely into two 128bit FPU's, one for each thread.

I think it's better to look at the Bulldozer as a vastly improved Hyperthreading architecture than anything else, even if each module is very close to being two cores. The saved "real estate" on-die because of the "single" FPU is a huge step in the right direction, IMO - there's no point in keep trying to make one-size-fits-all solutions when computers are used for such a variety of tasks.

Yes, your understanding is correct. The Windows scheduler does not "know" how to optimize core affinity for the new AMD architecture; thus it makes some poor decisions that impact performance.

Intel CPUs have damned big shared L2 (in Core 2) or L3 (in Nehalem and Sandy Bridge) caches. Core affinity doesn't too much matter, because any core can get to any row of cache without a significant performance penalty.

AMD CPUs have separate L2 caches. Exactly how the L2 cache is arranged differs among their various architectures. A core that needs a row that is not in its cache has to go through HyperTransport to get it from the core that has it. So on AMD CPUs, core affinity matters a lot.

There are two specific (limitations, blunders, call them what you want) in the Windows scheduler that particularly affect AMD FX CPUs. One is, the scheduler pays little attention to core affinity even when the number of threads that need scheduling is small. You end up with, say, a 4-thread load with each of the threads being bounced between cores 1, 2, 3, and 4. The other is, the scheduler doesn't pay attention to what the core parking logic is doing. Windows is aggressive about parking cores (putting them in a low-power state), but when the scheduler blindly assigns a thread to a parked core, the core has to be powered up again, and this is slow.

Microsoft tried to fix this in Windows 7. The original fix was withdrawn, but it looks like they recently (January 12) got it right. It is now up again as KB2645594. It also requires KB2646060 (selectively disable core parking).

These fixes were already made in Windows 8. Microsoft had to backport them to Windows 7. The performance improvement is said to be 10% on loads that are "lightly" threaded (that is, up to 4 CPU-bound threads). This performance improvement can be seen in current builds of Windows 8, so it's definitely there.

Bulldozer is still a disappointment, if not such a waste of good sand as the original Phenoms were ;)

I don't know, if you can pick up soda cans or do an extra job for a neighbor and get an extra 100 USD you are starting to hit the range of some of the prebuilt iBuyPower rigs off of Newegg. If you're not going to be happy with the result I wouldn't invest into something that's going to piss me off for the next 3 years until I can justify replacing it. So ask yourself if you'd be happy with the craigslist result, or the Walmart PC, or any other choice. If the answer is no, stick the 300 USD in a shoebox, pretend you spent it (for real, no stealing it from yourself for an emergency) and stick in every extra dollar you can earn through the rest of Winter and Spring, see if it's not enough to buy what you want.

Thanks pH7 and cjrgreen, I learned new things from each of your posts. :)

Personally I'd still stay away from Bulldozer CPUs (except for some rather exotic usages, like pure rendering PCs) at least until benchmarks show that the performance gap between them and the Sandy Bridge CPUs has closed considerably. And I'm still doubtful whether that's possible, since the gap is surprisingly large, especially considering that in theory the architecture indeed sounded like a good middle road between Hyperthreading and full hardware cores. I can understand and accept though that others may be more optimistic in that regard. :)

Whooops, totally went past the first link, my bad! That's a nice roomy case and will fit everything in quite nicely. Jump on the new computer bandwagon. ;)

I second that - I like the idea of Bulldozer, but not the current implementation; it's going to take some time to get enough things right. (And even when they get it right, I'll probably stick to Intel personally, because of what I use it for; but that doesn't mean it'll be right for others.)

This kind of reminds me of what one of my CS professors told in class once. Intel, a few decades ago, gave their engineers the task of figuring out how to improve performance. Both engineers working on the hardware as well as the compilers where tasked with this, and they worked separately on it for about two years. When they got together and started testing, it turned out that the new cpu/compiler combo wasn't any faster than the old.

Part of the reason was that the compiler team had looked at what made their resulting code slow and concluded it was the use of complex instructions/op-codes that was the culprit. So they rewrote the compiler to use series of simpler (and faster) op-codes to do in software what the complex op-codes did for them in hardware.

The other part was that the CPU team had done the same, identifying the complex op-codes as being too slow, so they'd been working for two years on improving these. Of course, when the compiler stopped using the complex instructions, this didn't help at all..

Anyway, the point is that hardware and use of hardware should match, otherwise you'll get performance that is unfair to both.