RTB

I'm doing good.

If you want to test your maximum bandwidth using wireless is often a poor choice. Wireless g has this theoretical speed of 56mbit, which is in practice at the very least halved. Wireless n might have enough bandwidth to cap out a 75/35 connection, but only if the laptop is really close to the router. The ethernet result shows that something is wrong. It can be a hyperactive anti-virus, a virus, a cheap router unable to hit 75mbit, congestion, routing at the ISP side, or a bad signal for the modem, etc. Definitely worth investigating, since you're not even close to what you're paying for.

Much depends on where you need the system for, so I'll just go for the generic details. They're both refurbished and probably returned for a reason. Both have Nehalem CPUs, which offer good performance but at a cost. The W3520 has a TDP of 130 watt, and if the cooling system was skimped on, it will overheat. The other system has 2x60W CPUs, which should be easier to cool. The dual socket system has 8 slower cores, so it will only provide a notable speed increase over the single socket system if the software you run is heavily multithreaded. Oddly enough, both systems are triple channel yet run 4GB, so that's dual channel at best. The single socket system has a mere 500GB harddisk, and the dual socket one is anemic at 250GB, which would probably require you to replace it. The FX1800 is based on the peculiar 8800GS, and I have no idea if it's useful in any way.

That would not solve the issue, and would essentially give you a single average number that is better left as a set of numbers for more information regarding routing. There are two speed graphs that I think are most important for an internet connection: a graph of speed tests to an internal server (within the ISPs network) to determine the average and variance during the day/week/year/millennium of your connection to your ISP; and a graph of speed tests to a server outside the ISPs network, to find out how good your ISP is at handling that. Latency tests can also be very useful. You are right in that a single server can be hampered by a single bad network, but it does point to a problem that is not supposed to exist, and will affect many more servers which you want to interact with.

It's safe to say that x16 gen2 isn't fully utilized yet either. Depends on the game and your CPU power. One 7970 is going to be overkill for 99% of all games out there, with exceptions being stuff like Metro 2033 and extreme battles in Starcraft 2 (CPU limitation). Averaging over many games, they're about equal. Nvidia gets CUDA and physx, because it owns both sadly. Custom water cooling is going to provide better temperatures at a much higher cost and maintenance. Then there are the closed-loop systems like the H100, which tend to be just a bit better than top-end air cooling. The advantage of such systems is that you don't have a huge block of aluminium/copper hanging on your motherboard. At this point, 8GB is downright affordable, so for a high-end system the goal is now 16GB (2x8) or even 32GB (4x8). It'll be overkill if you don't overclock your CPU to 4.5+Ghz. The only time that wouldn't be overkill would be OC'd CPU + OC'd crossfire 79070s. Sounds like you wouldn't have much use for an expensive high-end rig, it would be better to save the money. That's just the official maximum. You can run the RAM at 1866 speed, but you'll probably have to set it manually in the BIOS.

Let's see how well it works: As opposed to east coast server: A little surprised that it won't go faster to Washington. Would it be an idea and even possible to include a small ping test inside the download/upload test, so one could see whether latency is the bottleneck or not?

It's not especially hard to drive two or even three displays of that size, but it's still a very welcome upgrade to the laptop screen market. If only this was done 5 years ago, imagine how much further monitor specs would be compared to current day screens.

The stock market value isn't the same thing as a company's worth though, it's what speculators think it's worth.

I can't see how you come to your conclusion at all. Both aircooling and liquid cooling get temperature problems when a fan dies or a pump dies. The CPU will shut down/throttle either way. In terms of endurance, a fan should give out way earlier than a pump, which more or less justifies the price difference (which is not as big as you indicate by the way). In terms of moving heat away from a heat source, liquid is a superior medium to air by miles. Regardless of how you cool your CPU, you still need a fan or two to get sufficient airflow in the case to make sure the remaining components (VRMs in particular) get enough airflow so they maintain efficiency and don't overheat. If you liquid cool your CPU, you are making sure that the heat generated by the CPU is transported more efficiently to a different location, keeping temperatures down at the CPU sockets, but at the same time you don't get the airflow going down on the CPU socket that is typical for aircoolers. I don't really see how the tubes interfere with airflow that much more than the usual cable mess in a case, but oh well. The testing I've seen indicates that average aircoolers are beaten by average closed-loop watercooling sets (Corsair H50 for example), top-end aircoolers can get close to closed-loop watercooling sets, and that custom watercooling is in a different league altogether. Finally, the temperatures you post are not realistic unless the ambient temperature in your house is about 5c or less. This is because both ordinary aircooling and ordinary watercooling cannot ever result in temperatures below ambient, it'd be a violation of the laws of thermodynamics, and we don't do that here at testmy.net. AMD is known to have inaccurate temperature sensors since the introduction of the original Phenom CPUs, so don't trust that.

