>>45719485 FX-6300. FM2+ is the only other socket option available and that is only used for Kaveri. I would stick to AM3+ if you want performance now or wait until the next socket comes out and get the first processor with a die shrink.
>>45719485 860K is a 7850K without the GPU cores that make it an "APU". If you want a slightly more up-to-date platform than AM3+ and you don't mind a less expensive but less powerful processor, an A88X board is probably what you want.
>>45724207 Because the durability and power advantages are nearly useless in a desktop, sleep mode eliminates the need for booting up, and the few seconds shaved off loading things isn't worth losing about 80% of your potential storage space.
>>45724207 desktop drives are usually faster with a larger cache compared to laptop drives.
>>45724491 That's because you suffer from HDD distortion field syndrome. The fact that you think we need to prove SSD's are far more useful then HDD's for anything but storage makes you retarded. This is already a majority opinion. You prove to me that using a HDD is better. You're literally ignoring one of the most beneficial parts of your build for >muh HDD's Literally babyboomer mentality
>>45724491 Because you're a fucking irrational autist who tries to dismiss facts by downplaying them. SSDs are faster, and that isn't a minor benefit like you try to play it off as. The entire system feels more responsive. When moving large files, many files, or decompressing something an SSD is drastically faster than a mechanical disk. An SSD transfers no vibrations into your case. An SSD will last longer than a mechanical drive will. An SSD will never have a motor die during spin up. An SSD will never have performance issues arise as the result of file fragmentation. Reliability rates of SSDs outpace mechanical storage, and should you ever need to do a system restore an SSD will allow you to have your system back and running in just minutes instead of hours.
>>45724569 >An SSD will last longer than a mechanical drive will. >An SSD will never have a motor die during spin up. Except that SSDs will just run out of writes eventually giving them a limited life span.
>>45724645 Then where are all of your games going to go?
>>45724569 >Reliability rates of SSDs outpace mechanical storage, and should you ever need to do a system restore an SSD will allow you to have your system back and running in just minutes instead of hours. That is however one advantage SSDs have over hard drive, you can install your OS faster, which you almost never do.
>>45724661 >Except that SSDs will just run out of writes eventually giving them a limited life span. Not in my lifetime You really have no idea do you? 256 GB drives are under 100$ right now. That leaves more then 200 for your games. >implying you're going to replay AAA console ports that eat up 30+ GB each Remove them after you're done with them faggot Chances are you have a harddrive laying around anyway.
>>45724745 Socket FM2+ has no high performance parts, and its uncertain if it will receive any new chips. Socket AM3+ doesn't have a comparable chipset, but you have a selection of higher performance CPUs.
AMD doesn't have anything to offer on the desktop unless you want a cheap and quiet HTPC, a little mail server, or a cheap video editing build.
>>45725874 HSA can do so many little things that its ridiculous. You can use it to have the IGP render file icons instantly when you open up a folder. Little things like that seriously go a long way towards making a system *feel* fast.
For like 99% of common workloads a cheap HSA system with an SSD is perfect, whether its an APU or an ARM SoC. Though people really need to start branching into acronyms with more than three letters. Jesus.
>>45725957 That third bench, POV-Ray, its just an error that Anandtech never bothered updating in their results database. Kaveri scores a lot higher than Trinity in POV-Ray, and Trinity was about on par with the Q9650 in said bench.
In performance per clock the intel chip is ahead, but ultimate performance is what matters, how you get there doesn't.
I want this for a budget build, but why do I feel like I'm going to purchase their end-of-the-line overstock. Not to mention, if I wait, it could possibly be buying 1st gen garbage of the new socket....
>>45726112 Its not even 2015 yet, and the Zen chips won't launch til sometime in 2016, an optimistic estimate would be Q2 2016. Can you wait a year and a half to build a new system?
An FX 6300 now is a fantastic deal, but it depends on what you're upgrading from, and what you're planning to do with it. You can always wait for the next best thing to come around, but if you do that you'd be waiting forever.
>>45726161 I've waited 7 years. I'm not sure 18 months would be a problem. >You can always wait for the next best thing to come around, but if you do that you'd be waiting forever. Aye. I should know better. I'll get a new R9 gpu too.
>>45726204 That's KAVERI. It's an APU, we all know that you need fast RAM for an APU because it has no dedicated VRAM. It's a pretty fucking specific configuration that's only good for low-end work anyway, nobody is playing Crysis 3 on an A6 or rendering videos on an A8. If I have an i7 or an 8350 the benefit of buying 2400MHz RAM over 1333 or 1600 is absolutely zero, it's a total placebo. The most gain I've ever seen in-game was 2% in Battlefield 4 going from 1600 to 2400. Paying out the ass for some XMP 3000MHz shit is a total waste of money.
