When is 64bit OS coming?

[xandros9]

?

He was asking if you read the previous posts that contained information regarding 32 vs 64-bit.

Your assertion isn't quite correct, but it isn't completely wrong since those cases can overlap and swap.

 

[Grant Taylor3]

He was asking if you read the previous posts that contained information regarding 32 vs 64-bit.

Your assertion isn't quite correct, but it isn't completely wrong since those cases can overlap and swap.

He came across as rude for no reason.

 

[MDK22]

He came across as rude for no reason.

Your perception notwithstanding ...
Because your over-simplification was not entirely CORRECT ...

 

[milkyway]

He came across as rude for no reason.

I'm sorry if I was rude but you came into the thread and replied with information which was already (partly) debunked

 

[EspHack]

I know this is off-topic, but can you tell me if there's a significant impact in terms of PC performance if I upgrade my PC's RAM to still the same amount of RAM, but faster clock time. How significant is the performance gain?

I thought of asking this since I just saw YouTube recommending a gaming video asking the question: "MORE RAM or FASTER RAM?". It was late at night so I haven't watched it yet. Will probably watch it later after work.

faster or bigger ram you ask? depends...

if you're talking about gaming, and your graphics depend on an integrated chip, like an AMD APU or simply the standard "HD" graphics embedded to core i intel processors, then your ram SPEED will have a huge impact on grapfics(fps) so if you go from single channel to dual channel memory for example, you will see an enormous improvement most of the time, graphically speaking, since that kind of graphics chip is using your cpu ram, that's why you always see you have X amount of ram, but only a lesser amount of "useable" ram, the missing part is dedicated to the integrated graphics

now if its gaming with a dedicated graphics card, like one going into a pci-e slot, then ram speeds wont change a thing most of the time, BUT, if your cpu is being a limiting factor, faster ram will help it too, lets say you run a game at 40fps no matter what graphics settings you set, that means your cpu probably cant handle anything more than that, so if you're on single channel memory and install more ram, making it dual channel, or simply swapping slower 1600mhz ram to 2400mhz ram, your cpu will get faster, just not as much as an integrated gpu would

by the way, there's also "timings" for ram speeds, and they tend to balance out the overall ram speed, sadly, lets say you have 1600mhz ram and 1866mhz ram, the 1866 ram will probably have worse timings, meaning your actual bandwidth will be very close if not the same as the 1600mhz one, so no real upgrade there

 

[a5cent]

faster or bigger ram you ask? depends...

Ho hum... and once again I've fallen on my face trying to make things simple. I should know by now that every answer should start with "it depends". Good point on the importance of low-latency high-bandwidth RAM when using integrated graphics. I don't know a single gamer who relies on integrated graphics, so I always miss that angle, but it's an important aspect, particularly for people who game on laptops and can't just plug in a dedicated GPU.

 

[EspHack]

Ho hum... and once again I've fallen on my face trying to make things simple. I should know by now that every answer should start with "it depends". Good point on the importance of low-latency high-bandwidth RAM when using integrated graphics. I don't know a single gamer who relies on integrated graphics, so I always miss that angle, but it's an important aspect, particularly for people who game on laptops and can't just plug in a dedicated GPU.

yeah things get complicated in that regard, with products such as AMD's APU which essentially put a GPU die inside a cpu, they tend to use gpu chips that are actual "dedicated" pci-e ones, maybe not the highest end ones, but discrete enough for decent gaming, and then we have the current gen consoles which are essentially an APU on steroids with GDDR5 ram, its a very clever choice, the cpu part could get a substantial boost thanks to that which makes comparisons with it and the would-be standalone cpu part impractical since the standalone cpu & motherboard can't use GDDR5

 

[Ariel Takom]

It does make a difference, but the differences aren't huge. If you can go from really bad RAM to great RAM, it might be worth it, but only if you do a lot of stuff where every bit of performance counts. I'd recommend reading this:

DDR4 Haswell-E Scaling Review: 2133 to 3200 with G.Skill, Corsair, ADATA and Crucial

I see. Thank you for the reply! I'll have a read at the link you provided.

