Future Computer Parts / General Computer Discussions

[kitamesume]

^sweet dongle, sad though its quite expensive for a thumb-PC.

on that note i wish they stop doing that full stick HDMI plug-in, i mean if the port is on the back on the monitor the stick would stick out too much.

[halfelite]

most tv's have the one side hdmi port that makes it fit perfect

[kitamesume]

true, but quite a bit of others does have a back port, and quite a lot of monitors have back ports over side ports.

its annoying really x.x, but do they include extension cables just in case for those retarded screen designs?

[halfelite]

true, but quite a bit of others does have a back port, and quite a lot of monitors have back ports over side ports.

its annoying really x.x, but do they include extension cables just in case for those retarded screen designs?

I dont think it does. Coming from aliexpress even more so. If it retailed in the US I bet it would come with one, But I have a few um them extension cables around.

[3Rton]

Wait that is an actually functional full PC? O_O (Well actually functional... with those specs not so much but still)

Also worth it to ask for extension cable in the comments for the order (and send them email about it) it's 50% whether they'll actually send one or not but it's worth asking.

[kitamesume]

Wait that is an actually functional full PC? O_O (Well actually functional... with those specs not so much but still)

what are you talking about it being not so much of a functional PC?
its very good, imho barely slower than a mobile i3s at least, just make sure its the 2GB ram otherwise it's stuck with single-tasks at a time.

and for the record, 16GB internal (purely for OS) is very much fine, eMMC is very slow anyway.
best use case for this is buying a samsung evo C10-UHS1 64GB uSD for your apps, and a NAS for everything else.

[halfelite]

as stated yes its fully functional. You will not be playing no AAA title games on it but it should function for most stuff

[Engr.Alvin]

OMG AMD is falling out of the competition and now they are in the way to cut their manpower by 7%:

quoted!
The biggest problem of AMD today is inefficient Bulldozer micro-architecture of its central processing units, which is a key reason it has been losing market share and money in the recent years. Keeping that AMD will have to offer Bulldozer-based chips for at least 1.5 years now, it will hardly be very competitive against Intel. On the other hand, the company has a number of promising projects in the plans, including ARM-based server chips, semi-custom offerings and some other. What is not good is that AMD itself does not seem to believe that its initiatives will actually take off, which is why it cuts costs…

[kitamesume]

their low-power puma processors are quite good though, in terms of IPC its faster than bulldozer at least, and massively better in terms of power efficiency.
if only they could clock one to 4Ghz we might just see how this would fare at mid-tier chips.


edit: speaking of architectures, is it possible to merge two different processors?
e.g. ARM and Intel Haswell, not in a sense that you cram an ARM processor in an Intel chip, but integrate ARM instruction sets themselves.

[lololitas]

I don't think that's been done, and it would probably be either very unstable or very complex. Afaik instructions sets are usually either hardwired, installed as tables in on chip memory or a mixture of both. So adding an instruction set either means hardwiring more into the chip or adding memory. Maybe if soFIA manages to turn up a revenue we might see some dual instruction set processors.

[kitamesume]

yeah it'd be quite complex, they'd have to merge both pipeline to support two different group of instruction sets.

but from what i'm imagining, if they do manage to get it going, the results could be a boosted ARM-set performance.
ARM at the moment is on-par with baytrail, imagine if ARM is on-par with an i7.

and a highly boosted efficiency in terms of general tasks, since ARM instruction sets can run some of the codes at a far better efficiency than Intel-set could.
plus they could shut-off x86 instruction set while on power-saving mode, which in theory are more power hungry than ARM instruction sets.
though from where i put it at "windows" would run in a RT-like state.

[lololitas]

Is x86 more power hungry though? Or is ARM more efficient? Afaik there where tests comparing tablets with Atom Z3740 to tablets with the (now outdated) Snapdragon S4 pro and Tegra 3. The Atom won the CPU performance test by a wide margin with a power consumption of 2.9W, whereas the Snapdragon and Tegra won in the graphics department with 2.0/2.8W respectively. While it is hard to find accurate power consumption tests for ARM (apparently everyone blindly believes that they are more power efficient, so no actual tests are done), I did manage to find one side claiming that the Snapdragon 800 consumes 3W in Smartphones and 4W in Tablets.