>>45726502 I was just shitposting, but you really cannot tell anyone that you can humanly see the difference between 1600 and 1833 ram on any system with a dedicated cpu and gpu. APU of course, but he wasn't talking about that.
>>45726587 A couple different firms have produced 3ghz Cortex A9 based server chips as of a year or two ago. A 3ghz Cortex A57 chip would absolutely ruin the current Kabini/Beema and Baytrail chips. The Exynos 5433 based ODROID coming out in a few months will be it. Stock clock will be 1.9ghz, but you should be able to hit 2.5ghz with a little heat sink.
rocking an fx-9370 and loving it. Have it paired with an r9 290. Getting a 2nd 290 soon. Plays everything I want. minimum 45 fps. I Might be buying the new x99 system from intel though. I've been using AMD since phenom II was big and I'd like to see what intel offers. I have an Ivy bridge pentium running my file server as is. Very energy efficient and runs cool. I don't know. I'm just really hoping AMD releases something for the enthusiast market soon. Because having to clock my 9370 to 5 ghz or higher to match a 3.5ghz intel chip is just embarrassing.
>>45726678 Its not really about the architecture in itself, but more so about how many cores the program is designed to utilize. Look at Photoshop, Sony vegas, etc. They love moar coars. FX-83xx chips shine in those scenerios. Games? Usually 2-3 cores at most. The architecture is crap because of its horrible single core performance, not that it is harder to code for. Intel is merely faster at single core. Same could be said for intel's hyperthreading. Few programs actually use the 4 hyperthreading "cores" in an i7.
>we need moar cores >but how? these K10.5 cores are pretty big, getting 8 of them on single die with all the required cache is hard >nah, let's just use two weak integer units with shared FPU and cache and call it 2-core module >brilliant, what could go wrong :^)
>>45726713 2ALU/2AGU and terrible front end prefetch/branch prediction. The whole Bulldozer family wastes a lot of cycles dealing with miss penalties. Steamroller isn't nearly as bad as Bulldozer, if it had L3 cache it would be pretty decent, but you might as well wish to win the lotto.
The front end that feeds the cores, and the heart of the cores themselves, the execution pipelines, are extremely limited in Bulldozer family chips. Not just the Flex FPU but the integer cores themselves. They were designed originally with the intention of creating a super small core that made some sacrifices in order to improve compute power per mm2. Unfortunately however the Bulldozer module ended up only being a couple mm2 smaller than 2 K10.5 cores with L2 cache, and the K10.5 cores have much better performance per clock.
>>45726699 I've been in contact with the hardkernel guy for like 5 months now, and we've been talking about their future dev boards. They plan on adding more IO in future models, more USB 3.0, and a USB type C port. There was some talk of e-SATA too. The ARM SoCs are definitely getting to where they need to be in order to be a viable desktop system, and the IO is getting there as well, ARM software support just has to improve and everything will be golden.
>>45726671 intel chips are meant to be used with ddr3 1600 mhz. AMD chips memory controller can handle 1866 as a standard. Anything above 1600 for intel is considered an "overclocked" speed, regardless of what speed you bought it at. The chip itself can actually limit how fast of RAM you can actually use. You might buy 2400mhz RAM, but find the PC won't boot with anything over 2133. Kind of like how every chip is different with overclocking requirements. Voltage, max speed, heat output etc.
as >>45726747 said, 2133 ddr3 is no longer insanely expensive unless you're buying riced out RAM such as corsair dominator platinums. Then you're paying $30-$40 more. Usually faster RAM is about $5-$10 more than 1600 RAM of the same company. G-skill usually has good ram for decent price.
>>45726840 didn't say it did. Faster RAM can actually increase the score of synthetic benchmarks, but that is pretty much it. But, look at the picture in >>45726822, can you honestly tell me you'd buy the slower RAM? For $4 you get faster better binned RAM which can always be downclocked (usually with tighter timings than lesser speed variants) if need be. At that price point, its a matter of "why not have the faster RAM?"
>>45726764 K10.5 cores aren't big though. Two K10.5 cores with 2MB of L2 cache @32nm is 32mm2. One BD/PD module with 2MB of L2 cache @32nm is 30.9mm2.
The uncore necessary to link the K10.5 cores together is completely marginal, and could have been reduced by other means. The decision to use the Bulldozer arch was 100% a management mistake.