In that case, I'm going to second getting a trackball. If you then fiddle with the orientation and placement of it you can very much limit any movement of your elbow and wrist, which should be beneficial for you. By the sound of it, you're not in a state anymore where you care for competitive gaming.

Given how specific your requirements are, you have the choice between making it yourself, or buying one for 30 bucks. Perhaps the cheapest method is coating your desk with a material that is similar to mousepads.

Slowdowns with wireless are normal, due to too much interference from all sorts of devices. You should check signal strength (with a program like inssider) or the connection speed whenever it's lagging. Which router are you using?

It's rare to see anyone using Bulldozer, considering that they hardly match Phenom 2 CPUs. Since the introduction of turbuboost (and the AMD equivalent), overclocking has become a bit tricky, because the CPU can be in various modes which all have different voltages and clock speeds. Considering how cool you're currently running, you should certainly be able to get it to run at 4ghz on all cores. The convenience of overclocking inside windows is useful, because you're not waiting for windows to load over and over, but it also means that settings don't stick after a reboot. So once you find a stable OC, you'd have to set it properly in the BIOS.

Worse, the channel expired. So it's not his bot doing this, it's the network. I guess #tmn died.

If option 1 is "SSD, single card" and option 2 is "no SSD, SLI/crossfire", then you really want option 1. You will notice the speed of the SSD on everything you do on that computer, as opposed to better average fps at the cost of higher energy consumption, which you won't even notice if the fps is over 60ish to begin with. As for 80 gold versus 80 bronze: label Load-> 20% 50% 100% 80 PLUS Bronze 82% 85% 82% 80 PLUS Gold 87% 90% 87% 650 watt, single 580 idle: ~130 watt load: ~400 watt bronze idle real wattage: 130 / 0.82 = 158.5 bronze load real wattage: 400 / 0.84 = 476.2 (84% assuming graph is linear for about 61% load, which it is not, but close enough) gold idle real wattage: 130 / 0.87 = 149.4 gold load real wattage: 400 / 0.89 = 449.4 (61% load -> 89% efficient assumption) difference at idle: 158.5 - 149.4 = 9.1 watt difference at load: 476.2 - 449.4 = 26.8 watt if kept on 24 hours a day, difference at idle: 9.1 * 24 = 218.4watt = 0.2184 KWh thus yearly, the difference is at least 0.2184 * 365 = 79.72 KWh which is roughly 79.72 * 0.15 dollar = 11.96 dollar a year likewise for full load 24/7/365 it's 26.8/1000 * 24 * 365 * 0.15 = 35.22 dollar/year[/code] Note that a 7970 is more efficient at idle and at load, so you won't even need a 650w PSU for it.

Why this change? What would a Dell U2312HM cost there? Feature comparison here. Seems to be a fairly decent PSU, for SLI 580 you'll need 1000 watts too. What is this? Afraid to fry it? For 1080p gaming this is way overkill, so again it's the question if you want to fold 24/7 at huge speed/cost or you want the power for gaming. Note that a single 580 is enough for now, and that it's cheaper to keep the 500 dollar and buy a new videocard when it's no longer sufficient. Of course, the only reason to get a 580 at this point is for folding, because the 7970 is plainly better (and still overkill for 1080p).

Sadly, they run sandforce controllers. Just be aware that it's a cheap PSU, and will not do the "80 plus gold" advertised. I'm of the opinion that there isn't such a thing as a quality TN screen. It's just absurd that in almost a decade there has been so little progress for monitor technology. Given that the manufacturers received fines for price fixing yet continued on their merry way after that, I doubt we'll be seeing good progress from them ever.

Yes, air cooling works fine, and some towers are on par with Corsair's closed-loop sets. Looks like an average tower aircooler to me, and it's really only 'the best' there because it's being compared to coolers I've never even heard of before.The more famous ones like the thermalright ultra 120 aren't listed. In the end the choice is yours, but I'd still like to stress that Corsair is only one of many companies that sell drives with sandforce controllers. Likewise, the 500+MB/s read and write is a theoretical maximum and only applies to very compressible data. Last time I checked, RMAs for such drives are unusually common, whereas Intel, Crucial and Samsung have much lower return rates. Had to look it up, and this is apparently a rather dodgy brand. Such a pity that 80+ certification is already being cheated with, because it had a decent start. It's still worth it to restrict PSU choice to supplies with the certification, because the good PSUs will have it anyway, and the bad ones will fail during reviews. The question is, what are you after in videocards? You plan to get a cheap 1080p screen, so a single 580 is pretty much overkill to begin with for gaming at that resolution. Don't think that video cards don't last long nowadays, as video games are mostly console ports now, so they don't have excessive graphics. If you're after folding performance, of course SLI 580s will be huge, but that comes at a price. 24/7 folding with two 580s is like 600 watt plus PSU inefficiency, which will cost quickly cost more in electricity than the price of the video cards themselves. Not to mention the heat.