The way I see it an Intel CPU core with one of the ARM graphics chips could work wonders if it doesn't explode while soldering them together.

[kureshii]

Is x86 more power hungry though? Or is ARM more efficient? Afaik there where tests comparing tablets with Atom Z3740 to tablets with the (now outdated) Snapdragon S4 pro and Tegra 3. The Atom won the CPU performance test by a wide margin with a power consumption of 2.9W, whereas the Snapdragon and Tegra won in the graphics department with 2.0/2.8W respectively. While it is hard to find accurate power consumption tests for ARM (apparently everyone blindly believes that they are more power efficient, so no actual tests are done), I did manage to find one side claiming that the Snapdragon 800 consumes 3W in Smartphones and 4W in Tablets.

You might want to take a look at one of Anandtech's articles on x86 vs ARM, which tries to delve into this power consumption/efficiency myth.

The x86 Power Myth Busted: In-Depth Clover Trail Power Analysis
The ARM vs x86 Wars Have Begun: In-Depth Power Analysis of Atom, Krait & Cortex A15

Mind that the articles are almost two years old now, and won't cover the latest architectures.

The way I see it an Intel CPU core with one of the ARM graphics chips could work wonders if it doesn't explode while soldering them together.

By ARM graphics chips, do you mean the Mali? Not the best of graphics options... The Merrifield Atoms are actually using Imagination PowerVR graphics, by the way. The Intel HD Graphics in the Z3770 aren't too shabby, although most of the benchmarks have pitted it against other x86 chips and not so much against ARM licensees.

[from]

edit: speaking of architectures, is it possible to merge two different processors?
e.g. ARM and Intel Haswell, not in a sense that you cram an ARM processor in an Intel chip, but integrate ARM instruction sets themselves.

Yeah. No. Two different thoughts - x86 is CISC, ARM is RISC. This means that ARM basically consists of a very small set of instructions, which is why ARM processors are comparatively easy to make small and cheap. This also means that ARM does not, and never will, support more "specific" use-cases, e.g. how Intel included the AES instructions to speed up AES encryption a few years back. This means cycle-for-cycle, x86 CPUs are generally much more powerful than ARM ones - for instance, in the AES case an ARM CPU will use ten cycles to one on a modern Intel x86.

The instructions themselves in either set, however, are called upon (named) in machine code using numbers. When you use addition, the compiler will (eventually) translate that into the number associated with the addition operation. These numbers mean different things in x86 and ARM. That is, the number "8" could be associated with the addition operation in x86 and the subtraction operation in ARM, which would cause problems.

This could be bypassed by recompiling the code for the new platform, which is generally how these things are handled.

So no, the ARM instructions can't be integrated into an x86 computing core. What could be done is the other case you mention but dismiss, integrating an ARM core on the same chip as an x86 one, so either handles instructions that it understands. It's highly unlikely it'll happen, because of recompilation and/or virtualization, but it shouldn't be technically impossible.

[kureshii]

their low-power puma processors are quite good though, in terms of IPC its faster than bulldozer at least, and massively better in terms of power efficiency.
if only they could clock one to 4Ghz we might just see how this would fare at mid-tier chips.

Except that’s now how things work. Try to bump up clock speed by raising the multiplier, and all you’re doing is minimising stalls. Your buffers and cache sizes and branch predictors and memory bandwidth and almost all other potential limiting factors remain much the same (without a change in base clock). If you want to clear instructions at a higher rate, you need an architecture that can handle more instructions in flight, not just find one that can handle a decent number at low power and then try to overclock the hell out of it.

Not to mention that the voltage bump you’ll need to bump up base clock speeds and hit 4GHz would pretty much wipe out all hopes of better power efficiency. If getting it to scale was so easy, AMD wouldn’t need a separate architecture for the mainstream/enthusiast desktop segment.

edit: speaking of architectures, is it possible to merge two different processors?
e.g. ARM and Intel Haswell, not in a sense that you cram an ARM processor in an Intel chip, but integrate ARM instruction sets themselves.

