>>43043950 As an owner of a Xeon, i can tell you now that for consumer workloads, there is no advantage.
Now, if you needed more than 6 cores, a xeon is where you go, especially if you need to go multi-socket. Just a heads up though: an 8 core Xeon w/ a decent core clock is going to run one hell of a pretty penny ($2k USD easy). Want even more cores? Its going to cost even more. Want to do a monster quad LGA 2011 box? Dont. Each CPU alone is pushing 4 grand, 8 gigs of ram for each CPU (4x2GB DDR3-1600 ECC Registered) would run another 500-1000 USD, and finally a barebones 4P server to hold it all another 2k.
Want moar coars, especially past the dual socket point? Go AMD. Its a metric fuckton cheaper, to the point where anyone with about $2500 USD can purchase a full AMD quad socket system, complete with ram, CPUs, board, heatsinks, power supplies, and a suitable case.
>>43044401 Depends on how many threads you need and your budget. AMD chips are naturally weaker than intel chips, so if you are doing something that isnt so heavily threaded, stick with intel. If you are doing something that benefits massively from multiple threads (folding, rendering, etc) go with AMD.
Theres a reason why distributed computing teams almost always use AMD 4P rigs. Not only are they cheap and easy to get, the supermicro boards they are typically run on can be flashed to an unlocked custom BIOS that would allow for overvolting and overclocking. IIRC the cinebench world record is currently held by an overclocked AMD 4P rig, and has remained unbroken for several years. Word of advice though if you go AMD 4P and overclock: expect a fuckton of power consumption, and make sure the board's VRMs are properly heatsinked. Fuckton of power consumption: depending on how high you push the system, you'll be looking at a consumption of about 300W per socket. Hyper 212+ heatsinks can handle the load, but make sure the rig itself has adequate airflow or is in a suitably cool place (basements work). VRMs properly heatsinked: the supermicro boards werent designed for that kind of power load in mind, and thus if the VRMs arent heatsinked they will overheat and catch fire. Friend of mine who had an AMD 4P lost it that way. If they ARE properly heatsinked though, the system will last much longer.
>>43044883 Using x264 with --preset placebo daily makes me dream of having a quad CPU opteron setup. :(
Anyway, right now you can find Sandy Bridge Xeons (8c/16t) at $250 a piece on ebay. So for $500 you could run a dual CPU Xeon rig with 16 cores / 32 threads for HALF the price of what Intel is going to charge for an 8core Haswell-E part. The only downside is that entry level Xeons are clocked very low (2ghz, 2.2ghz if you're lucky) and support no overclocking whatsoever.
>>43045129 Huh? Xeons aren't particularly power hungry and electricity is dirt cheap here in murrica. I'm sure you spent more money on a new cpu and mobo than the extra power cost on your xeon rig over its lifetime
>>43044996 Well, Folding@home scales very well with more threads, be they real, fake, weak or strong. Id have to go look around for a bit, but a properly configured AMD 4P rig could easily push the half million points per day mark. Its getting the hardware properly set up and running that is the issue. Once that is done and the box has a link to the net, its a fairly simple matter of getting the folding@home client configured and running.
Im not so sure about other distributed computing projects, but i think einstein@home also does well with plenty of threads. I'd have to start up my desktop (which, by luck and a fair bit of generosity, has a E5-2690 Xeon in it, basically the highest SB-E Xeon possible) and look thorugh the BOINC client to see what of the 3 projects i am running does well with multiple threads.
>>43045103 Those cheap xeons are crap, and most of them are of a stepping that is not guaranteed to work properly. And 2P LGA 2011 boards arent cheap. If he wants to go 2P Intel, he'd be much better off getting a pair of 1366 hex core Xeons, which can be cheaply had, and an SR-2 if he can find one.
>>43044911 The need for ECC is vastly overstated. Running less than 1TB RAM and the odds are nearly 0 that you'll experience any errors in a year. But go ahead and believe what you want. The only reason for ECC in my opinion is because memory is available in higher density.
>>43045247 A very large-scale study based on Google's very large number of servers was presented at the SIGMETRICS/Performance’09 conference. The actual error rate found was several orders of magnitude higher than previous small-scale or laboratory studies, with 25,000 to 70,000 errors per billion device hours per megabit (about 2.5–7 × 10?11 error/bit·h) (i.e. about 5 single bit errors in 8 Gigabytes of RAM per hour using the top-end error rate), and more than 8% of DIMM memory modules affected by errors per year.
>>43043950 > Do Xeon processors offer advantages over i7s?
Have you ever run into something you know would work better with a Xeon? Are you running a high end workstation or server? Do you need moar cores? If the answer is no then there's no advantage for you.
An i7 is great for most users and for the purpose of gaming actually works better than a Xeon.
>>43045334 Or you could go full retard and spend money for two processors or more cores for the hell of it.
>>43045302 A lot if you optimize your application for it. CryptoNight mining demands a lot of cache for instance, a bit more than 2 MB per core. The algorithm totally tanks if you don't have that much cache.
