In the most recent "Maximum PC" there was an editorial about AMD's new 64 bit chip and Intel's delayed entry into the consumer 64 bit game. It also pointed out that there are both 80 bit and 128 bit extensions in the current Intel processors for both floating point operations and multimedia. Taking this into account how much of a performance increase would a true 64bit chip present. I have not heard anything amazing coming from the jump to 64 bit, and this is mostly blamed on a lack of software to take advantage of the new chips. When 64bit software begins to become more prolific will we see any major changes? I am not very well versed on the topic so I would just like to hear some different opinions or comments. 😕
-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
The 64 bit question?
- Thread starter ThomS
- Start date
sandorski
No Lifer
There a few 64bit Linux performance tests out there(can't remember where exactly, checkout Ace's Hardware for some possible tests) . In those tests that I have seen, the Athlon 64 was able to beat the P4 in Multimedia Encoding and other areas where the P4 has always outperformed its' AMD counter parts, so it appears(so far) that 64bit OSs and Apps could make a significant difference.
DaveSimmons
Elite Member
Here's one long thread on 64-bit intel says you don't need it
You can find a few more threads if you search on "64" or "bit," some have already shuffled off to archived-posts land.
You can find a few more threads if you search on "64" or "bit," some have already shuffled off to archived-posts land.
sciencewhiz
Diamond Member
I doubt the increases in speed have come from moving to 64 bits. However, the Integrated memory controller and the extra registers that are accessible in 64 bit mode do increase speed.
sandorski
No Lifer
True enough, it is not the 64 bitness, just the way that AMD has implemented 64bitness into their cpus. In the end the difference exists between the Practical(64 bit implementation) and the Theoretical(64bit advantages), these differences seem to indicate though, that going with a 64bit OS and Software will benefit the Athlon 64 significantly.
InlineFive
Diamond Member
In the article they are talking about the fact that ALL processors convert 64-Bit x87 instructions to 80-Bits internally for data precision. It's not competitive advantage for anyone.
Not much. The AMD64 architecture is designed so that you can run 100% of your current software in it, and also run any 64-bit software you want to in the future. There are already 64-bit versions of Oracle, IBM DB2, Mandrake, SuSE and RedHat Linux, and everything that is open-source (99% of Linux distributions), since you only recompile it with the appropriate flags. There are 64-bit betas of WindowsXP already out there, but I think you have to be a subscriber to MSDN to get a (legal) copy.
Intel and the press in general is trying to downplay the importance of 64-bits, since they say we still don't need it. This would be like saying 15 years ago that nobody needed a 386! What would you think if they told you this? Intel's basic reason behind doing this is that their flagship 64-bit chip, the Itanium, which is incompatible with today's software, is selling abysmally, and they do not want to introduce a compatible 64-bit chip that would instantly obsolete Itanium. Ironic that, 15 years ago, Intel was the one enticing everybody to move to 32-bits, but today Intel is telling everybody that you don't need 64-bits!
Marcos
The 64 bitness is for the integer arithmethic only since standart FP operations can do even 80bit precision. There are very few applications that can benefit strongly from 64bit int - mainly encryption/decription algorithms (can show more than 2-3 times speedup since single 64 multiplication can be done in minimum 4 32bit muls + several adds).
But the big deal about 64bitness is ability to adress memory flat - current maximum memory adressable by 32 bit Windows is 4GB of which only 2GB are availble for process (Advanced Server has option that allows 3GB per process but it cost $$$). So if you are using more than 2GB in some application like SQL Server, CAD, 3D Modeling, or even large image processing having more than 2GB RAM on your machine is of no help - windows will start paging on the HDD to give your application a desired window of 2GB.
Of course most people are not running SQL server or CAD system so this is not so beneficial for them but luckily Athlon 64 bring more advantages to the table: it doubles the registers to 16 (~5% general speed increase in non stream applications but only in 64bit mode) and has onboard memory controller that reduces memory latencies signifficantly (can bring additional 10-15% speed in both 32bit/64bit)
So far all these advantages of 64bitness are not so big for the normal users (unlike business users where 2GB/4GB limitation is real problem) but most people forgets the biggest advantage x86-64 brings: in order for the application to run in x86-64 it has to be recompiled for Athlon 64 since this is the only chip supporting x86-64 now (Intel does not have such a chip). The good think about that is that application is automatically optimized for AMD CPU - so far most of the applications have been optimized for PIII and in very few instances for P4 (where Intel poured money - mostly benchmarks). This recompilation for a target CPU can bring easily 30-40% speed increase in many applications - for example on the A64 launch even in San Francisco AMD demonstrated 64bit DivX 5 running in some instances 30-40% faster than 32bit version, similar speed increase is with 64 bit version of Unreal.
And BTW there is current running beta version of 64bit Win XP - I even have one. It is supposed to go retail somewhere in Q2 2004 (as it is now it runs pretty stable but all the drivers has to be rewritten for 64bit so it takes time to port all 32bit XP drivers to 64bit).
Alright, someone seriously needs to dispel this RAM myth. The RAM addresability of the current 32-bit processors is not a problem yet and it won't be for the next few years.