Since you weren't very clear in what you mean by "merging two processors", I'm just going to drop this link here: http://www.anandtech.com/show/6007/amd-2013-apus-to-include-arm-cortexa5-processor-for-trustzone-capabilities

Things like this happen all the time, by the way. If this is not what you meant, excuse my misinterpretation of your vague question.

ARM is also not a processor. It is an instruction set, or rather a family of them. When you say "an ARM processor", do you mean:

1) One of ARM's ready-made SoCs?
2) One of the various OEM SoCs that incorporate bits of ARM IP, e.g. Exynos, Opteron A1100?
3) Or one of the various OEM SoCs that are compatible with an ARM instruction set, but are their own architecture entirely, e.g. Apple A6/7/8, Qualcomm Snapdragon?

Some useful reading: http://www.anandtech.com/show/7112/the-arm-diaries-part-1-how-arms-business-model-works

[kitamesume]

quite true, but i'd still expect it to consume half as much power than bulldozer at the same clock.


i meant ARM processors as an entirety of processors using ARM instruction sets, like Intel processors.
and "ARM" itself is a corporation name imho.

on that regard, by merging two processor i wasn't saying to just slap in a working ARM core in an Intel chip.
but rather integrating ARM instruction sets into Intel's, could make a weird processor though.

if i remember correctly, ARM doesn't really discourage this sort of thing didn't they? Samsung, Nvidia and qualcomm had been making their own abominations imho.
i do wonder what'd it result with if ARM had fully beefed pipelines like those of Intel i Cores, and x86 instruction set.
the price though, quite likely will it be as pricey as their i core series.

edit: on a side note, from what i've heard Intel's strategy with android OS is using translated codes (e.g. written for x86 instead of ARM) to run android in their x86 chips.
and i heard that is it wasn't quite efficient, and quite the reason why clovertrail were a bad result, subpar and hella laggy even though windows runs quite nice on those chips.
so would it be likely if Intel grabbed an ARM license, made an ARM+x86 abomination with code compatibility result in a real abomination?

[from]

What could be done is the other case you mention but dismiss, integrating an ARM core on the same chip as an x86 one, so either handles instructions that it understands. It's highly unlikely it'll happen, because of recompilation and/or virtualization, but it shouldn't be technically impossible.

I'm just going to drop this link here: http://www.anandtech.com/show/6007/amd-2013-apus-to-include-arm-cortexa5-processor-for-trustzone-capabilities

Welp, shows how good I am at anticipating developments, and how up-to-date I am with AMD.

I'll stand by my non-crystal-ball thoughts, though.

[kitamesume]

but they're doing a co-processor integration though, wasn't quite the idea i had in mind.


http://wccftech.com/intel-skylake-processors-integrate-iris-gt4e-level-graphics-chip-features-72-execution-units-128-mb-edram-llc/
nice, i wonder if their entire line will have the 64MB/128MB cache, or is it only those GT4e?

http://www.kitguru.net/components/cpu/anton-shilov/intel-begins-to-send-skylake-s-processor-with-disappointing-frequencies-to-partners/
but this seems like of a concern.


http://techreport.com/forums/viewtopic.php?f=2&t=93911
on the notes of skylake, theres a speculatory thread about core configurations, i'd be very amused to see a quad-core i3.

(click to show/hide)

i3/i3-K quad-core without hyper-thread and 6MB? L3 shared cache
$150~$180 4C/4T = 2.4Ghz~3.0Ghz base clock | no turbo boost
$200 4C/4T(K) =  3.0Ghz with max 40multiplier (4Ghz)

and while i'm at it.
i5/i5-K quad-core with&without hyper-thread and 8MB? L3 shared cache
$180~$225 4C/4T = 3.2Ghz~3.6Ghz | +200Mhz turbo boost
$200~$250 4C/8T = 3.2Ghz~3.6Ghz | +400Mhz turbo boost
$275 4C/8T(K) = 3.6Ghz max 63multiplier (6.3Ghz) | +600Mhz turbo boost

i7/i7-K hexa-core without hyper-thread and 15MB? L3 shared cache
$300~$350 6C/6T = 3.4Ghz~3.6Ghz | +400Mhz turbo boost
$400 6C/6T(K) = 3.6Ghz with max 63multiplier (6.3Ghz) | +600Mhz turbo boost

note: take note that this is just assumptions, and a highly optimistic one.

