Egad

You have a LGA 1150 mobo and a 1151 CPU? I hope for your sake that was a typo.

If a new 4790K is 340 euros, I vote don't do it. I'd want more than 90 euros off for a chip whose only test was he did was watch a rig boot.

Part of this of course is how much you trust him and how much this is a business norm for your area. For me if I heard a guy was selling hundreds and hundreds of parts I'd expect to see a Facebook page with reviews from happy customers, etc. Something to show he's not just bullshitting on his history, but that's just me. The second questions is, if this i7 turns out to be bad and you've sold your i5 to your friend, how screwed are you? Can you get another i5 or do you no longer have a functioning computer. Because if you can't afford to replace the i5, don't do this deal. Ultimately it's risk tolerance, the i7 is better but used is always a risk. You can also sit on your current i5 and watch i7 prices come down as more Skylakes, Kabylakes, etc release and 4th gen hardware gets cheaper. By the way what does that i7 cost new for you?

@mattebad All you have to do is have your drivers detect the other team and enter 'Petulant Bitch Performance Mode' or have the card's BIOS react. The manufacturer still controls clock speeds, memory timings, etc. nVidia already does with with turning off PhysX, Hairworks, etc whenever they detect that AMD Catalyst drivers are running on the system (meaning an AMD card is present).

I'd take that deal if you can get there in person and see the CPU in action. Bring up AIDA64 on it and run for a bit or really any tester that isn't Prime. If it's stable, take the deal. Yeah you need to do your GPU, but 40 euros for an i7 is totally worth it. You can worry about the GPU in the summer when next gens launch (or whenever you have the money for it really). A 760 is getting old, but isn't not obsolete.

Yeah the flight mechanics weren't particularly impressive in the beta for me. So unless they had things in the engine they weren't showing, seems like it will definitely suck.

So theoretically if your running an i5 (4 cores) and your program is only using two, you could see slight lag if suddenly your program needed two other two cores to do something and report it immediately. Theoretically being the operative word. Games tend to just jump on all four cores/all the cores they need at the first loading screen, so all the cores are unparked. For long running renders or other tasks that suddenly spawn more threads, the unpark time is so small it doesn't matter in the grand scheme of things. In theory in a fast twitch game with poor optimization you could see it stutter for a second because it had to wait for an unpark, but I think these days the proper response is to yell at the developer to get their shit together, not both with registry hacks.

We need a new Star Wars flight sim. Take all the best parts of the X-Wing series and the (superior) TIE Fighter series, stick them together, and release it. It's even up EA's ally because they can give in to their inner urges to be utter bastards and charge extra for the TIE Defender DLC, etc. Or a chrome plated X-Wing, or whatever.

While cool, one concern I have is that they only did this with comparable Tier GPUs. I'd like to see a test where a GTX 960 or such was tossed in with a R9 390, just to get some nVidia features. If that type of setup works it might be more attractive for AMD and nVidia to allow this to continue, as it gives them a foothold in the competitor's system. Such as a guy with a R9 390 seeing a GTX 9## on sale and grabbing it for Hairworks, PhysX, etc. However when it is comparable cards together I wonder if AMD and nVidia wouldn't just take their chances on trying to force you to buy two of their cards. nVidia especially right now since they're in a stronger position with a pair of nVidias working on lower wattage power supplies and having better linked GPU support. Perhaps more realistically though this speaks to the fact iGPUs might not be wastes of space and if you're suddenly using your iGPU suddenly your RAM speed, channels and amount matters more since system memory can often limit iGPU performance. Microsoft already did a tech demo where using the iGPU for post processing tasks boosted Unreal Engine's FPS by 4 FPS. Intel has much less incentive in being a bastard toward any discrete GPUs you install. It will be interesting to see if enough engines get supply for unlinked mixed to the point were we buy RAM different (namely buy dual or card channel kits rather than buy one stick, run single channel, upgrade later) and if a 390X can come hang out with my 980Ti SLI. Or two 980Tis and two 390Xs, with the identical cards linked and thus to the system you only have two GPUs.

