Why should gamers go 64bit?

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
And whatever happens to the cpu that doesn't support 64 bit? I still have an old p3 and AMD chips? But the system's are mainly for internet?
Click to expand...

Just like 16bit 286s and 386s, they will slowly fade into the sunset.

No one has addressed my Conroe point yet. Does anyone know if Conroe will be an EM-64T processor or if it has a new 64bit technology for Conroe? Or if Conroe might not be so fast in 64bit mode because if EM-64T.

Yes, it will be 64 bit.
http://www.theregister.co.uk/2004/03/17/intel_plots_4mb_l2_64bit/

I know Conroe is going to be 64bit, the question (that wusy is working on) is, will Intel's version of the 64bit technology hinder it in 64bit mode or have they corrected that yet?
That's something nobody can answer until it releases...


Weren't the 386's (not sx) the 1st 32 bit procs?



Intel 386DX/SX/SL/EX etc and AMD 386DX CPUs are 32bit

The 386SX is 32bit but has a 16bit bus.
 
Ok, in a serious note, I think the end justifies the means.....I don't care who has the best 64bit design as long as they perform better then the other that's all that matters in the end. By the way, I thought that amd64 and emt64 were the same extensions. I believe that amd and intel have a clause that dates back to a 1995 supreme court decision that states that amd can use the i386 microcode as long as certain technologies can be shared, i.e. sse and amd64. Makes you wonder what's really going on.....
 
Ok, in a serious note, I think the end justifies the means.....I don't care who has the best 64bit design as long as they perform better then the other that's all that matters in the end. By the way, I thought that amd64 and emt64 were the same extensions. I believe that amd and intel have a clause that dates back to a 1995 supreme court decision that states that amd can use the i386 microcode as long as certain technologies can be shared, i.e. sse and amd64. Makes you wonder what's really going on.....
Click to expand...


I understand Intel -- for the lack of a better word -- "borrowed" AMD's x86_64 extensions renamed them EMT64 and used them in their CPUs.

I think AMD was rather flattered.
 
Ok first thread and might as well start a war here.

Why should a gamer go with 64bit? Most games dont support it yet. You have to get a new OS and MoBO. The 32bit chips are faster than the 64bit in 32bit mode. IE get a P4 3.6 for 300 bucks, or get an AMD dualcore 4800 (which only runs at 2.4 in 32bit mode) for 600 bucks.
Click to expand...
Most of the answers in this forum were correct.

And, yes, as a gamer you should currently look at getting a good AMD Athlon 64 3500+ or higher for today's games. Dual core is nice for other tasks, and theoretically for future games, but by the time games start being well optimized for multithreading you'll want a new processor anyway.

What 32 bit chips are faster than the 64bit chips in 32 bit mode? Sorry to dissalusion you but the AMD64 chips have been king in 32 bit mode since they were introduced.
Click to expand...
The reason for his disilllusion is likely because when running on a 64bit platform such as Windows XP x64 Edition... 32bit applications are run through an API that acts like a translator. Since the OS kernel is 64bit and the game is 32bit, it's not as fast as it would be running on a 32bit OS.

However, this has nothing to do with hardware support and the difference disappears when you run 64bit games on a 64bit OS.

It is true that for gaming today you do not need to go to a 64bit platform. However, 64bit chips do not cost any more. Hell, even the $70 Celeron D chips are 64bit. The hardware support of these newer chips improves the reusability and the Total Cost of Ownership of the chips.

The only speed increase you're going to get with games on x86-64 is by using more than 4GB of memory(including PF) and utilizing the extra registries which AMD64 has which every 64-bit game will anyway.

EM-64T DOES NOT have those extra registries to increase speed in x86-64 environment.
Click to expand...
Intel EM64T is functionally identical to AMD64. It does have the extra registers. They are the general purpose registers R8-R15 as well as the SSE registers labelled XMM8-XMM15.

Please keep in mind that x86-64 is basically a processor instruction set language. It is not a widening of internal pathways, bus widths, etc. On the physical side it is merely the addition of extra and wider registers which only help performance when programs are rewritten and optimized to use them. And even then, other factors will likely bottleneck their use.