edit: and something about their discussion reminded me of what i asked a few years ago, anyone remembered about the stuffs i said about a single-core supporting more than 2threads?
apparently there was an implementation of it - http://en.wikipedia.org/wiki/Simultaneous_multithreading - and xeon phi uses a 4-way multi-thread (4threads per core).

[from]

edit: and something about their discussion reminded me of what i asked a few years ago, anyone remembered about the stuffs i said about a single-core supporting more than 2threads?
apparently there was an implementation of it - http://en.wikipedia.org/wiki/Simultaneous_multithreading - and xeon phi uses a 4-way multi-thread (4threads per core).

Single-cores can run many threads, and generally do. Recall starting programs in any pre-dual core Windows, for instance Win98 - the OS keeps running in the background while the program runs, that's two threads. And W98 could handle a decent amount, not just two.

The most inefficient way of handling multiple threads is simply shunting all registers and caches used by one thread into RAM, load those of the new thread into the registers/caches and go boldly forth. Works, but we can do better. SMT, most famously used by Intel in their HyperThreading CPUs, is one way to be more efficient about threading. HyperThreading mostly negates the cost of swapping between two threads and can even allow some actual concurrency, which is nice. Other implementations (E.g., the Xeon Phi you mentioned) can do the same with even more threads.

[kitamesume]

yes, single-cores can run multiple threads but not simultaneously, this architecture in particular can make single-cores run multiple threads simultaneously.
its like a person writting a couple of essays, while a single person can write multiple essays, he cannot write them all at once as if having multiple writting hands and brains.

and to note, this "multi-threading" you're putting into example isn't exactly multi-threading at all, see preemptive multitasking.

where as in sumultaneous multi-threading.

Multithreading is similar in concept to preemptive multitasking but is implemented at the thread level of execution in modern superscalar processors.

In simultaneous multithreading, instructions from more than one thread can be executing in any given pipeline stage at a time.

PS: a few years back, when i asked about running more than two threads on a single core whether it'd be more efficient, everyone said it was an absurd idea.

edit: i cant seem to dig up that old thread (pun intended).
this one's a recent post, did they prune the other one?


edit2: speaking of multi-threading, if anyone remember a few pages back, some of us were discussing the effects of "more cores" with gaming.
and theres a tiny-bit of improvement with minimum frame-rate and frame-time with more cores.

to point out, and looking into certain processor architecture, or however you call that.
a barrel processor's pros and cons had pointed out something really interesting in a sense that most of us might've already forgotten about such issues.
http://en.wikipedia.org/wiki/Barrel_processor

Advantages compared to single threaded processors

A single-tasking processor spends a lot of time idle, not doing anything useful whenever a cache miss or pipeline stall occurs. Advantages to employing barrel processors over single-tasking processors include:

• The ability to do useful work on the other threads while the stalled thread is waiting.
• Designing an n-way barrel processor with n-deep pipelines is much simpler than designing a single-tasking processor because a barrel processor never has a pipeline stall and doesn't need feed-forward circuits.
• For real-time applications, a barrel processor can guarantee that a "real-time" thread can execute with precise timing, no matter what happens to the other threads, even if some other thread locks up in an infinite loop or is continuously interrupted by hardware interrupts.

notice the parts "stalled thread", "interrupts" and "realtime", these issues are indeed still present with current modern CPUs on both intel and AMD.
to experience it yourself, open taskmanager and choose a light but consistent load on CPU resource, and push it's priority to "realtime", this will cause everything including itself to lag mind you.

