• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

EM64T vs. AMD64

E7130

Limp Gawd
Joined
Nov 17, 2006
Messages
330
Ok, just a quick question. My understanding of EM64T (which intel copied from AMD64 because of licencing they had) is really no different and does the same exact thing. There are a few differences, but really all the same.

Another tech guy from a company that I was talking to about servers (he is an AMD fanboy) and told me that Intel's are not true 64-bit. They are 32-bit that emulates 64-bit. At the time (we where in a meeting) I just kind of nodded and went on my way.

However, my understanding of it is that EM64T doesn't require this "emulation" and that it runs 64-bit code just like the AMD does. Or works with a 64-bit server OS the same too?

Am I mistaken?
 
neither AMD64 or EM64T are true 64-bit, in the sense that they were designed from the start to support 32-bit directly. They are really for all practical purposes identical architectures.
 
neither AMD64 or EM64T are true 64-bit, in the sense that they were designed from the start to support 32-bit directly. They are really for all practical purposes identical architectures.

How in the world are they not true 64 bit? They have 64 bit instructions that operate on 64 bit data and they have 64 bit registers.

Using your line of thinking, the 32bit 80386-Pentium 4 (sans EM64T) aren't true 32-bit processor because they were designed to support legacy 16bit code and data of the 808x and the 80286.
 
How in the world are they not true 64 bit? They have 64 bit instructions that operate on 64 bit data and they have 64 bit registers.

Using your line of thinking, the 32bit 80386-Pentium 4 (sans EM64T) aren't true 32-bit processor because they were designed to support legacy 16bit code and data of the 808x and the 80286.

All of the processors since 286, x86 processors use 16bit WORDS, for 32 bit its a DWORD or a double word, in 64 bit its a QWORD, That why they are backwards computable with 32 and 16bit. The intel itanium for example uses 64 bit words which is why its a true 64 bit processor, and which is why normal OSes don't work on it.

this is absolutely zero difference between EM64T and AMD64, since its both the same instruction set designed by AMD.
 
All of the processors since 286, x86 processors use 16bit WORDS, for 32 bit its a DWORD or a double word, in 64 bit its a QWORD, That why they are backwards computable with 32 and 16bit. The intel itanium for example uses 64 bit words which is why its a true 64 bit processor, and which is why normal OSes don't work on it.

this is absolutely zero difference between EM64T and AMD64, since its both the same instruction set designed by AMD.

Don't forget the 8088 and 8086 were also 16bit. The reason normal OSes don't work on the Itanium is because the ISA is completely different. If you stripped out the 32bit instruction set from the AMD64 and EM64T it'd be a completely 64bit only processor would it not?
 
There is a considerable amount of confusion still with these types of technology.

64-bit... PSH!

Neither of the 2 technologies are 'true' 64 bit computing, as compared to IBM PowerPC or Intel Itanium, or a multitude of other CPUs out there (since they only execute 64-bit code). Each are CAPABLE of executing 64-bit code and addressing more than 4096 MB (2^32) of RAM. Each are a 32-bit core with 16-bit legacy, and 64-bit Extended Memory CAPABILITY. And truthfully neither can actually address 2^64 of memory either, which would be an insane amount though... it's around 2^40 and Intel was 2^36, but have improved.

To take advantage of the 64-bit CAPABILITY you'd need the CPU, a 64-bit enabled BIOS, 64-bit OS, 64-bit Drivers and Applications. Doesn't sound so hard now? But there is still confusion... It irritates me to no extent when people refer to their AMD Athlon 64's as 64-bit CPUs... GRR!

Now AMD does have an advantage whilst running 32-bit code in a 64-bit environment, eg using WoW32 in Vista x64, as it does not take as great of a performance decrease when executing the 32-bit code, compared to the Intel counterpart.

Though to answer your question, with the latest iteration of the technologies AMD64 and EM64T they are *almost* the same instruction set and *almost* the same capability for memory addressing, so you'd be hard pressed to see a difference on a computer running Vista x64 on either an Intel or AMD CPU... as is the case today with an equally matched CPU from either camp in regular Vista or XP.
 
There is a considerable amount of confusion still with these types of technology.

64-bit... PSH!