 

[Ariel Takom]

faster or bigger ram you ask? depends...

if you're talking about gaming, and your graphics depend on an integrated chip, like an AMD APU or simply the standard "HD" graphics embedded to core i intel processors, then your ram SPEED will have a huge impact on grapfics(fps) so if you go from single channel to dual channel memory for example, you will see an enormous improvement most of the time, graphically speaking, since that kind of graphics chip is using your cpu ram, that's why you always see you have X amount of ram, but only a lesser amount of "useable" ram, the missing part is dedicated to the integrated graphics

now if its gaming with a dedicated graphics card, like one going into a pci-e slot, then ram speeds wont change a thing most of the time, BUT, if your cpu is being a limiting factor, faster ram will help it too, lets say you run a game at 40fps no matter what graphics settings you set, that means your cpu probably cant handle anything more than that, so if you're on single channel memory and install more ram, making it dual channel, or simply swapping slower 1600mhz ram to 2400mhz ram, your cpu will get faster, just not as much as an integrated gpu would

by the way, there's also "timings" for ram speeds, and they tend to balance out the overall ram speed, sadly, lets say you have 1600mhz ram and 1866mhz ram, the 1866 ram will probably have worse timings, meaning your actual bandwidth will be very close if not the same as the 1600mhz one, so no real upgrade there

Thanks for the reply!

Btw, could you explain more regarding the timings of the RAM you mentioned at the end of your reply? Or maybe provide a link where I could read up on it? This is my first time tbh to hear of the RAM timings.

And once again, thank you for your reply!

 

[a5cent]

^ RAM timings are discussed in the article I already linked to. Otherwise just searching for RAM and timings already gives you a good selection to choose from. This was at the top of my google search:



http://www.hardwaresecrets.com/understanding-ram-timings/

 

[Cruncher04]

From what my lecturer told me in a coding class I took as an elective, 64-bit allows the OS to utilize more than 3-4GB of RAM, and that's basically the only advantage it has over 32-bit (he did say there are some other advantages, but this is the main one).

Just to be clear here. This proposition only holds when everything else being the same. However we are talking about ARMv7 Thumb2 vs ARMv8 AArch64. There are quite a few differences outside of a pure address range extension, which impacts performance.
So you lecturer should have put a few disclaimers next to his statement. Possibly your lecturer was specifically referring to x86 vs x64/AMD64...who knows.

 

[Grant Taylor3]

I'm sorry if I was rude but you came into the thread and replied with information which was already (partly) debunked

Thank you for saying sorry.

 

[unichips]

Just to be clear here. This proposition only holds when everything else being the same. However we are talking about ARMv7 Thumb2 vs ARMv8 AArch64. There are quite a few differences outside of a pure address range extension, which impacts performance.
So you lecturer should have put a few disclaimers next to his statement. Possibly your lecturer was specifically referring to x86 vs x64/AMD64...who knows.

Heck, even with the IA32 vs AMD64 transition, there were other factors at play. Most notably, in 64-bit mode, an x86_64 CPU has more both larger general-purpose registers (obviously), but also presents the programmer/compiler with more of them, than that same CPU would offer in 16- or 32-bit mode. (Which, In turn, would be the same number of general-purpose registers as any native IA32 or older processor.)

In some circumstances, for example tight loops iterating many times over a small set of data, you can get better performance out of the increased number of registers, due to a reduction in the frequency of CPU-to-cache transfers and cache-to-RAM transfers.

Of course, this distinction may not be particularly noticeable in the general case.

 

[Cruncher04]

In some circumstances, for example tight loops iterating many times over a small set of data, you can get better performance out of the increased number of registers, due to a reduction in the frequency of CPU-to-cache transfers and cache-to-RAM transfers.

This can be more generalized. If you look at x86 (assembly) code, you literally see memory references all over the place, because you are lacking architectural registers. With other words symbol values hold in registers have an extreme short live time. When looking at programs at the function level, symbols can have an extreme long life time. The compiler generally determines the live time for each symbolic value and constructs an interference graph where adjacent nodes have overlapping lifetimes. If the architecture is lacking registers this graph cannot be colored. The compiler generates spilling code accordingly (or just using memory operands in case of x86)
I assume there is a sweet spot of how many architectural registers are optimal. I have seen interference graphs of some programs with a chromatic numbers higher than 32 (w.g. which means you need more than 32 register for optimum allocation) however at some point you have diminishing returns due to the cost of registers (both physically and in terms of instruction set coding space). Looking at different architectures my guess is that 32 GP registers is such a sweet spot.

Speaking of IA32 vs AMD64 another advantage is that the ABI has changed from passing functions arguments per stack to passing function arguments via register.