64-bit client

M

Maverick

Guest
Does anyone know if there will be two versions of Half Life 2 - for instance a 64-bit version and a normal 32-bit version?

I have seen it mentioned somewhere that HL2 will not support SMP, So does this mean I will not get a performance increase using a dual Opteron box?
 
look man 64bit cpu will not have any increase in performance for games, only application, i repeat only application.....:borg:
 
Originally posted by Gorgon
look man 64bit cpu will not have any increase in performance for games, only application, i repeat only application.....:borg:

What are you talking about buddy? For one, a game is an application. And anything will have performance increase as long as it has 64-bit optimization.
 
no budy, games are not application they are 3d application.......... application without the work 3d means business application....:cool:
 
Originally posted by Gorgon
no budy, games are not application they are 3d application.......... application without the work 3d means business application....:cool:

Ok thanks little buddy, How about you step aside and let the grown ups talk.
 
Yes, ther ewill be a 64bit port of Half-Life 2.

64bit processors increase performance across the boards. They increase speeds in everyhting
 
Originally posted by Maverick
Ok thanks little buddy, How about you step aside and let the grown ups talk.

I will kik ur ass when i get HL2 and in HL2 multi do not forget my name Gorgon my name in Multi will be Gorgon so rememebr me when i kik ur butt
 
increase performance only in business appli.....
for games would increase per only 10% and that is nothing
 
Originally posted by SidewinderX143
Yes, ther ewill be a 64bit port of Half-Life 2.

64bit processors increase performance across the boards. They increase speeds in everyhting

Umm no, it actually decreases speed in certain applications and in games which are gpu limited it wont matter if youre processor is 64-bit or 4504985ghz
 
Originally posted by Gorgon
increase performance only in business appli.....
for games would increase per only 10% and that is nothing

Please stop ruining my thread, You obviously have no idea what your talking about.
 
Originally posted by Gorgon
I will kik ur ass when i get HL2 and in HL2 multi do not forget my name Gorgon my name in Multi will be Gorgon so rememebr me when i kik ur butt

damn. That's worse than I spell.

Can you post some fact that say 64bit processors don't increase gaming performance much?
 
Originally posted by Gorgon
look man 64bit cpu will not have any increase in performance for games, only application, i repeat only application.....:borg:

Oh that's too funny.

Games are still 32-Bit applications.

64-bit optimized games allow for 2 times the bandwidth of 32-bit games.
 
Originally posted by reever2
Umm no, it actually decreases speed in certain applications and in games which are gpu limited it wont matter if youre processor is 64-bit or 4504985ghz

No- Your thinking of SMP with a tradational 32-bit CPU... Games that support SMP now would spilt Graphics, Sound, and AI into different threads - this is what causes performance decrease in some games.
 
but their will be no ****ing diffrence between IMAGE quality for 32bit and 64bit.........:cool:
 
Originally posted by Gorgon
but their will be no ****ing diffrence between IMAGE quality for 32bit and 64bit.........:cool:

Quality? Maybe not. Your processor lends a pretty big hand to your graphics card, so graphical performance increase? Definately.

And guess what kind of non-graphical processes are possible.
 
Originally posted by BlumenKohl
Quality? Maybe not. Image performance? Definately.

And guess what kind of non-graphical processes are possible.

I mage quality is the last think the player will see and enjoy.. not the image per....
 
Image quality is pretty limited without the necessary performance. :dozey:
 
true.... but trust me their will be no diff between 32bit game and 64....
 
Originally posted by Maverick
No- Your thinking of SMP with a tradational 32-bit CPU... Games that support SMP now would spilt Graphics, Sound, and AI into different threads - this is what causes performance decrease in some games.

No, im not talking about games here when i say performance. You said it increased performance for everything, using the 64-bit verions can slow certain applications i.e. NOT games, down.
 
Originally posted by reever2
No, im not talking about games here when i say performance. You said it increased performance for everything, using the 64-bit verions can slow certain applications i.e. NOT games, down.

I didnt say everything- I said all 64 bit optimized applications
 
I guess I should have phrased my original question better. Say I go to the store on the realease date - Will there be two different versions on the shelf (64-bit, 32 -bit)
 
I'm not sure but i do believe i read somewhere there would be a 64bit server but not a client.... but like i said i'm not really sure since i dont remember where i read it. Might have been about another game tho....

It might have been in the valve sticky thread... try doing a search.
 
Your right, the 64bit processors won't help image quality, but it will allow your comp to runt hat sma eimage quality faster and more efficently.
 
From the Valve info thread:

"I would expect we would run about 30% faster clock for clock comparing an Athlon running 32-bit code and an Athlon 64 running 64-bit code. Release of the 64-bit client will be gated on MS releasing 64-bit Windows."

Gabe says 30% increase in performance.
 