Neither of the 2 technologies are 'true' 64 bit computing, as compared to IBM PowerPC or Intel Itanium, or a multitude of other CPUs out there (since they only execute 64-bit code). Each are CAPABLE of executing 64-bit code and addressing more than 4096 MB (2^32) of RAM. Each are a 32-bit core with 16-bit legacy, and 64-bit Extended Memory CAPABILITY. And truthfully neither can actually address 2^64 of memory either, which would be an insane amount though... it's around 2^40 and Intel was 2^36, but have improved.

To take advantage of the 64-bit CAPABILITY you'd need the CPU, a 64-bit enabled BIOS, 64-bit OS, 64-bit Drivers and Applications. Doesn't sound so hard now? But there is still confusion... It irritates me to no extent when people refer to their AMD Athlon 64's as 64-bit CPUs... GRR!

Now AMD does have an advantage whilst running 32-bit code in a 64-bit environment, eg using WoW32 in Vista x64, as it does not take as great of a performance decrease when executing the 32-bit code, compared to the Intel counterpart.

Though to answer your question, with the latest iteration of the technologies AMD64 and EM64T they are *almost* the same instruction set and *almost* the same capability for memory addressing, so you'd be hard pressed to see a difference on a computer running Vista x64 on either an Intel or AMD CPU... as is the case today with an equally matched CPU from either camp in regular Vista or XP.

Would you not also need a 64-bit BIOS, 64-bit drivers, and 64-bit OS to take advantage of a "true" 64-bit CPU like the Itanium? The lack of a direct 32bit compatibility makes it a true 64-bit CPU and the inclusion of 32bit capability makes it a "fake" 64-bit CPU? That's a bunch of bullshit to me. If it can execute 64bit instructions that operate on 64bits of data at a time, and store the results in 64bit registers, it a 64bit CPU period.

(Car analogy warning) To say it's not is like saying that GM's Displacement on Demand enginges aren't really V8's but just four cylinders that can run with 8 when needed.
 
Would you not also need a 64-bit BIOS, 64-bit drivers, and 64-bit OS to take advantage of a "true" 64-bit CPU like the Itanium? The lack of a direct 32bit compatibility makes it a true 64-bit CPU and the inclusion of 32bit capability makes it a "fake" 64-bit CPU? That's a bunch of bullshit to me. If it can execute 64bit instructions that operate on 64bits of data at a time, and store the results in 64bit registers, it a 64bit CPU period.

(Car analogy warning) To say it's not is like saying that GM's Displacement on Demand enginges aren't really V8's but just four cylinders that can run with 8 when needed.

Indeed you make a good point... so after some research I shall concede. We could argue semantics for a long time if need be, but I still personally say those CPUs aren't truly 64-bit. Since they cannot address 2^64 bytes of memory... which more ram than I could ever: imagine, use, or afford. Now compared to the Itanium's which use the ISA IA-64, which can address 2^64 bytes of memory... Granted this is only 1 aspect of my point.

But I guess if it walks like a duck and sounds like a duck, it must be a duck. If it has 64-bit wide (virtual) registers and executes 64-bit code, then it's a 64-bit CPU.

Hmm... maybe it's just Intel saying they're not 'true' 64-bit' CPUs so people will spend an assload for their Itaniums.

Gosh... my head hurts.

Wait, did we answer the question yet?
 
Indeed you make a good point... so after some research I shall concede. We could argue semantics for a long time if need be, but I still personally say those CPUs aren't truly 64-bit. Since they cannot address 2^64 bytes of memory... which more ram than I could ever: imagine, use, or afford. Now compared to the Itanium's which use the ISA IA-64, which can address 2^64 bytes of memory... Granted this is only 1 aspect of my point.

But I guess if it walks like a duck and sounds like a duck, it must be a duck. If it has 64-bit wide (virtual) registers and executes 64-bit code, then it's a 64-bit CPU.

Hmm... maybe it's just Intel saying they're not 'true' 64-bit' CPUs so people will spend an assload for their Itaniums.

Gosh... my head hurts.

Wait, did we answer the question yet?

I don't really consider address space the determining factor of the bitness of a CPU. The 16-bit 808x address bus was a 20-bits, the 80286 was 24-bits, Pentium Pro, 2, 3 and early 4's had a 36-bit address buses using PAE.