>>45726838 This played a huge role in things as well, it ended up hurting Llano seriously as well. GloFo's 32nm PD-SOI node had horrendous yields and binning issues with cut down the clocks of every chip produced on it. The Original Bulldozer chips were supposed to have a stock clock in the range of 4.5ghz inside of their TDP. Llano could barely hit 3ghz on the node without power consumption exploding when K10 chips on the prior 45nm process could hit 3.7ghz without major issues. A sub par foundry under delivering coupled with absent minded management is like the recipe for the perfect storm of mediocrity.
>>45726826 Whenever someone takes the risk and fronts the cash for it. Theres literally no technical limitation in place, it all comes down to whether or not they'd be able to sell enough units to recoup the design costs and then turn a profit. Give it a year or two and I think we'll see larger form factor ARM systems than these current dev boards.
>>45726877 ah well that is a predicament most don't face. Nothing wrong with kingston, they're usually just expensive. >>45726886 I've nothing to say. You're saying you'd pass up on better hardware for $4. If you're penny pinching to that extent, building a computer is not for you.
>I'm no rich money blowing person either. I built my system on an AMD platform for the price:performance. but $4? really?
I haven't built a computer for myself in the last 6 or so years but I've spent over 31,696.36 USD building computers for other people.
Anything from sub $300 to $1.5k+ builds, sometimes bulk orders, yes, fuckers will complain about every fucking penny.
Anybody from home users, to multi factory printing presses to advertisement design, fucking factory owners driving BMW X6s blowing hundreds of dollars on a single meal, yet complain about the price of a single dimm fucking ram.
>>45726990 >I personally think there is no tangible ceiling for the iGPU scaling with ram speed yet. I would love to see a ddr4 APU. >>DDR4 4000mhz. >>Getting 55+ fps @ 1080p on high with what amounts to a $400ish build. This will be in the future most definitely. Sadly, not yet. For now, one must shit out cash for a $300 room heater to play new games.
>>45727017 I know what you mean. I build computers for people myself. Guy I work with didn't bat an eye when I told him the final price of his work station PC he was using for photo and video editing.
>fx-8320 with hyper 212 evo >gtx 660 ti (was on sale, good for 3 monitors and cheap CUDA accerleration) >750 watt PSU >2 23inch 1080 monitors >3 2TB hdds >2 250 gb crucial ssds >1 128gb crucial boot ssd >32gb ram
He spent $1300 for everything.
Other co worker wanted an office build with ssd and I manage to plan out a $400 build that would do everything he wanted and more with reliable parts >scoffs at the idea of paying my $80 to hand pick the parts, go get the parts from the store on my own time in my own vehicle using my gas, build it, install windows 7 (that I provide), and give him a fully functional PC with all anti-virus etc.
>>45727051 I only get $32 USD per computer build, and people try to talk me down to $9.5
Fucking third world shit holes.
One of the worst jobs I took was reinstalling a computer, rewiring an office network and then having to somehow re-setup their email server when they didn't know the fucking account and password to their own fucking servers.
Had to call their past tech guy and luckily he kept a record of that shit.
All for $16 bux.
Needless to say I don't really take on those jobs anymore. I do translating work now although that only nets me $300 USD average a month.
BBA degree with full honors and a minor in English means jack shit.
>>45727050 On a dual channel IMC, 4000mhz DDR4 would give you 64GB/s bandwidth. The Reference Radeon 7750 has 72GB/s if I recall. However with the color compression implemented in Tonga you could have an effective 89GB/s which would give you enough bandwidth to more than adequately feed 8 GCN 1.1 CU, you could get away with 10 even.
HBM however is the answer to the IGP memory bottleneck, and we won't have it on an APU until the Zen chips launch.
Pic related, the DDR3 variant of the Radeon 7750 is very close to how the Kaveri A10 7850k performs with 2400mhz RAM. Just giving those shaders adequate bandwidth would net a huge performance increase, though even that isn't enough. For appreciable gaming performance you need more CU, and that requires even more bandwidth. HBM is the only answer.
I wouldn't bother, I'm as big an AMD fanboy as you'll ever meet and even I had to go with Intel. I couldn't wait on my Phenom II X4 955 BE for AMD to final pull their thumb out their ass and make a decent chip that isn't just MORE CORES
I do have a R9 290 though, cause I got an insane deal on it at the time and the 900 series hadn't been released yet
>>45727118 did it because he knows his programs were CPU intensive and knew the CPU fan would go crazy trying to keep it all cool on the stock cooler. First guy was my ideal customer example. i.e. not an idiot. Not everyone overclocks their CPU but 2 120mm fans spinning at 1k rpm > 70mm fan spinning at 4k rpm any day.
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.