Allows me to educate you boobs

Okay, since i've got a few years college degree in computers, and 9 years experience building them, I suppose I should educate you all, for the betterment of mankind.

First, i'll explain what 64-bit processing actually means, since clearly none of you know.

When anything is reffered to as 'X bit', it means how many bits of information are sent per trip. A bit, as we know, is an 'on' or 'off', a binary number, 1 (on) or 0 (off).
32-bits means a device sends and receives 32-bits at once. Example of a 32-bit BUS:
L1- 0
L2- 1
L3- 1
L4- 0
L5- 1
...
L32- 0

This means a processor of 32-bits received 32-bits (4 Bytes) per clock cycle.
Clock Cycle means the amount of 'ticks' that occur per second. This is measured in Hertz.
If a processor's multiplied clock setting is 1 GHz, that means it performs 1,000,000,000 'ticks' per second:
32 bits per cycle, that makes 32,000,000,000 bits (4 GB/s)

So, if a 1 GHz processor is 64-bits, it moves 8 GB/s, twice the data at the same speed.

Now, consider that the fastest 64-bit processors are not even 2 GHz yet. When you do the math, 64-bit processors are /not yet/ faster than the higher end 32-bit processors.

Now, another thing that 64-bit processors have is a higher memory register address maximum. Any processor can only access Bytes of 2^n, where n=The bitrate.
So a 2-bit processor can only access 4 Bytes of data.
A single 32-bit processor can only access 4 GB of RAM max.

However, a single 64-bit processor can handle 14 EB (Exabytes, 1,000 TB [Terabytes, 1,000 GB])




Bitrate is also NOT the only thing which goes into a processor's speed.

A further understand is that the 'speed' of a processor refers to the Internal Multiplied Clock Frequency. This is the speed of data INSIDE the processor. Everything outside still operates at the /core/ processor speed. Also called 'System BUS' or 'Front Side Bus'.

So if you have a 2 GHz processor with 400 MHz FSB. That means that data /inside/ the processor, accessed from L1 and L2 Cache (explained later) is the only data on the PC moving that fast.
Data OUTSIDE the CPU, such as RAM, moves at the FSB maximum.
And devices such as HDD, move at the ATA BUS maximum (the Southbridge)
So when you overclock your PC by an extra 33 MHz? Your not getting 33 MHz out of the entire machine, only about 128 KB of it.


The /biggest/ indicators of speed are in Cache Size. Cache is memory built into the CPU's chip, which operates at the full speed of the CPU's multiplier.
L1 Cache is the first, and fastest. It moves at 100% of the CPU's speed. So if you have a 2 GHz CPU with 400 MHz FSB, it runs at 2 GHz.
L1 cache is, usually, either 64 KB or 128 KB, and divided into two sections:
Data Cache (for storing small peices of data)
and Instruction Cache (Which stores a long listed 'queue' of orders the CPU is sending to devices)
Durons use 64 KB, while most Celerons use 128 KB.

Next is L2 Cache. L2 Cache Operates at a lower speed than L1, but it's still nearly the speed of the multiplier. A 2 GHz CPU would have it's L2 run at roughly 1 GHz. However, some CPUs (such as AMDs) use the same type of memory for L2, and L2 moves at the full multiplier. THis is why most people beleive AMD to be faster. And indeed, it does make it mvoe much faster.
Most processors use L2 Cache. The only ones that don't are some Celerons and all Durons. L2 Cache varies from processor to processor, but is usually 256 KB or 512 KB (in newer ones).


Now, data outside the CPU is MUCH slower.
Most motherboards (all new ones) have on-board Cache, sometimes called L3 Cache. (Note that some processors for servers, such as the Xeon, has L3 Cache on the CPU. This is not to be confused.)
L3 Cache usually moves at the FSB maximum of the motherboard. If the motherboard (NOT the CPU) is 400 MHz FSB, the L3 moves at 400 MHz FSB, even if the CPU's FSB is lower.
Yes, a new mobo can make your PC faster.
L3 cache sizes range GREATLY between PCs. Most cheaper boards ($50 an down) use 256 KB to 1 MB of L3 Cache. More expensive use 2 to 32 MB. And the top of the line can use more. Mine (Asus A7N8X Deluxe) uses 64 MB of 400 MHz FSB L3 Cache.
The fastest L3 Cache right now is on the P4 800 MHz FSB boards. However, note that these boards use 64-bit memory, not 128-bit memory like other boards.
The fastest memory in terms of actual data movement, are 400 MHz AMD boards. The data movement specific, however, is not important when dealing with games: Only for server operations, databasing, and other high-yield high-demand programs, such as Photoshop.