When comparing AMD64 to EM64T technologies from the standpoint of AMD vs. Intel, keep in mind that the evolution of the x86 architecture is not a technological feat. It is a slow and easy progression. The extra registers and widened registers of this architecture only added about 224 bytes to the processor die, whereas a Conroe chip houses over 4 million bytes of cache alone.

Simply put, Intel did not introduce 64bit architecture prematurely because there was no need at that time. (That, and they wanted to promote their wonderfully architected by overly priced Itanium chip.) By the time Microsoft Windows XP x64 came out, Intel's 64-bit chips had already hit the market. Just in time. And yet even now Windows XP x64 is almost never used by the average consumers because the software they use will not benefit.

If Intel had wanted to adopt AMD's "64bit is faster" marketing technique they used when the Athlon 64 debuted, they could simply redesign their processors to support the future x86-128 and implement it with backward compatibility. Of course this would be pointless and stupid (unless you want to be able to address a concurrent 281474976710656 YiB (yes, yobibytes) of address space.

I understand the implications that the chicken and the egg theory has to this, which is why I will go over the advantages of a 64bit execution environment.

The two advantages to x86-64 are what I will call extended addressing of memory and storing values with a high dynamic range.

1. Extended Addressing

This is the obvious advantage which most people understand. Standard 32bit processors can only access as many as four binary gigabytes (4 GiB) of data. x86 uses byte addressing, and does not address individual bits, which improves the addressability (grouping data bits into clumps of eight made a huge difference back in the 16-bit days). Here are some examples as to why this can improve performance:

Say you have a large 13GiB database running on a server and you need to be able to address the data contained inside it. With a 32bit CPU, you would have to split the addressing up into chunks of 4 GiB each. You would have to perform software address translations when reading and writing to the data which would cause a lot of CPU overhead, especially when accessing data which is spanned across multiple chunks. Of course, much business software like Oracle is optimized to do this very efficiently, and avoids storing concurrent data structures across these chunks. However, a 64bit platform provides a huge performance improvement.

Similarly, say you are a home user who wants to open up a HD video recording of 5-15 GiB in size. The same situation applies, only most non-enterprise grade software is not nearly as good at handling these platform limitations and therefore take a performance hit.

The other benefit to this kind of operation is that 64bit code no longer uses segment registers to access offets in memory. Normally on even a 32bit program running in a 4 GiB address space, the program's memory was divided into seperate segments. At least one for code, one for data, etc. Even though these segments overlapped the same 32bit address space in flat addressing mode (which is the mode modern apps and games run on), the overhead was still there. Any time an instruction referenced something, it was either implied which offset it was referencing from, or it was overrided by what is called an instruction prefix. The downside to such a prefix is that it increases code size, which takes up more space inside the microprocessor's internal pipeline as well as its cache memories.

In 64bit mode, an address is an address is an address. Address translation is still done by the processor, at the hardware level in order to translate the 16 EiB (exbibytes) of logical program memory into real physical memory such as your 2 GiB of system RAM.

2. High Dynamic Range

High dynamic range is the lesser advantage of the two.

Computer data is basically used to represent data in the real world. Like a person's age or the dimensions of an object. One example of data is audio representation. Audio is basically represented with 16bit/24bit values because these are more than large enough to individually represent the range of hearing a human being is capable of. And to my understanding, even 24bit is overkill at that.

However, in scientific and engineering, and graphical applications, this dynamic range is very useful to respresent very precise real numbers.

However, most multimedia does not need this higher dynamic range. This is why 32bit colour, 32bit textures, etc. are about as high as they are going to get. Human eyes can only perceive so many unique colours.

When talking of bitness in terms of, say, a 256bit graphics card, that simply means groups of smaller data elements (each normally 8bit, 16bit, or 32bit in size) are able to transfer from the GPUs registers and buffers to your graphics memory faster since the bus is wider. Don't be confused and think that means that software bitness equals speed.

However, in relation to the needs of dynamic range in scientific and engineering applications, I believe I have been misleading. To better explore the uselessness of x86-64 and what it offers for a dynamic range to store data, let us consider that:

- It widens the general purpose registers to 64bits, but they are used only for addressing and integer arithmetic, not floating-point arithmetic.

- We have had 64bit registers since the original 80486 CPU with the introduction of the x87 FPU.

- Since the introduction of the original SSE, x86 CPUs had 128bit registers in order to store these values of high dynamic range. However, they were rarely used for that.

