First of all, I'd like to say something about the raw power of AMD Bulldozer when it comes to Fractal Extreme. People have been repeatedly stating again and again that the Sandy Bridge 2600 outperforms the AMD Bulldozer platform, but I believe that is a misnomer because most benchmarking software uses float calcs, not integer.
The Bulldozer platform gets less operations per clock per core than the Phenom II, but it makes up for this with more cores. I have also noted through experimentation (by using the "affinity" setting in Task Manager) that loading two integer threads on a single module, without varying the clock speed, takes about a 20% performance penalty compared to running the two threads on separate pipelines. As far as Windows 7 Task Manager is concerned, the threads are grouped this way, (0,4) (1,5) (2,6) (3,7) with (x,x) denoting a single pipeline (2 cores). I imagine floating point would incur a nearly 50% performance penalty, or something close to that.
You can go to "render" and click "pause", or just press p. If you then zoom in, and undo, it won't render because it's paused. If you save the file and re-open it, it will automatically start rendering. I found this out when I was trying to solve this save problem for the benchmark test with the i7 2600.
I did the test again and it turns out my time is now 7:34,8.
Your time: 2 * 60 + 48 = 168 seconds
My time: 7 * 60 + 35 = 455 seconds
455 / 168 = 2,708 times faster
That's with overclock on your processor, a very nice speed-up. Thank you for this information, really helps me. If I could just have 2 of those in one computer.
As it turns out, assuming Dinkydau's i7 2600 processor is configured correctly, he's got 4 cores (and a total of 8 threads due to hyperthreading, which will only provide performance benefit if the other thread stalling the CPU, which isn't likely with Fractal Extreme) operating at 3.4Ghz
My FX-8150 was overclocked at 4.2Ghz. I did some testing at different clock cycles by adjusting the multiplier, and the change in render time was near exactly inversely proportional to the change in clock speed. Since everything (RAM, Northbridge, etc) else stayed exactly the same, the only bottleneck in this scenario was the raw clock cycles.
Dinkydau, I did some crunching on our render times to see what our render results would have looked like per core per Ghz (lower is better).
AMD FX-8150 @ 4.2Ghz (8 cores) = 2:48 (168 seconds)
AMD Phenom II 955 @ 3.2Ghz (4 cores) = 4:55 (255 seconds)
Intel Sandy Bridge i7 2600k (4 cores) = 7:35 (455 seconds)
Bulldozer single core @ 1Ghz = 168 x 8 x 4.2 =
5644.8Phenom II (K10) single core @ 1Ghz = 255 x 4 x 3.2 =
3264Sandy Bridge single core @ 1Ghz = 455 x 4 x 3.4 =
6188Despite it's shortcomings, the 1Ghz bulldozer core actually beat the Sandy Bridge, but the older Phenom II wins hands down, so it may actually be in your best interest to get the last gen Opteron (if you can somehow equal the number of cores and clock cycles). By my calculations, if I take the 3264 figure for the Phenom II, and divide it by 6 cores and again by 3.2Ghz, then I would presumably get a render time of 170 seconds for a Phenom X6 1090T processor running at stock clock speed. That's nearly the same performance value as my overclocked FX-8150!
3264 / 6 cores / 3.2Ghz = 170 seconds
Or, if I simply set the multiplier to 18 and run the X6 at 3.6Ghz, essentially forcing all cores to run simultaneously in turbo mode (at rated turbo speed they should all be able to run completely stable), which is possible and safe to do with a good aftermarket heatsink, then my theoretical render time should be:
3264 / 6 / 3.6 = 151 seconds, which would actually outperform the 8 Bulldozer cores!
The reason why I bring this up, about the older Phenom II outperforming the bulldozer, is that you may be still be able to purchase the last generation AMD Opteron and get similar performance value out of it. Sadly, it appears based on the out-of-stock status on the NewEgg website, that the Phenom X6-1090T desktop processor is no longer being manufactured anymore, but people were recommending that over the Bulldozer, and there may be some truth to those statements.