Now, the next slowest is the RAM. RAM cannot operate higher than the motherboard's FSB rating. However, it CAN go over the CPU's FSB. But this is a dangerous setup, and can cause severe and frequent errors.
Nonetheless, the fastest memory right now in terms of data movement is only 500 MHz, 128-bit DDR.
For games, though, since games use very small, simple instructions, one might get better results from RDRAM, a proprietary and very expensive RAM from Rambus, a co-company of Intel. RDRAM moves, at max, of 1066 MHz. It is, however, 16-bit memory. So it moves very, very little data.


Beyond the Memory comes a lot of other issues, but we don't go into it.



The biggest thign 64-bit processors has to offer right now is slightly increased speeds in multiplier, but VERY VERY GREATLY INCREASED FSB. A newer Opteron has 1.6 GHz FSB, and 1.8 GHz total multiplier core speed.

Also, the biggest thing it has to offer, is a HELL of a lot more memory. 14,000,000 GB of RAM is a really amazing amount. It brings the Wierd Al Yankovich song 'All about the pentiums' to mind. I'll quote: "I got a million gigabytes of RAM!"

Indeed. One day, we'll have millions. Perhaps all 14 million gigs of RAM.







As for just making games faster, 64-bit or 32-bit will make no difference. The code will still be just as efficient in 32-bit. Half Life 2 will NOT be able to, nor need to address more than 4 GB of RAM. If it does, I seriously quesiton their sanity.
 
Originally posted by Zoltar
From the Valve info thread:

"I would expect we would run about 30% faster clock for clock comparing an Athlon running 32-bit code and an Athlon 64 running 64-bit code. Release of the 64-bit client will be gated on MS releasing 64-bit Windows."

Gabe says 30% increase in performance.

But is he talking about game performance or server performance?
 
Thanks Jangular, I was wondering what a 64-bit CPU could do over a 32-bit CPU. You should write a book or something...really, that was a great description.
 
Thanks, but i'm really not qualified to write books. ^.^

That was a very watered down explanation. There's millions of other things that go into it. Namely why PCs may never reach the full 100% memory address (ever seen a real 1 32-bit processor PC with 4 GB of ram? That's because some of the registers have to be used for devices, like AGP and PCI slots!)


But I hope I set any of the exaggerated numbers to rest.

Upgrading bitrate isn't making the water faster, it's making the hose bigger.
 
Re: Allows me to educate you boobs

Originally posted by Janglur
However, some CPUs (such as AMDs) use the same type of memory for L2, and L2 moves at the full multiplier. THis is why most people beleive AMD to be faster. And indeed, it does make it mvoe much faster.

Uhh what are you talking about? Faster than what? You really should include a section about cache configuration, it has a much larger effect on cache bandwidth than mhz and bus width.

And the P4's L2 bandwidth is more than Amd L2 and L1 bandwidth, more than 3 times faster than its L2

http://www.aceshardware.com/read.jsp?id=50000364
 
Wow - I'm printing that post out! (in the morning, I don't want to wake anybody up with my obnoxious printer)
Thanks for explaining, Janglur - yes you do explain things well; I feel very educated!
 
I don't know where you pulled from figures from, but that's SURE as hell not what's being taught in college.


As for benchmarking programs, i'm afraid they are NOT accurate.
ALL benchmarking programs, being programs, have certain areas which have patterns. Hardware will handle these patterns.


For example: If you specifically code hardware to process certain commands, such as 'addition' faster, then a program that does a lot of addition will move faster. That does NOT mean it's faster than another peice of hardware.

Benchmarks should never be trusted. All benchmarks, being software, are run through hardware, which has natural preferences.

This is why two benchmarks of the same type, but from different programs, can have VASTLY different results.

As well, tiny things can change benchmarks: A CPU on a motherboard with extremely fast FSB and L3 Cache is going to always score better than on a slower. Tiny adjustments like this can interfere with the benchmark results to the point of invalidating them.


The only thing we know for certain is the hardware's physical limitation. And in that, we rely. Anyone who puts their money into benchmarks WILL find themselves cheated.

For those who digress, please refer to Cyrix processors.
 
Re: Allows me to educate you boobs

Originally posted by Janglur
blah.....blah....blah

Great post Janglur, I wish my hardware course taught me these things instead of recipies for making half-adders and assembly for motorolla 68000 chip.

A couple of things I don't get though are, firstly, if you cite the Opteron to have a 1.8Ghz clock then with a 64 bit word size and all other things being equal it's equivalent to a 3.6Ghz 32-bit box. These are not very widely available AFAIK.

Also, we have had the same instruction set for the last ten years or so and the 64bit chips are supposed to have a new instruction set. This will probably result in a performance increase although I don't know the specifics to be able to say this authoritatively.

Finally, much faster FSB(1.4Ghz) should mean faster transfer between RAM and CPU(as long as RAM supports the speed) which is a major bottle neck so another win for Opteron here.