Get the i3, although ideally if you have room in your budget, am i5 would yield some improvement. For your FPS goals and budget though an i3 will work. Couple AMD owners here like float into these threads and start rambling on about edge cases where more weaker cores is better (since that is what AMD is). The games you cited aren't the kind that see a major benefit from additional cores. In additional the AMD CPUs have no upgrade path. You can get a reasonable priced motherboard for your i3 that can also support an i5 or i7 from the same generation if you need a stronger CPU. Whereas if you go out and buy a top of the line AMD CPU, it/'s already outclassed by an i5. AMD is releasing a new line of CPUs in 2016, but they'll need different motherboards. So all the CPUs AMD are selling right now are dead ends.

4th gen i5, unless you need Skylake IPC. If your client has a research grant or such with a budget for equipment, talk to him or her about buying multiple i5s and picking the best. Or just being aggressive with voltage so when they die, they just get replaced. One thought in passing, is what is the length of the task. Because if this is a long running program that shoves stuff back in RAM, you have a balance between speed and each time you go the RAM there is a potential error. Minuscule but if the code does that extensively over a long period of time it's a threat. At that point there might be the need to consider ECC even if it lowers the rate at which the task is completed.

Yup, most FX-8350s make up to high 4s or 5 Ghz. You can also pay a premium for a 9350, which is just a factory overclocked 8350. Normally not worth it but if clock is a priority you can consider paying it as a sort of binning. In A10-6800K reviews a decent number of reviewers talked about hitting 5 GHz. Since this forum is generally 'I want to use my CPU for gaming or video editing' the OC doesn't offset the IPC lag, but of course it all comes down to exactly what you're running.

I would suggest you reread the part where he said a non intensive thread. Read up here. If he's confident the code is written such that the execution phase is not bottlenecked by the IPC of the AMD chips because he's doing a simpler instruction set, then more cycles is king, not IPC. If you're doing one stupidly easy thing, IPC stops mattering and you just want the most cycles per second.

I just occasionally switch my 980Tis out of SLI mode when playing a game with a shit SLI profile (you can do it via the control panel, no opening the case needed). I just set my second card to be the PhysX card (next to useless I know) and play the game without any problem. Most SLI issues though are people who have to play on day one. Really poor day one SLI performance is still very much a thing. It can take weeks or months for the publisher to patch in good SLI support. If you have two 980Tis though you'll be fine, since one Ti alone will handle 1440p. It's becoming fairly uncommon for devs to leave SLI permanently broken.

FX-8350, a quality board, and a good AOI cooler is something to consider if as you say it doesn't execute many instructions per clock. If he's only doing a few things and then reacting to them next clock this is probably the best way to be assured 5 GHz aside from a Pentium anniversary. Otherwise take a 4790K, a great board, plenty of cooling, and overvolt it unless the client needs to run this for years without interruption. I assume he wants the task done and doesn't care about chip lifespan.

That's about it. Basically hope there was some goofiness when you added the 780 to the system and doing just the 970 to start will clean up any driver or optimization oddities. If not, bad card probably. On a side note, PhysX isn't really worth it, only a couple games ever really made extensive uses of it. If your case has sufficient ventilation you can stick the 780 in purely for fun once you have your 970 up and running. However realistically any minimal benefit the 780 gives you is probably more than offset by the degree to which it heats your case up.