I could be mistaken, but if the (45nm, K10) six-core Phenom II X6 1090T can outperform the (32nm, Bulldozer) 8-core Bulldozer FX-8150 on the desktop platform, then there is a good possibility that the (45nm, K10) Magny-Cours 12-core may be able to outperform the (32nm, Bulldozer) 16-core Interlagos at similar clock speeds. And that is for Integer performance. Floating point will definitely get an even higher yield on the older architecture, since the newer Bulldozer processors cut the actual FPU count in half.
A couple of more thoughts on this...
If you can wait a few months, the second generation of AMD Bulldozer processors are right around the corner, codenamed "PileDriver". I don't know about the server market, but here's a quote from Wikipedia regarding the Desktop processors. I may or may not decide to upgrade my existing setup to PileDriver at a later date, but it's supposed to solve the issues regarding lower throughput per clock per core.
2nd Generation
AMD Financial Analyst Day 2010[30] revealed the 2nd generation is scheduled for 2012; AMD referred to this generation as Enhanced Bulldozer. This later generation of Bulldozer core is codenamed Piledriver. The focus of Piledriver is to improve instructions per clock and frequency.[31] It is intended for specific desktop and notebook markets:
2nd Generation FX-series CPU - Desktop Performance market (Volan platform):[32] Zambezi's replacement is Vishera, with up to 8 cores; with Turbo Core 3.0 while using the existing Socket AM3+ format and 9xx series chipset of the 1st generation FX-series Zambezi processor. AMD says that this 2nd-generation FX-series processor would offer up to 20% to 30% better performance increase under digital media workloads. Vishera will continue to use dual-channel DDR3 memory interface.[33][34] The 2nd generation FX-series is scheduled to be released in Q3 2012 (July to Sept) with the FX-8350, FX-8320, FX-6300, and FX-4320 models.[35]
Also, if you are going to be spending all this money on a new workstation PC, you might as well get a good EATX case as well. I recommend:
http://www.yycase.com/yy-w2xx.htm#It's heavy as hell, but it's worth the money, plus it dampens the internal sound/vibrations much better than a lightweight aluminum case. I got the yy-w201 with option B (hard drive bay with 92mm intake fan and mounts for up to 5 HDs). You'll likely never use up all of the expansion bays. Two of mine I replaced with drawers where I keep loose screws, driver CDs, and small parts.
Another thing to consider, is it really worth the amount of money required to build a high-performance server platform which will likely be bested by desktops in a couple of years? The reason why I am asking you this, is that I spent a large sum of money ($5000 USD) to build a similar system back in 2005. In fact, I am still using the case, power supply, and DVD-Drive of that original system. At the time of build, I actually had to wait a couple of months after building before I could really start using it until Microsoft Windows XP x64 was released. Originally, I had Red-Hat Linux running, which I didn't like too much and wasn't compatible with Windows software. The system had two 2.4Ghz Opteron processors (single core) and 4 Gigs of RAM. Had I waited just a year or two more, I could have gotten a cheap dual-core Athlon 64 desktop processor which would have danced circles around the old Opteron setup. Another issue at the time was lack of support for 64-bit Windows XP drivers for most of my hardware. I had a silent PC with no audio because the motherboard manufacturer (Tyan) hadn't even released 64-bit XP drivers for the integrated sound. I'm not sure if they ever did or not. And most software that existed at the time I built the PC didn't support multithreading, so the single-threaded performance gain I got out of the behemoth was only slightly better than my old Athlon XP rig.
Of course, a lot has changed in the past 6-7 years. 64-bit software drivers are everywhere now, and many software apps are available which support multiple threads/CPUs. A couple of neat little quirks about server processors: You can run like 20 instances of a program which is a known resource hog, and the server processor won't choke on it like a desktop will. And I say this meaning my original system had only two physical single-core CPUs, so it supported 2 execution threads, but I could load my computer down with 12 separate instances of Bryce 5 (a 3D ray-tracing software I liked to use back then, which unfortunately did not support multithreading until version 6 came out a couple years later) and still be able to play my favorite N64 games in a full-screen emulator without dropping much fps. My Athlon PC could barely handle the emulator running in a window with nothing else running in the background, though the video card may have been a factor in that. Basically, server CPUs can handle a lot more thread thrashing before they start to choke than a desktop CPU can (they're basically designed from the ground up to be overloaded), but single-threaded performance of anything (like games) will be abysmal. That's why desktop processors have fewer cores but much higher clocks.