(click to show/hide)

now looking back, windows and other apps does cram more than a hundred threads on so few of your cores, you can see it in taskmanager's performance tab.
just one of these threads stalling can cause some microsecond jitters, and so does interrupts.
a few microsecond of jitter might not be much evident, but in terms of raw compute its practically wrecked the timing of each other threads in line.
and is also the reason why games stutters so madly with less than 4cores.

tl;dr - this might be why multiple cores feel a lot more smoother, regardless of overall performance.


on the topic of multi-threading, theres the GPUs as well.
if anyone recalled, the GPU has more than a hundred cores in it.
and though it appears a single core(not quite sure about this), it uses a very fast scheduler to spread the threads.
so a stream of threads is funneled through a single input, and a scheduler spreads it amongst the multiple cores.

but the graphics is quite unique in it's own, a single process contains thousands if not millions of threads by its own.
which means it can fairly and evenly spread itself amongst the hundreds of cores.
unlike your typical desktop process which only contains a single or some few threads by its own.


though from this, it makes me wonder how a CPU would perform when adopting that style of execution.
an example of this is the VISC processor where a single-thread, after breaking it down, could be spread amongst multiple cores.



the advantage of this, from the way i see it, is easier implementation of multi-threading of apps, whether it'd be more efficient than hard-coded multi-threads is another matter all together.


http://www.extremetech.com/computing/193000-the-alienware-graphics-amplifier-finally-desktop-quality-graphics-on-your-laptop
if they made this USB capable, though i doubt we'd get enough bandwidth or even a low enough latency to utilize a GTX980, maybe just a GTX750Ti might make things worth it.

http://www.extremetech.com/computing/192711-samsungs-new-20nm-ddr4-clears-the-way-for-massive-128gb-dimms
ridiculous, its large, too large, i wish i had the money for a pair.
this reminds me, has there been a review about DDR4's multiple dimm bandwidth?
apparently each slot is it's own channel, e.g. an 8slot ddr4 is an octa-channel.

http://www.extremetech.com/extreme/192929-255tbps-worlds-fastest-network-could-carry-all-the-internet-traffic-single-fiber
this, and the deep sea pipelines, just do it please.


http://www.tomshardware.com/news/philips-4k-bdm4065uc-monitor,27977.html
http://techreport.com/news/27290/new-philips-monitor-serves-up-4k-on-a-40-va-panel
oh my, its pretty nice to see 40" 4K going below $1K.

http://techreport.com/news/27277/amd-a-series-price-cuts-still-arent-in-effect
i get the feeling that AMD is pulling a sly one, announcing a "price cut" in advance (more than a few weeks) could technically stagger purchases on other items.
e.g. "ohh they're slashing prices, i'll hold off my i3 purchase for now."

http://techreport.com/news/27263/microsoft-quarterly-revenue-up-25-on-strong-surface-xbox-sales#metal
quite surprising that they're still being stingy about their consumer pricing, considering theres this much growth in revenue.
though i didn't know microsoft had a seer in their team, who would've guessed they'll rake in $23.2B revenue in Q1 2015.
i'm pretty sure its a typo though and this is a fiscal year budget.

http://www.tomshardware.com/news/lian-li-mini-itx-nas-pc-q26,27950.html
right, if only most ITX boards has more than 10sata slots, or BGA itx boards.

http://techreport.com/news/27295/get-ready-for-more-graphics-cards-with-8gb-of-ram
moar VRAM, i just wish game coders could make their engines load entire texture packs to saturate the entire VRAM regardless of size, think android RAM usage case.

http://techreport.com/news/27299/sage-microelectronics-enables-5tb-sata-ssds
dat SSD, and dat price, its very slow though at SATAII cap sustained speed, but at least the IOPs isn't bottlenecked by that.

http://www.globalpr.com.tw/press-room/lian-li/press-releases/article/lian-li/lian-li-unleashes-hell-on-the-pc-market-with-the-pc-d666-1/
dual system on a single case, NAS+mainsystem would work well in this, ridiculously overpriced though.

http://www.techpowerup.com/reviews/NVIDIA/GTX_980_PCI-Express_Scaling/1.html

(click to show/hide)

holy crap thats mighty low bandwidth consumption, whatever Nvidia did to it's compression algorithms its doing a very good job.
this card could technically run on the bottom 4x 2.0pcie slot without much issues at all.
now if only they step up the extra slots to 3.0pcie we'd see a lot less bottlenecking all together.
considering that 4.0pcie will be predominant on the next batch of motherboards.