Image and video processing would be another example. You want that image to fit inside the L2 cache
>>43045155 >>43045175 not him but my energy bill was in the hundreds every month. Switched a few lightbulbs and got rid of the ol' Xeon home server, replacing it with a low-voltage i3 (fanless). Saved myself over a grand per year. >hurr muh poorfag dur ok
>>43045302 Depends on a large number of factors. In some applications having a fuckton of L3 cache is greatly useful. In others it serves little purpose other than allow for on die communication between cores. L1 and L2 cache are far more critical though, as they are much faster and lower latency, and thus closer to the CPU cores in the memory hierarchy. Toms hardware did a review with an Athlon II quad core and a Phenom II quad core, set both CPUs to the same settings (clocks, voltages, ram speeds etc) and pit them against each other. In some cases the Phenom II pulled ahead, sometimes by a significant amount, othertimes it was neck and neck.
>>43045287 Cosmic rays and flaws in manufacturing can be a right bitch. Especially cosmic rays. If the ray itself doesnt do the damage, the subatomic shrapnel of it smacking into the atmosphere at relativistic velocities definitely will. Most cosmic rays move with a velocity > .95c, and the fastest one detected to date was a single proton that hit with the kinetic energy equivalent to a baseball moving at 60mph.
>>43045466 > The terms “producer” and “consumer” come from economics and its treatment of material products. Thus, using them leads people to mistakenly apply to copiable digital data all that they know about the economics of uncopiable material products. Of course, this error is exactly the one proprietary software developers want people to make. Were not talking about digital data though were talking about physical items autist
>>43045464 >Cosmic rays and flaws in manufacturing can be a right bitch. Especially cosmic rays. If the ray itself doesnt do the damage, the subatomic shrapnel of it smacking into the atmosphere at relativistic velocities definitely will. Most cosmic rays move with a velocity > .95c, and the fastest one detected to date was a single proton that hit with the kinetic energy equivalent to a baseball moving at 60mph. this is either the world's best shitpost or a brief introduction to an extremely interesting topic
>>43045582 Implying I don't. This is not workstation vs gaming GPUs.
There's nothing stopping a Xeon from performing better than a core series processor with any type of task, Intel doesn't gimp them for one thing or the other like nvidia/amd do with workstation/gaming GPUs, they only add benefits to the Xeons.
>>43045503 Ok then. On cosmic rays: They are a variety of subatomic objects, ranging from single protons to entire atomic nuclei to even antimatter. What they all have in common is that they are moving at very very high speeds, typically fast enough that relativity and all the weird shit that happens to matter moving that fast occurs. Now, most of the time these subatomic particles and shit are flying around without a care in the universe. The bulk will miss our solar system entirely. Some will hit the sun's magnetic field and be deflected back into space. The faster ones will penetrate the field but miss our planet. The extremely (un)lucky ones will manage a successful strike against our planet's atmosphere.
Now, you have to keep in mind, these particles are hitting a fairly dense layer of matter at energies that would make the LHC weep with envy. When they hit, they and whatever unfortunate molecule that got struck basically get torn into their constituent components and fired deeper into the atmosphere towards the surface, still moving at high velocity. These fragments continue to "shotgun" deeper into the atmosphere, shredding atoms as they go. Many will stop before they reach the surface. Some though will make it. Cont next post A link for reading material in the meantime: http://en.wikipedia.org/wiki/Cosmic_rays
>>43045532 Of course it handles loads great. It's a great processor. But the i7 Extreme is great too, and it's great in slightly different ways.
Maybe you don't realize what I'm talking about. I don't mean this: >i7-4790, LGA 1150, $340 http://www.newegg.com/Product/Product.aspx?Item=N82E16819117369&nm_mc=AFC-C8Junction&cm_mmc=AFC-C8Junction-_-na-_-na-_-na&cm_sp=&AID=10446076&PID=3938566&SID=
I mean this: >i7-4960X, LGA 2011, $1050 http://www.newegg.com/Product/Product.aspx?Item=N82E16819116938&nm_mc=AFC-C8Junction&cm_mmc=AFC-C8Junction-_-na-_-na-_-na&cm_sp=&AID=10446076&PID=3938566&SID=
The i7-Extreme is what happens when Intel gets out the A list and makes a desktop processor out of it.
>>43045733 Cont: And heres where the problem lies. Each one of those fragments will still pack one hell of a punch for their size when they end their journey. To a human or other organism living on the surface, we can take a hit from such an event and basically shrug it off. To a computer though, such a shower of subatomic debris can be hell. See, transistors nowadays have very very fine tolerances when it comes to their construction, with some of the layers involved being only a couple atoms thick at best. Lets throw in the wrench that is the aforementioned shower of subatomic debris moving at a significant fraction of c. Suddenly said transistors are getting the shit whacked out of them by debris passing through, knocking atoms from the crystal lattices and generally causing chaos. Most of the time the transistor can continue to function after a hit. Sometimes though the impact is enough to cause the transistor to switch its state, which can lead to silent corruption and data loss. That server running in the stock exchange doing high speed trading just had a shitfit from that single memory error, causing millions in losses while the system restarts. Or that rendering project you were running just had a parameter changed that shouldnt have been, corrupting the whole thing. Basically, cosmic rays can fuck up a computer, and the problem is getting worse as transistor sizes continue to fall.
Oh, and a bit on that super particle: http://en.wikipedia.org/wiki/Oh-My-God_particle
All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.
This is a 4chan archive - all of the content originated from them. If you need IP information for a Poster - you need to contact them. This website shows only archived content.
If a post contains personal/copyrighted/illegal content you can contact me at email@example.com with that post and thread number and it will be removed as soon as possible.