Of course, most of these will be useless until Longhorn or whatever it's called comes out.
 
A note:

The war between 'which is faster, AMD or Intel?' does have an answer:


It depends.


If you going for low-power low-cost, Celeron beats Duron in power and speed. However, Duron is much cheaper.

The Pentium 4 is optimised for games, especially those utilizing hyperthreading. In reality, hyperthreading increases instability and decreases total yield in applications which are data intensive. But it increases the speed of applications which are cycle dependant, such as games.

AMD Athlon XPs are otimised for the business universe, numbercrunching, and work. They work slower per clock, but run stabler, move more data, and are priced for business: Cheaper.



The answer is pretty simple overall. Intel for play, AMD for work.

The only exception is in mp3 encoding and decoding. AMD utilizes some specific code to increase the speed of mp3 playback, making it 70-150% faster than Intel chips. However, it's still not much of a different: An mp3 loads just as fast on either, due to the speed of system Memory, and most primarily, the abysmally slow HDD.
 
They dont teach you cache associativity and its larger effect on bandwidth than bus/size/mhz increases? You really should ask whoever is teaching you about it or you will be missing out on some info there
 
all i have to say is whoever doubts that 64 bit system wont increase performance is dumb.

if u dont believe me just look at these numbers


2^32 = 4294967296

2^64 = 18446744073709551616

without going in depth, that explains it.
also the max amount of ram for a 32 bit system is 4294967296 bytes(arround 4 gigs.)
the max amount of ram for a 64 bit system is 18446744073709551616 bytes (higher than the terabyte. u count it out)

go here if u reeaaally want to read lol
http://www.arstechnica.com/cpu/03q1/x86-64/x86-64-1.html

edit: someone already explained this in GREAT depth but anyways at least this made me feal a little bit smart.
 
Originally posted by Gorgon
no budy, games are not application they are 3d application.......... application without the work 3d means business application....:cool:

I bet you felt all smart when you posted that. Too bad it's not true. :dozey: :dozey:
 
Re: Re: Allows me to educate you boobs

Originally posted by Llevar
Great post Janglur, I wish my hardware course taught me these things instead of recipies for making half-adders and assembly for motorolla 68000 chip.

A couple of things I don't get though are, firstly, if you cite the Opteron to have a 1.8Ghz clock then with a 64 bit word size and all other things being equal it's equivalent to a 3.6Ghz 32-bit box. These are not very widely available AFAIK.

Also, we have had the same instruction set for the last ten years or so and the 64bit chips are supposed to have a new instruction set. This will probably result in a performance increase although I don't know the specifics to be able to say this authoritatively.

Finally, much faster FSB(1.4Ghz) should mean faster transfer between RAM and CPU(as long as RAM supports the speed) which is a major bottle neck so another win for Opteron here.

Of course, most of these will be useless until Longhorn or whatever it's called comes out.


Correct, on most parts.
I cited /an/ Opteron. There's several typed out there. 220, 420, 820, etc.

And, these ultra-fast Opterons are alost some $2000+ dollars, for the CPU alone.
Who here wants a few extra FPS /that/ badly?


As for the 64-bit data instruction sets, it's true this will become more efficient in time. But seeing as how Half-Life 2 was programmed intiially in 32-bit instructions, to utilize 64-bit instructions would mean entirely redesigning it. I doubt they will go through that trouble for a 5-10% speed increase, especially since they'll not come near using 100% of a high-end 32-bit processor /anyway/.

The 64-bit data isntruction won't become very useful until we've had it for another ten years, same as 32-bit.

Remember: We have 128, and even 256 bit console systems now. But they're mostly still using 32-bit instruction sets!
These extended instruction sets have little place in computers, yet. A few might make a novel save in typing here and there, but one set isn't going to make a game suddenly run twice as fast.

Bitrate evolution and programming evolution are seperate worlds. And the latter moves much, much slower.


Remember: NASA uses mostly 16-bit processors for it's most important controls. Because the instructions, while limited, are reliable and more easily predicted.
 
Another thing, too.

Programs coded in 32-bit, even if ported to 64-bit, isn't like changing XLS to HTML.
It's changing the very machine code it's made up in.


What this means is that programs, native 32-bit and 64-bit 'altered' 32-bit programs, are still only using primarily the 32-bit resources.

So far, the 64-bit data instruction set hasn't added any new processes that have been applied for use today, with one exception: Bigger memory addresses.

Basically, 64-bit and 32-bit is going to end up like Windows and DOS.

DOS loads in real-time mode, and loads the Windows OS Kernal, which switches to protected mode (32-bit mode). 16 bit programs are still only able to use 16-bit resources. (Anyone who plays old games on 98 has surely encountered conventional memory shortages)

64-bit OSs will effectively make 'bubbles' to emulate a 32-bit environment.
 
Back
Top