http://www.phoronix.com/scan.php?page=article&item=linux_raid_fs4&num=1
ohhhhhhhhhhhh, SSDs benchmarked on linux using different filesystems.

http://wccftech.com/intel-launching-skylakes-unlocked-broadwell-q2-2015/
well, at least skylake-S is still in Q2'15, and here i was thinking that they had postponed it to Q4'15.

http://www.extremetech.com/computing/193628-what-does-a-cpu-do-when-its-doing-nothing
that was enlightening, i can see where this would go with windows 10 on tablets.
i do wonder though, if you can manipulate the interrupt timer to interrupt more often, will it make the system respond faster? at the cost of power efficiency of course.

http://www.extremetech.com/extreme/193794-new-phase-change-material-lights-the-way-to-all-optical-super-fast-computing
another vaporware?

http://www.anandtech.com/show/8496/dell-previews-27inch-5k-ultrasharp-monitor-5120x2880
mmmmmh, 5K. wasted on a bad-rep manufacturer, they haven't earned their credibility on the UHD market yet.


damn it, they took out the EA-550 from their sell list.

though on the other hand a seasonic 500BT ECO had popped out for $52, this versus the Antec VP550P V2.
anyone know whether this one is better than antec's? i heard it was based on Seasonic's S12 series, not the S12II.

edit: adding a few more PSUs on that list of local supplies.

does anyone know whether Antec EA-550 Platinum is reliable? well Antec in general too, curiously i found it locally for $81, and it's 450W version for $65.
according to tomshardware it seems well-built, minus some regulation quirks.

if its fairly reliable, it'd replace the FSP Raider 550W i have thats already rocking 2years now (almost 24/7 at that).
to list whats locally available to me though:

(click to show/hide)

Corsair VS550 = $50
Corsair CX500 = $64
Corsair CX500M = $73
Corsair GS600 2013 = $73
Corsair CS550M = $91
Corsair TX650M = $102
Corsair TX650 v2 = $104

Cooler Master Elite V2 500watts = $42
Cooler Master Thunder M 520W = $57
Cooler Master GX2 550watts = $68
Cooler Master Silent Pro Gold 550watts = $69

Seasonic 400BT ECO 400watts = $46
Seasonic 500BT ECO 500watts = $52
Seasonic 600BT ECO 600watts = $57
Seasonic S12II 520watts = $62
Seasonic S12II 620Watts = $75
Seasonic M12II-520 EVO = $75
Seasonic M12II-620 EVO = $96
Seasonic G550 = $96
Seasonic P-520FL Platinum = $125

Antec VP550P V2 = $48 techpowerup seems to regard it well, for it's ridiculously low price its an option for the NAS.
Antec TruePower TP-550 = $61
Antec VP650P V2 = $67
Antec HCG-620M = $74
Antec EA-550 Platinum = $81 tomshardware says its quite a good PSU, but hows the reliability?

EVGA 500B = $45

FSP Raider 550watts = $57
FSP AURUM S 600Watts = $87 interesting price point for a gold 600W PSU.
FSP AU-550M AURUM CM = $91
FSP AU-650M AURUM CM = $99

Hec Cougar RS 550W = $54
Hec Cougar PowerX 550W = $58
Hec Cougar PowerX 700W = $75
Hec Cougar CMX700 700W = $96

note: i'll have to point out that the seasonic platinum is far outside of what i can consider worth it, i'd find it worth something if it was at least $100.
PS: budget limit is $100 for a PSU, i could stretch it a bit more though it wouldn't be worth it considering S12II 620W is only $75, efficiency aside.
PSS: as much as i'd like to search and link every single one's reviews i'm not really that bored.

i think i just found something that'd simplify my project o.o
basically, i could possibly do a HTPC+NAS+Router on a Windows platform.

what tipped me off on this is TMeter, and connectify, well i only need an AP with DHCP and a Traffic shaper (QoS).

on that note security might be a concern, but it isn't really that bad though.