Anyway, just think carefully before you buy, and I hope you get excellent results with whatever you decide to purchase. LOL, long post...
EDIT: If anyone is interested, I have uploaded a copy of my benchmark *.fx file, you can find it here and compare your computer's performance:
http://www.fractalforums.com/fractal-exteme/fractal-extreme-performance-benchmark/Now the "crappy old" desktop render more fast then my "good" computer and never crash, no because is more powerful (as i said is the opposite a obsolete thingy) but because its OS has nothing else to do
So in few words iyou may discover that a computer found for few bucks at a garage sale or even in a trash can may render quickly and better than a pricey top level monster
if you just use it only for that
i still tweaking to see if i can streamline even more i found some utility (
http://en.wikipedia.org/wiki/NLite_and_vLite ) to easier the process
i am talking about Win because i am mostly using Mandlebulb3D but for linux is even more easy ( in case some cool info and utilities are on
http://en.wikipedia.org/wiki/List_of_remastering_software ).
So in a few words before spend money on a new PC i would suggest you to start practicing costless hacks,only once you got them may be worth spend money ona new PC (but may be worth spend no more then 50$ to buy some computer destined to the trash can just to pratice )
----------------------------
still i am very interested to suggestion about the most powerful computer at reasonable price...
but even the most powerful can't beat the most crappy if the powerful is overloaded by tons of always running services while the crappy has to do just 1 thing
Sorry, but some of the info presented here is inaccurate. Back in the olden days when CPUs had only one core, and computers had very little RAM, your statements about uninstalling unneeded software may have been accurate. And some of what you are saying definitely holds true if you have some sort of malware that is wasting CPU or network resources. But as long as your harddrive has a reasonable amount of free space (15% or more) and has been defragmented withing the last month, and you have 2Gb or more of RAM, and a multicore CPU, these things shouldn't be an issue. To test whether or not some rouge application is wasting CPU resources, open the task manager and select the processes tab. Click the CPU load button at the top to arrange the list of programs by their CPU load. Most idle processes will show little to zero CPU load when they're not doing anything. If not, you can terminate the process, or better yet, find out what software it's associated with and disable it. I use my large desktop machine to web surf and perform other tasks all the time, and it rarely if ever affects the render progress. If your render application is bogging down your web surfing or other activities, you can change the render application's priority to "low" and that should free up CPU cycles for other programs like web browsers. Don't ever change your render program higher than "normal" or your PC may become unresponsive. But it is in most cases unnecessary to uninstall things like LAN adapters and stuff. Windows 7 does needs to be set not to sleep if you plan on doing render jobs while you are away from the computer, and Automatic updates should use the option "download updates, but let me decide when to install them" otherwise the computer will automatically restart after it applies the updates and you will lose your work.
But on a modern computer with modern operating system, your CPU usage will be practically zero or at most 1% at idle, and about 50%-75% of your RAM should be freely available. A quad-core processor will typically do a render job twice as fast than a dual core at the same clock speed, so faster clocks and more cores typically mean shorter render times. Also newer processors often have updated instruction sets and more efficient pipe-lines, larger caches, etc, so your argument of salvaging an old computer from the scrap bin and then expecting major performance gains simply by modifying or removing unneeded software, without opening up the case and replacing a lot of hardware, is pretty naive. However, if you just want a box to surf the web on, an old dinosaur will probably do fine. You may get between a 2-5% gain in performance by tinkering with software/ cleaning up the hard drive, as opposed to a 100%-200% or higher gain by simply purchasing and installing a new CPU or CPU/Motherboard combo.