For reference, if you build it yourself you can get this: PCPartPicker part list / Price breakdown by merchant CPU: Intel Core i7-5820K 3.3GHz 6-Core Processor ($369.99 @ SuperBiiz) CPU Cooler: Corsair H110i GTX 104.7 CFM Liquid CPU Cooler ($129.99 @ Newegg) Motherboard: Gigabyte GA-X99-SLI ATX LGA2011-3 Motherboard ($183.98 @ Newegg) Memory: Kingston HyperX Fury Black 16GB (2 x 8GB) DDR4-2133 Memory ($101.99 @ SuperBiiz) Storage: Samsung 850 EVO-Series 250GB 2.5" Solid State Drive ($84.89 @ OutletPC) Storage: Western Digital Caviar Blue 1TB 3.5" 7200RPM Internal Hard Drive ($52.33 @ OutletPC) Video Card: XFX Radeon R9 390X 8GB Double Dissipation Video Card ($413.98 @ Newegg) Case: NZXT S340 (Black) ATX Mid Tower Case ($59.99 @ Micro Center) Power Supply: EVGA SuperNOVA G2 650W 80+ Gold Certified Fully-Modular ATX Power Supply ($89.99 @ Amazon) Total: $1487.13 Prices include shipping, taxes, and discounts when available Generated by PCPartPicker 2015-10-24 20:17 EDT-0400 Either go from a 390X to a 390 or drop the SSD if you money for a new Windows license. You can also step down to 8 GB of RAM if you want. If you do an all NCIX build it would cost a bit more and you might need to drop a couple of features. You could also go Skylake platform instead of X99 in this price range and upp down an i7-6700K. Seriously they should ship you come lube with the fuck job your builder is giving you. 1,500 and you only get a 4th gen i7, a 970, and no SSD.

Whoever is saying that is incorrect. R9 390 is the stronger card. If you absolutely refuse to build your own, really you should build it and get a much better X99 rig for 1,500, use NCIX and their assembly it for you option: http://www.ncixus.com/products/?sku=7842 Which is a mere 50 bucks instead of the markup other vendors will charge.

Redo the drivers withou the 780 present, full reinstall. Remove the 780 Uninstall all nVidia drivers. Reboot Go to nVidia.com and get the latest drivers Also make sure you have every connector on the GTX 970 plugged in. If you still have problems after all this, it's probably time to RMA it (or return it to where you bought if possible).

What's the sale price of a 970 vs current price on a R9 390 in your neck of the woods? 390 is better, but at a sufficiently attractive sale price the GTX 970 is still worthwhile. More powerful card first, SLI second is always the way to go.

Back plate. Or for the cheaper solution, mount it lower, buy a can of black plastidip, ask for extra chopsticks next time you order Asian takeout and cut them to size, then paint them.

GTX 970, Adobe says a lot of shit about a lot of things and only some of it comes to pass. OpenCL looks better and Adobe did some OpenCL improvements back in May, but CUDA still has the edge right now. I hope more OpenCL work comes out of Adobe and everyone else, but I hope for lots of things. Honestly if you do lots of editing, rendering, etc then no GPU is going to last you four to five years. The time saved by getting a new one puts most professionals on a yearly or biyearly upgrade cycle. 4 to 5 years would just be painful. We do a couple of our primary non 10 bit render boxes every six months or so and it pays for itself.

I care about power usage to the point I'll try to overvolt a 512 MB niVidia card to avoid running my HD, but I bought a FX CPU. sigh You're probably better coming at this from the other end and coming up a MSI Afterburner profile that undervolts and detunes your HD for your classic games. Load that profile for games and load its normal profile for modern games. Lots of bitcoiner miners would volt down down into .7 to .85 volts with clocks in 600 to 775 MHz range to save power and temp output.

Given the state of AMD's finances right now, I'm not exactly convinced we'll see them invest serious amounts in improving their Linux drivers. The're bleeding cash right now and the reality is a very small segment of the gaming market is going to run out and simply abandon Windows for Linux/SteamOS. It's cool that ~1,500 Steam games work in Linux, but they all work in Windows (aside from maybe some weird OS X exclusives? Can you buy pre Halo Bungie games on Steam maybe?) As others have said it's not surprising that AMD blows here given their general Linux support. Hopefully AMD steps it up, but realistically it is in their best interest to wait and see if nVidia and Valve can make SteamOS a success. If it's gaining market share they can go out and put in the time and money to make their Arctic Islands offerings work well.

@fabiosapplou Problem is GeForce and Quadro drivers aren't the same. You can make it work but you'd have to do some tweaking and you probably lose 10 bit color in the process. In these scenarios lowest common feature tends to win. Typically people will pair a weaker, but still decent, Quadro with a 'gaming' card and run them both on GeForce drivers. For sanity's sake I'd suggest a M4000 SLI. Or if your specific software has good multi GPU support you can unlink them and let the software push out jobs to each GPU as it sees fits.