I haven't looked into intel's smart response stuff, so you'll have to ask someone else. If it's a caching thing, then I doubt it's going to make a difference, especially with a small SSD. Custom watercooling is going to take maintanaince, but there are coolers like the Corsair H80, which have a closed loop, so they doesn't need any. For traditional HDDs SATA3 offers no real benefits, but there's probably no price difference at the same size harddisk anymore. Sounds fancy, which means it probably doesn't do much If you say you want the system to last as long as possible, consider a blu-ray reader. More fancy words for a fan You'll want the SSD as dedicated boot drive, not as caching thingy. You also picked an SSD with a sandforce controller, which are not that reliable. If you want the machine to last, you should look for Intel 510, Crucial M4, or Samsung 830. Those are considered significantly more reliable, and are latest generation drives. The size depends on how much you want to install on it, 64GB might be a bit small. In general Intel's sandy bridge CPUs are not that sensitive to memory speed, I only mentioned that for serious video editing work. 1600mhz is probably a bit cheaper. Power supplies are the most efficient when supplying a wattage at or above 20% of their maximum. This means that you'll want to pick a PSU for which the idle wattage of your system is at about 20%, and full load is at roughly 50%. Even though you plan on folding with the card 24/7 I still think that 1000 watts is overkill, and 800-850 watt would be fine. Secondly, at high wattage the efficiency rating of the PSU becomes more important in my eyes. The difference between 80% and 90% at full load for your system is quite a lot. If possible, look into adding a PSU that has a better 80+ certification. At this point I find it difficult to recommend a video card for gaming sake, but if you're after folding speed then I guess the 590 can't be beat. It's still a dual GPU card, which I find iffy, because average fps can be deceiving.

I don't see the advantage there, both chips are more than sufficient for 'multitasking'. The reason to go with a 3930k would be to render CPU-limited things faster if your time is more important than the extra 300 dollar investment. Though, if you go that way you'll want more and faster ram, something like 1866mhz 16GB. The one thing you didn't add which helps multitasking more than a bigger CPU is an SSD. At this point I can't really recommend sandforce drives due to their instability, which leaves the Intel 510 series, Crucial M4 and Samsung 830. If I'd have to pick between investing 300 dollar in the CPU versus an SSD, I'd pick the SSD for sure. Likewise, both CPUs are unlocked and OC nicely, but you picked the no 'OC option'. Do you plan on OCing it yourself?

Your choice of video card really depends on what resolution you play at. At the moment pretty much any midrange current-generation card from nvidia/ati can run any game at 1920*1080 and the higher models have no problems with 2560*1600. As for the CPU, currently there's no real need to go beyond an i7-2600k or even a i5-2500k with a good cooler to run either chip at 4.5Ghz or so, because as you increase the graphics settings any new game becomes mostly GPU-limited. I'm no fan of crossfire/sli because a single card is always more efficient, and it's less buggy too. On the laptop side, there isn't much to specifically recommend as it's a real mess. As long as you get a 1366*768 screen (which by the way is pathetic in various ways) most midrange cards should work, if you go 1920*1080 things get tricky because the card needs to be beefier and that causes lots more heat and power usage. For CPU you don't have to look beyond intel's i5-2xxx or i7-2xxx series, which can be dual core (with hyperthreading) or quad core (with and without hyperthreading). If you care about battery time nvidia seems to have the advantage with Optimus, but there is no way to see if a laptop has implemented that before buying and testing it. Optimus shuts off the video card entirely reducing idle usage. Ati has a similar technology but doesn't advertise it as well, and it's hard to find info on which laptops have it. Edit: Note to self: when checking out a thread, don't forget to click the second page before replying. Facepalm-inducing, really.

Or wait for the spambots to take over, whichever comes first

Can't believe I missed this thread. I will test your server CA3LE, mark my words.

Or maybe he's messing with us... This might be photoshopped, or he's running a program to simulate a broken screen. I don't see any shards of glass on the keyboard.

Actually, having both an internal and an external server lets you find out more specifically where a problem might exist. If the internal server is slow to begin with, the problem is local. If it's fast, but the external server is slow (might have to test at multiple locations) then there's a problem with peering. Of course... this is assuming such a bog-standard flash test bothers to do an accurate test. Of course, a fast internal test only says something about internal speed, and claims that it's an accurate representation of browsing speed are merely lies. The reason why 90% is considered perfectly fine is that sending data over the internet has overhead. There are headers/footers added to data as it's being sent/received, and those count in the maximum throughput but not for download speed, because they're not part of the data being sent/received.