I beg you to show me one home user that actually NEEDS (not wants for showing off purposes, but NEEDS) more than 4GB of RAM. As far as things stand today 1GB is just BARELY BARELY starting to be needed by some home users. The vast majority are perfectly fine with 512MB. There's the occasional hardcore media encoder/rendered that has 2GB of RAM, but that's about it. As things stand today, the 4GB ram addresability issue on P4s is a non-issue since 99.9999999% of home users aren't even close to needing one half of that limit, let alone more than the limit.
On the business end, Xeon processors have been able to address more than 4GB of ram for some time now. I believe their limit is 32GB or possibly even more, so the 4GB RAM addressability limit, isn't a limit on the business end. People need to stop using this as a justification for the need for 64-bit processors today.
And like the starter of the thread pointed out, FP calculations on today's 32-bit processors are done using 64-bits or more, so there will be no gains in that area. Also, can anyone tell me an application where 64-bit Integer calculations are NEEDED today, or might be needed in the next 5+ years on the DESKTOP (not business, DESKTOP) PC? And I'm talking about an application that is actually NEEDED by more than just a tiny fraction of a single percent of desktop users.
When Intel said that 64-bit processors aren't needed on desktop computers yet, and won't be for the next few years, they were NOT lying. Whether or not they're saying that as part of a marketing hype doesn't matter, the statement is TRUE. It's just that too many people on this and other enthusiast forums are always eager to bash Intel/Microsoft and always eager to take anything AMD/Linus Torvalds tells them as gospel to be able to look at a subject from an unbiased point of view and see who is actually telling the truth in the situation.
I beg you to show me one home user that actually NEEDS (not wants for showing off purposes, but NEEDS) more than 4GB of RAM. As far as things stand today 1GB is just BARELY BARELY starting to be needed by some home users. The vast majority are perfectly fine with 512MB. There's the occasional hardcore media encoder/rendered that has 2GB of RAM, but that's about it. As things stand today, the 4GB ram addresability issue on P4s is a non-issue since 99.9999999% of home users aren't even close to needing one half of that limit, let alone more than the limit.
On the business end, Xeon processors have been able to address more than 4GB of ram for some time now. I believe their limit is 32GB or possibly even more, so the 4GB RAM addressability limit, isn't a limit on the business end. People need to stop using this as a justification for the need for 64-bit processors today.
And like the starter of the thread pointed out, FP calculations on today's 32-bit processors are done using 64-bits or more, so there will be no gains in that area. Also, can anyone tell me an application where 64-bit Integer calculations are NEEDED today, or might be needed in the next 5+ years on the DESKTOP (not business, DESKTOP) PC? And I'm talking about an application that is actually NEEDED by more than just a tiny fraction of a single percent of desktop users.
When Intel said that 64-bit processors aren't needed on desktop computers yet, and won't be for the next few years, they were NOT lying. Whether or not they're saying that as part of a marketing hype doesn't matter, the statement is TRUE. It's just that too many people on this and other enthusiast forums are always eager to bash Intel/Microsoft and always eager to take anything AMD/Linus Torvalds tells them as gospel to be able to look at a subject from an unbiased point of view and see who is actually telling the truth in the situation.
As I have stated home users barly needs the 64bit functionality but they can still use some of the K8 features like 2x registers and on-board memory controller (and recompilation for specific processor is always wellcome).
Where you are completely wrong (or spreading FUD) is you part about Xeon and business users. Yes it is true that Xeons support more than 4GB but you somehow forgets to metion how this memory is used - in order to be used you need both PAE (Physical Adress Extensions) and AWE (Adress Window Extensions) and let me tell you as a database developer that paging SUCK$. You also forgets to mention that even with AWE you can not have any single process to acess more that 4GB of memory at the same time. So your (Xeons) PAE/AWE will help you zilch if you have very large model (either CAD/3D or whatever) and need to have frequent access to various parts of your model (for example rendering or physics calculation where you have to jump from object to object in that model). You can actually ask Unreal developers how they feel about so called "larger than 4GB memory support" of Xeon (no it is not something nice!). Actually PAE/AWE is even worse than the times when you can access memory above 1MB on the AT (yes the old 286 & 386 if you still remember it) via memory extenders (QEMM anyone 😉 ?) - don't take my word for it since I haven't develop anything for it - take it from my friends that works in MS in the SQL server team.
There is a reason why almost the only program that use for real AWE is SQL server - no other company can spend so much resources optimizing their product for little gain.
And finally you are making a bold claim as how in next five years home users will not need 4GB (actually it is 2GB per process to be correct) but let me ask you - how do you know how much memory next wave of applications are going to use? And how cheap the memory is now, and how cheap it will be tomorrow? Now most of the machines are sold with 512MB RAM while power (still home, at work having 2GB per small 1CPU SQL server just for development purposes is a must) users like me needs atleast 1GB to work comfortably.
He he - there are even games right now that does not work very well with just 512MB and you are telling me that it will be 5 years before home users will start using more than 2GB??? I bet that if next year 1GB RAM is <$100 most of the new sold machines will be with 1GB of ram. Also don't worry as long as such amount of memory is cheap developers will increase complexity of their models to match the common harware - you will see larger worlds/more detailed features /better simulation.
JustAnAverageGuy
Diamond Member
It was also once said that nobody would need more than 640k of memory. Better early than last minute.
I rest my case 🙂
Sir, at the speed computers advance (which is also increasing) 4GB of ram will be what 32MB is today 3 - 4 years from now. Games will require HUGE amounts of ram, and so will many programs. Besides video cards, ram is the fastest moving market for computer hardware. I mean seriously, an athlon xp 1600+ with 1GB of ram and radeon 9600 pro and some cheap motherboard can do ANY game out there well.
Matthias99
Diamond Member
Being able to do 64-bit floating-point calculations in a larger register set could provide a noticeable performance boost for many applications. Double-precision floats are *slow* in a 32-bit architecture, since they have to be packed/unpacked from 32-bit words. Scientific computing applications (including distributed programs like SETI@Home and Folding@Home, which you may classify as useless but many do not) would make excellent use of 64-bit processing, both integer and floating-point. The neuroscience lab I worked in last summer was very interested in getting more than 2-4GB of RAM for their simulations, but it just wasn't affordable or feasable for them to go to an Itanium server solution -- an Opteron running a 64-bit Linux would have been just the thing to fix their problem. Image and video-editing applications could make good use of 64-bit registers (and would also be likely to benefit from greater amounts of RAM). I imagine that 3D rendering/raytracing would also be a prime candidate for this sort of thing.
Are any of these what you would consider "mainstream" apps? Probably not. Are office apps going to need 64-bit computing any time soon? No, but one could argue that they didn't need 32-bit computing, either. More complex hardware drives more complex software, and vice versa. There's also good money to be made in selling people 3Ghz x86-64 CPUs with 12GB of RAM so they can surf the web and check their email. 🙂
There is no packing and unpacking in FP calculations.
Maybe, but the current 64-bit client of SETI is slower than the 32-bit.
http://www.amdzone.com/articleview.cfm?articleid=1315&page=2
Matthias99
Diamond Member
There is when they've moved back and forth from memory or cache, or if they have to be stored in 32-bit registers. The FPU and its handful of registers may be 64-bit, but the rest of the system is not.
This is a problem with the 64-bit SETI client, not 64-bit computing! Read the article... it sounds like they didn't really do a full "port" of the application, just a recompilation for the new processor.
The memory bus on modern platforms are 64-bit or 128-bit, the P4 has a 256-bit bus to the L2 cache, and has cache-lines that are 128-bit in size. x87 utilizes 80-bit registers and the P4 has a huge number of 128-bit rename registers for FP/SSE2. The main determination between 32-bit/64-bit CPUs is that a 64-bit CPU has integer registers that are 64-bit in size, and has nothing to do with FP at all.
Or that SETI, like many applications, has no need for 64-bit computing.
Matthias99
Diamond Member
The memory bus on modern platforms are 64-bit or 128-bit, the P4 has a 256-bit bus to the L2 cache, and has cache-lines that are 128-bit in size. x87 utilizes 80-bit registers and the P4 has a huge number of 128-bit rename registers for FP/SSE2. The main determination between 32-bit/64-bit CPUs is that a 64-bit CPU has integer registers that are 64-bit in size, and has nothing to do with FP at all. [/quote]
Okay, okay, so I guess caching floats doesn't hurt performance as much as I thought it did. 🙂
I don't really want to get into a pissing contest over this. As best as I can find, the P4 has 8 32-bit GPRs, 8 80-bit x87 registers, 8 64-bit MMX registers, and 8 128-bit SSE1/2 registers. I'm not sure whether or not 8 is in the "huge" range 😛. I also thought that the SSE registers couldn't be used directly for non-SSE floating-point operations (something like you couldn't have it be both source and destination for an FP op, but maybe that's only while you're using MMX or SSE1/2). If I'm wrong here, or you have better info, please point me towards a technical document or something, as I'm somehow not having much luck finding anything useful with Google right now.
Opteron extended registers:
http://www.tomshardware.com/cpu/20030422/opteron-06.html#a_big_deal_opterons_64bit_registers
Will it translate to better FPU performance? Double the GPRs, with each able to hold a 64-bit float if necessary -- along with twice as many SSE1/2 registers. I would think that performance with double precision floats in 64-bit mode would be significantly better, although by how much I couldn't tell you without benchmarks and/or more info on the rest of the floating point architecture.
Considering that SETI is pretty much nothing but FFTs and other forms of analog signal processing, I would think it would have a *lot* to gain from a 64-bit architecture if it significantly improves FPU performance. But unless you have the source code handy, it's really just speculation at this point.
Will 64-bit desktop computing pan out? I have no idea -- and neither does anyone else, really. If it significantly improves performance in real-world applications and the specialized fields that can use it (and we'll have to wait for 64-bit Windows next year to tell) at a minimal cost, then it will. If it does nothing but cost more, then it probably won't. 😛 We'll have to see what happens when Intel gets Prescott rolling -- a (hopefully) mature, super-fast 32-bit processor versus an unconventional, untested 64-bit newcomer. It ought to be fun to watch. 🙂
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