And technically x64's memory space is 64-bits wide. However, any address from 00008000-0000000000 to FFFF7FFF-FFFFFFFF will raise an exception in the processor and thus limits the useable address width to 48bits.
 
Thanks to all of you on your input, it really has given me some more insight into this. That is probably why I was a bit confused when he (the gentleman I was referring to in my first post) said "Intels' are not true 64-bit processors" when I knew that Intel had copied the architecture from AMD (much like when AMD started by copying from Intel, isn't licensing fun.)
 
Indeed you make a good point... so after some research I shall concede. We could argue semantics for a long time if need be, but I still personally say those CPUs aren't truly 64-bit. Since they cannot address 2^64 bytes of memory... which more ram than I could ever: imagine, use, or afford. Now compared to the Itanium's which use the ISA IA-64, which can address 2^64 bytes of memory... Granted this is only 1 aspect of my point.
Actually, while the Itanium, like AMD64 or EMT64 has a logical address space of 2^64, the actual physical memory address bus is only 2^44 (http://www.tech-faq.com/address-bus.shtml).

Just like with the Opteron, this is merely a design limitation on the current hardware rather than a limit imposed by the architecture.
 
thats probably because in our current hardware and software, having
17,592,186,044,416 MB of ram is a bit over kill =p thats the limitation of a 64bit memory bus.

44bit mem bus =16,777,216 MB of ram a number that is a bit more realistic, but still FAR from what anyone will be using for a while =p

the bolded text is 4gb and 7gb, can you even count up to the highest number the 64bit mem address?

I thought the main reason that the IA-64 Could not run normal X86 software was due to the difference in the design IE. C2D / X64 are both CISC processors, the IA-64 is a EPIC design cpu
 
thats probably because in our current hardware and software, having
17,592,186,044,416 MB of ram is a bit over kill =p thats the limitation of a 64bit memory bus.

44bit mem bus =16,777,216 MB of ram a number that is a bit more realistic, but still FAR from what anyone will be using for a while =p

the bolded text is 4gb and 7gb, can you even count up to the highest number the 64bit mem address?

I thought the main reason that the IA-64 Could not run normal X86 software was due to the difference in the design IE. C2D / X64 are both CISC processors, the IA-64 is a EPIC design cpu

Back in 1986, did you even think personal computers would be this fast or use this much memory or have this much hard drive space?, watch in 20 years, 64 bit limits us again.
 
Back in 1986, did you even think personal computers would be this fast or use this much memory or have this much hard drive space?, watch in 20 years, 64 bit limits us again.
I don't know why but I think in 20 years, we won't really be where we think we will by now. I'm sure technology will evolve in all sorts of directions, and alot faster then it was 20 years ago till now, as the technolgy we have no allows us to further explore other things.
 
Now AMD does have an advantage whilst running 32-bit code in a 64-bit environment, eg using WoW32 in Vista x64, as it does not take as great of a performance decrease when executing the 32-bit code, compared to the Intel counterpart.

What do you mean?
I own an Core2 Duo, and my brother owns a Pentium D. We both have been using 64-bit OSes for quite a while now. XP x64 at first, and now Vista Home Premium x64.
We haven't noticed much of a difference going from 32-bit versions of Windows to their 64-bit counterparts. Some applications run slightly slower, others slightly faster. But the difference is never more than 2-3%, so negligible.
Ironically enough the biggest boost we've seen from running 64-bit software was with one of my 3d visualization routines, on the Pentium D. It has taken a lot of flack, but its 64-bit mode is actually quite amazing. Could be the SSE2-units running at the extreme clockspeeds, coupled with the extra registers in 64-bit mode... but it made a bigger improvement in 64-bit than the Athlons and Core2s we tested the code on.

If you ask me, people just *want* AMD to be better at 64-bit, because they were the first to market with the 64-bit extensions. But truth is that Intel's implementation, even in the Pentiums, does the job just as well, if not better.
 
Very interesting discussion guys. :)
I think another important question that arise many times is what advantage(s) does 64-bit environment brings us apart from 4GB+ of memory?

Sure, to be able to fully use 4GB and beyond of memory is great as 2GB of memory is bare minimum for gamers these days. What else does 64-bit desktop computing give us?
 
Very interesting discussion guys. :)
I think another important question that arise many times is what advantage(s) does 64-bit environment brings us apart from 4GB+ of memory?

Sure, to be able to fully use 4GB and beyond of memory is great as 2GB of memory is bare minimum for gamers these days. What else does 64-bit desktop computing give us?

It's supposed to be faster than 32bit, but in most things it's about the same speed. Might just be that programmers are still optimizing their own code for 32bit platforms and using the compiler to make 64bit code. Maybe once they start hand optimizing for 64bit we'll see some gains, but I don't know.
 
It's supposed to be faster? What part in 64-bit makes it faster than 32-bit?
 
It's supposed to be faster? What part in 64-bit makes it faster than 32-bit?

64bit mode can take advantage of all 16 registers in the CPU. 32bit mode is limited to only 8 of them. This means that processes can store more data closer to the execution units and won't have to bounce out to the cache or even main memory as much.

Then 64bit mode can crunch 64bit worth of data per cycle compare to 32bits in 32bit modes. Now these things don't necessarily mean faster, but in certain tasks the more data you can handle at once the faster things get done (e.g. video encoding).
 
I'll support the 'faster' claims. When I switched to the 64 bit version of Vista from 32, my boot and shutdown times decreased significantly.
 
doesn't AMD now have 128bit registers?(or 2 64bit ones) for SSE/SIMD/MMX etc =p?

The reason I think that we havn't seen the big improvements is because of software, not the hardware.
 
this is absolutely zero difference between EM64T and AMD64, since its both the same instruction set designed by AMD.

If I remember right from when I was doing a compiler design course a year ago, AMD64 supported some MOV instructions (direct memory to memory stuff) EMT64 didn't - and those instructions were handy too, but I may be remembering wrong.

Either way, the difference....is negligible.
 
Either way, the difference....is negligible.

I'd say the difference is non-existent in practice, because even though there are a few slight differences in the details of the implementation, every major 64-bit OS only uses the common subset of EM64T and AMD64 (Microsoft likes to call it x64, which I think is a good name for it. It is similar to x86, points out that it is 64-bit, it's not long and cryptic like x86-64, and it doesn't contain the name of a hardware vendor, so people won't be confused into thinking it might only work on one brand of hardware).
Therefore, in practice it doesn't matter which one you have. All 64-bit OSes and software will work fine on both. I don't think this will ever change, as compatibility has always been the primary feature of the x86 architecture. For well over two decades people have wanted to just upgrade their hardware and continue running their old applications.
 
Thanks, Ryan for the insight.

Anyway, guys, do we have any applications which actually perform better in 64-bit environment? What about games? I confess that the *only* 64-bit supported game I've tried is Source engine based ones (Half-Life2 / Counter Strike Source / Lost Coast). And I tried them last year on Windows XP 64bit Edition. Those games actually performed worse in 64-bit mode.
I don't know how 64-bit performance turned out recently with Source engine (especially with updated new Source engine 2007 powering HL2 Epsode2, Team Fortress2 and Portal). But given VALVe's slow update, I'm not holding my breath.
 
Thanks, Ryan for the insight.

Anyway, guys, do we have any applications which actually perform better in 64-bit environment? What about games? I confess that the *only* 64-bit supported game I've tried is Source engine based ones (Half-Life2 / Counter Strike Source / Lost Coast). And I tried them last year on Windows XP 64bit Edition. Those games actually performed worse in 64-bit mode.

I only know one other game in 64-bit, and that's Far Cry, which performs better in 64-bit (Crysis will also ship with 64-bit binaries, will be interesting to see how that one performs).
Most 64-bit applications also perform better, as they should. But as someone already pointed out... if people only optimize for 32-bit, and then compile to 64-bit (which I think is what Valve did), you don't always automatically end up with faster code. Some things are faster, some are slower. But in general things can be rewritten so they perform as good or better in 64-bit mode.
So I'm inclined to say that Source is the exception to the rule that games will run better in 64-bit... and Valve will probably get their 64-bit version right eventually. But since the Source engine isn't exactly state-of-the-art anymore, there's no point in putting effort into optimizing the 64-bit version. The 32-bit version already runs great, it doesn't need that many resources. When their next generation engine has poor 64-bit support, then they have a problem. We'll need 4+ GB soon for games, and 64-bit systems are going to deliver that.

I don't know how 64-bit performance turned out recently with Source engine (especially with updated new Source engine 2007 powering HL2 Epsode2, Team Fortress2 and Portal). But given VALVe's slow update, I'm not holding my breath.

They pulled the 64-bit engine from all their games except for Lost Coast, as far as I know. After a recent update my Half-Life 2 was no longer in 64-bit, and I've never had Episode One and Two in 64-bit at all. But considering the poor performance, you're not missing much.
The 64-bit versions did load levels quite a bit quicker than the 32-bit ones though... but the framerate was just poor. Back to the drawing board.
 
Scali, thanks for your reply. Your inputs/posts are always pleasure to read, and wealthy of information.

Yup, I'm very much looking forward to what Crysis brings us on table. This Friday can't come soon enough.

Cheers.
 
Well, I just got off the phone with the guy I was talking to and he told me that from a "the cost to own" view the AMD processor and motherboard chipset are better than Intel. And he still tells me that Intel's 64-bits are Extended Memory and AMD is true 64-bit with 64-bit registers where Intel only has 32, if I am hearing him right. He still clams that Intels emulate 64-bit through its architecture.

He also said that AMD guarantees that for each socket they make there will be processors for three years. Meaning that in three years time I could drop a 8-Core cpu in to replace a Quad core and it word work right out of the box. Where Intel it is not as upgradeable.

also AMD's cpu to fsb HT is better than Intel and I know intel is going to get this fix in 08.
 
Well, I just got off the phone with the guy I was talking to and he told me that from a "the cost to own" view the AMD processor and motherboard chipset are better than Intel. And he still tells me that Intel's 64-bits are Extended Memory and AMD is true 64-bit with 64-bit registers where Intel only has 32, if I am hearing him right. He still clams that Intels emulate 64-bit through its architecture.

He also said that AMD guarantees that for each socket they make there will be processors for three years. Meaning that in three years time I could drop a 8-Core cpu in to replace a Quad core and it word work right out of the box. Where Intel it is not as upgradeable.

also AMD's cpu to fsb HT is better than Intel and I know intel is going to get this fix in 08.
'

You friend knows not what he's talking about.
 
Well, I just got off the phone with the guy I was talking to and he told me that from a "the cost to own" view the AMD processor and motherboard chipset are better than Intel.

What exactly is "cost to own"? TCO? Those are such vague terms... They can be misused for anything.
As it stands, you can build Intel systems that deliver more performance per watt, better price/performance, or more performance in absolute terms, meaning you can be more productive.

And he still tells me that Intel's 64-bits are Extended Memory and AMD is true 64-bit with 64-bit registers where Intel only has 32, if I am hearing him right. He still clams that Intels emulate 64-bit through its architecture.

Then he is simply wrong.
Intel has never done this. I've worked with some of the first Pentiums to have EM64T, and they have always performed just fine in 64-bit mode. They always gain a bit from going to 64-bit mode, sometimes even more than an Athlon64.
If what he says were true, then 64-bit operations would take twice as long to perform than 32-bit operations. Trust me, you would notice that. You'd get a huge performance hit from running 64-bit software.

He also said that AMD guarantees that for each socket they make there will be processors for three years. Meaning that in three years time I could drop a 8-Core cpu in to replace a Quad core and it word work right out of the box.

Oh really? Tell that to the socket 939 users.
I'd like to see that in an official AMD document first.
Besides, the socket isn't everything, as early Barcelona users can tell you. You often need a BIOS update aswell, and for many boards these aren't available, so it doesn't work out of the box, and in many cases, not at all.

Where Intel it is not as upgradeable.

That all depends... Recently AMD hasn't been doing well with 939 and Am2.
Intel on the other hand has kept to socket 775, so people with newer boards can upgrade to a Core2 Duo or Quad. Some boards even support the new 1333 FSB with a BIOS upgrade.

also AMD's cpu to fsb HT is better than Intel and I know intel is going to get this fix in 08.

Who cares though? Intel's CPUs perform much better. Why? Intel simply has a lot more features that are better than AMD's (ALUs, SIMD, cache, etc etc). So the whole HT thing really is irrelevant in the bigger scope of things.
 
What exactly is "cost to own"? TCO? Those are such vague terms... They can be misused for anything.
As it stands, you can build Intel systems that deliver more performance per watt, better price/performance, or more performance in absolute terms, meaning you can be more productive.



Then he is simply wrong.
Intel has never done this. I've worked with some of the first Pentiums to have EM64T, and they have always performed just fine in 64-bit mode. They always gain a bit from going to 64-bit mode, sometimes even more than an Athlon64.
If what he says were true, then 64-bit operations would take twice as long to perform than 32-bit operations. Trust me, you would notice that. You'd get a huge performance hit from running 64-bit software.



Oh really? Tell that to the socket 939 users.
I'd like to see that in an official AMD document first.
Besides, the socket isn't everything, as early Barcelona users can tell you. You often need a BIOS update aswell, and for many boards these aren't available, so it doesn't work out of the box, and in many cases, not at all.



That all depends... Recently AMD hasn't been doing well with 939 and Am2.
Intel on the other hand has kept to socket 775, so people with newer boards can upgrade to a Core2 Duo or Quad. Some boards even support the new 1333 FSB with a BIOS upgrade.



Who cares though? Intel's CPUs perform much better. Why? Intel simply has a lot more features that are better than AMD's (ALUs, SIMD, cache, etc etc). So the whole HT thing really is irrelevant in the bigger scope of things.

thanks for you input. I know a lot of this stuff, however, I feel I must get feedback when presented with an opposite view. This guy is an "AMD fanatic" so I was taking what he said with a grain of salt.
 
'

You friend knows not what he's talking about.

Well, he is not my friend. He works for a IT company that we where going to have build a server for us (I would, however I am very busy most of the time at work and in my spare time I build high end gaming PCs).
 
scali, I believe he is referring to Servers when he talks about sockets lasting 3 years, even though intel has been on 775, the original 775s needed to be upgraded to support C2D, the second set needed upgrades for C2Q and 1333fsb support, 939 did get screwed over, but AM2 plowed the way after that. Maybe AMD switched to AM2 for the DDR2 price drops? who knows
 
Well, they priced a server with 2 Barcelona chips, 8GB RAM, 450GB Data storage RAID 10, 2 RAID 1 arrays for the boot and page file. WinServer 2k3 with upgrade to 2k8 for $9,000.

I talked to Systemax and they said they could build a server with same hardware (except with intel chips) for $6,243, w/o OS, but that isn't a big deal.

So which to chose. This server is running SQL.
 
Well, they priced a server with 2 Barcelona chips, 8GB RAM, 450GB Data storage RAID 10, 2 RAID 1 arrays for the boot and page file. WinServer 2k3 with upgrade to 2k8 for $9,000.

I talked to Systemax and they said they could build a server with same hardware (except with intel chips) for $6,243, w/o OS, but that isn't a big deal.

So which to chose. This server is running SQL.

I don't see how thats possible, intels chips are more $ and the ram to use with the intel chips is way more too. :|

the lowest barc out now is the 1.7 and its $209 vs the 1.6ghz xeon @ 290?
 
I don't see how thats possible, intels chips are more $ and the ram to use with the intel chips is way more too. :|

the lowest barc out now is the 1.7 and its $209 vs the 1.6ghz xeon @ 290?

The AMD system includes Win2k3 Server with an upgrade to Win2k8 while the Intel system doesn't include an OS at all
 
The AMD Server that we where given the $9000 price tag had 2 Opteron 2350 while the Intel machine would have 2 Xeon 5345 and both will have 8GB of RAM and 3 arrays for the hard drives (raptors). My supervisor is leaning towards the Intel setup.
 
The AMD Server that we where given the $9000 price tag had 2 Opteron 2350 while the Intel machine would have 2 Xeon 5345 and both will have 8GB of RAM and 3 arrays for the hard drives (raptors). My supervisor is leaning towards the Intel setup.

the opteron 2350 is 450
the xeon 5345 is 560 =p
even with the win2k3 server, does it cost more then 3k for a license? i thought it was around $800
 
Back
Top