Almost all applications that utilize any form of Streaming SIMD Extensions (SSE, SSE2, SSE3) are using them for their actual purpose which is storing 8bit, 16bit, 32bit, and 64bit values together in one big register and performing a single calculation on all of them at once. That's why they call it SIMD (Single Instruction Multiple Data).

Therefore, in conclusion, large virtual memory addressing is the main reason for the move to x86-64.

All criticisms aside, I support the change. There are almost no 32bit only CPUs on the market anyhow, not in current lineups. However, I think 64bit pushing zealots need to be shot for their ignorance.
 
Never mind the brand of processor, intel or amd...i'm looking at the 32bit prices over the 64bit prices. I'm looking for the debate over why a gamer should spend extra money going 64bit when games run at 32bit. The processors I used where picked for pricing matters. Should a gamer spend extra money on going 64bit when most games and even upcomming titles run on 32bit? Yeah there are a few games out there that run on 64bit, but I can only count those games on one hand.

continue to discuss.....

Cataclysm
Click to expand...
The best thing with AMD 64 CPU's IS NOT because it's a 64-bit CPU. It's because it has a integrated memory controller. Intel CPU's doesn't have that. The integrated memory controller is what makes AMD 64 so fast. And you can't count GHz when comparing different types of CPUs. When it comes to gaming a cheap AMD 64 3200+ (running at 2GHz) will kick the P4 3.6GHz in many games. The 4800+ dual core will eat the P4 3.6GHz...
 
Right now, there are no 64bit games out. And now CPUs are coming out 64bit, but it doesn't mean the 64bit fearure is used.
Click to expand...

- Half-Life 2 (Currently single player and lost coast only)

- Shadow Ops Red Mercury

- Far Cry

- Unreal 2004

- And many others

are all offering x64 modes, so long as people have a x64 Operating System to run them on.
 
I read somewhere that the amd64/emt64 was stolen from the nintendo64......just goes to show you how far we have come.
Click to expand...

Yeah, we've come so far people still believe everything they read on the Internet.

eg: "I read it on the Internet, therefore it must be true" 😛
 
Vista should support 64 bit from the ground up, which will help in the adoption of 64 bit code being used in programs and games in the future. It also, idealy, should be much more compatable than XP-64, since that is more a hack job on XP to get it to support 64 bit, instead of being built form the ground up.

64 bit is not automaticly faster than 32 bit, but it has the potential to be, particularly when dealing with large numbers or addressing large amounts of data. However, most consumers won't see the benifits of 64 bit processing for another year or two after Vista is released.

Will games adopt it? of course they will eventualy, but only when it makes sense to, or for bragging rights in FPS. Much like games will adopt multi-threading to take advantage of dual cores. Will it benifit every game? Well, not every game is a huge FPS intensive game.

Oddly enough, where dual cores will help the most will be in simulation games like Out of the Park Baseball, where complex calculations take place to simulate events, and often make the player wait till they're finished to move on (Simulate a day/week/year of baseball games for example). In most graphic intensive games, a good GPU will trump just about every other performance factor, be it clock speed, 32 vs 64 bit, duel vs single core, or what not.

Ultimately, 64 bit is more marketing than need at the desktop level, but still, consumers will benifit from it taking place now well before it's needed, allowing for a much smoother transition.
 
It’s like comparing a Harley with a 1000cc motor to a Honda Crotch Rocket with a 750cc motor. I promise you the Honda’s faster, because of the engine, weight, and design of the bike.
Click to expand...
This is a bad analogy for your example because the Honda is faster becuase it has a smaller, lighter engine, that spins at about 10,000 rpms, producing lots of horse power vs. the 5000 of the Harley. The Harley has lots more torque but doesn't spin as easily.
This is acutally the opposite of the processors, the AMD has more torque, aka it does more work per revolution, and the Intel has more horsepower, aka it can move the wheels really fast.
For gaming you want more torque because they are really heavy apps, which is why your GPU does lots of calculations per clock and not lots of clocks.
 
Huh? According to their graphs they're faster with 64 bit apps under a 64 bit OS and marginally slower with a 32bit app with a 64bit OS, woh woh woh!

I've seen benchmarks where they've been faster, slower and the same, couldn't be bothered digging any up for now.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom