Slashdot Mirror
AMD Athlon 64 Performance Preview
k-hell writes "It seems like X-bit Labs have gotten their hands on an 'engineering sample of the AMD Athlon 64 2800+ processor'. Damage at Tech Report is writing that 'This is really fun, but I am a little concerned about their memory latency numbers.'"
I don't think that word means what they think it does.
One line blog. I hear that they're called Twitters now.
Hey thanks. The site's down and flaming in a record 3 posts.
temporary mirror here
You should include the full quote of Damage, because just quoting out of context can be misleading. Here's the full paragraph (emphasis is mine):
This is really fun, but I am a little concerned about their memory latency numbers. They haven't specified what units those numbers are in, but latency numbers come out of programs like cachemem in CPU cycles. Obviously, processors with higher clock speeds will see more clock cycles pass per second than processors with lower clock speeds. One must convert those numbers into comparable units, such as nanoseconds, in order to compare CPUs at different clock speeds. I do expect the Athlon 64 to have low memory access latencies because of its integrated memory controller, but I don't think the gap will be so great as the X-bit numbers would seem to indicate.
So, the worry is about the units the latency numbers are expressed in. And when you'd see the numbers below, you get an idea why it is so:
Athlon 64 2800+
Athlon XP 1.6GHz
Pentium 4 2.8C
See it for yourself.
--
Error 500: Internal sig error
A number of other sites have gone ahead and linked it
naw.. probably just slashdot
Selling software wont make you money, selling a service will.
Did they burn their fingers?
For those of you too lazy to click through to the 11th page, here's their conclusion:
The major conclusion, which we can draw as a result of this test session, sounds as follows. Even though Athlon 64 processors have internal architecture very similar to that of Athlon XP processors, they still differ from their predecessors quite significantly from the practical point of view. We can't give you a definite answer to the question, if Athlon 64 has become any faster than Athlon XP. In fact, this is just a different processor.
Moreover, there is also not much we could say about the performance of Athlon 64 in 64bit applications or at least in 64bit operation systems and 32bit applications. Supposedly, x86-64 will ensure a significant performance improvement, but it is also quite possible that x86-64 will not receive a warm welcome from the software developers. AMD has already tried to promote its own instructions set a while ago, and this experience could hardly be regarded as a success, to tell the truth. 3DNow! instructions set failed to become widely spread even though it proved to be very convenient to work with. So, we can only wait here for the first signs for or against these suppositions.
Speaking about the performance of Athlon 64 in traditional 32bit applications we can say that this new CPU boasts a few very remarkable and strong features: large L2 cache, high-performance memory subsystem and SSE2 instructions support. On the other scale we see relatively low core frequency. As a result, we see either a performance boost or a performance drop depending on each particular application and its critical parameters.
For example, Athlon 64 is not very successful in traditional calculating tasks, such as scientific calculations or 3D rendering. But as soon as we get to games or info compression, it appears beyond any competition. In general, if we compare the performance of Athlon 64 2800+ with that of Athlon XP 2800+, we will have to admit that the latter appeared slower than our today's hero in quite a bit of benchmarks.
Mooniacs for iOS and Android
On the picture of the chip, it's stamped "Copyright 2001".
One line blog. I hear that they're called Twitters now.
So, what could be the problem? The problem is that AMD is trying to support too many instruction sets. AMD should have axed 3DNow! and swallowed its pride. Supporting MMX, SSE, and SSE2 is sufficient.
When you try to put everything and the chicken sink into a chip, you inevitably pay for it with a slower clock speed.
Does the high cache size of Sparc processors outweigh their slower clockspeeds? Do the instruction sets?
I'll soon be getting a server that does involved calculations for for members of Math Addicts Anonymous. These will involve things like prime factorizations of insanely large numbers and calculating pi to more digits than anyone could ever care about. Assuming a budget of $500(the server will have other duties) for the motherboard/ram/processor, what would be the best architecture for the job?
You can't judge a book by the way it wears its hair.
The processor doesn't take to many benchmarks, but you can't fault it too much. It's nice to see some numbers are the CPU in 32 bit mode, but let's not forget that EVERYTHING here is 32 bit (OS, programs, etc). I'd LOVE to see a comparison between 32 bit programs running under and identicle OS versions that do and don't support 64 bits (Win XP vs Win XP for x86-64 for example). I'd suspect that the performance would go up with a 64 bit OS (especially on the games, where drivers and such play such a big part). Considering it's clockspeed, it holds up very well. The fact that it's almost never far behind a current athlon with an identicle performance rating (which is actually like 400 mhz faster) shows that it can definatly run things well. This isn't the horridly crippled performance that we've all heard about with the Itanic.
So what's my take on all of this? I think that this shows that the x86-64 can really become a success. I know some of you out there are thinking "Why would I buy one? I've got a 2.4 ghz Octium 7 and my PC is faster than that thing." That may be true, but many people aren't like you. My fastest computer is a PIII 933, so even at 1.6 ghz that Athlon64 can run circles around my best PC. If you are using a PC that's even a year or so old, you can probably benefit alot if you were to move up to an Athlon64 when it comes out.
My notes on some specific benchmarks:
Now my objections to the benchmarking
My final thoughts are this: it looks quite promising, and I can't wait to see more. More and more people with comeout with benchmarks as time goes on, and with the Opteron released now, we'll soon see benchmarks of it in SMP mode against other chips in both 32 and 64 bit OSes with 32 and 64 bit code. Either way, it looks like it's more successful than the Itanic.
Comment forecast: Bits of genius surrounded by a sea of mediocrity.
'This is really fun, but I am a little concerned about their memory latency numbers.'
Note, this is not because they're bad numbers, but rather because the units aren't specified, and may be clock cycles, which wouldn't be a fair comparison to the other processors.
couldnt they have used any one of the 64bit linuxes? this sounds like a bs review to me;
I want 2D games back.
X-Bit was down all morning... I had assumed it was being slashdotted... only to find out the link was just posted.
If non-Slashdotted traffic can bring down their site.. god help them now.
Printable version here.
Escher was the first MC and Giger invented the HR department.
Really? I could have sworn...
Sig.i>
lemme sum up the article:
- WAHOO! CHECK dis shit OUT! we got an athlon 64 chipz0rz!
- it's beefcake, dood. memory controller insIIIIDE!
- we're just gonna test it with 32 bit shizzle.
- it's like, good at some things, not so good at others.
anyway, here's something to consider: the sample they tested is 2800+ per AMD's performance rating spec, and it runs at 1.6gHz. yeah. most of the tests and graphs n stuff show it running around the level of a P4@2.53gHz. alright, so it doesn't exactly match the P4@2.8gHz. but think about this:
it's running at 1.6gHz!
nevermind the fact that it doesn't squash the fastest P4 they tested it against into the ground. it's just amusing to see how good the architecture is of the A64. i dunno. i think it's pretty cool, anyway.
anyway, seriously speaking: what use is testing a processor touted as being a 32-bit compatible 64-bit chip, when _NO_ 64 bit apps were used in testing?!
"uh. well. it ran the 32 bit stuff fine. and uh. it didn't fry."
Remember that this "preview" probably violates one or more NDA's, and it is of a desktop x86-64 chip that is scheduled for September release. In the meantime, it's bigger brother, the Opteron, who has more memory bandwidth, (usually) more cache, and multiple processor support will be released in less than a week (Tuesday to be exact).
Now the reviews that out in 4 days time should be much more interesting reads. I expect to see someone do a solid x86-32 vs. x86-64 comparison using Linux, maybe other OS's too. And yes, probably even Quake frame rate results. =)
http://saveie6.com/
Well the important thing here is how fast they can ramp up the CPU speed. If they can get it up to 3 3.5 GHz it will undoubtedly be a monster at those speeds, but as we all know, there are some operations where architecture can't replace raw GHz.
I also understand that the pipeline is a bit longer so I wonder if this decreases the IPC of the chip in comparison to the AXP or whether the other enhancements like the memory controller and HT link make up for that.
Dissolve... Resolve... Evolve...
To add more fuel to the "its only an engineering sample", check out the date on the
processor itself.
Imagine, with nearly two years of time to improve on this piece of silicon just what is in store for the Clawhammer. Personally, i'm waiting for it so I can finally upgrade my Athlon 600.
I always get the shakes before a drop.
Being that I am an AMD employee, I know for a fact that B0 (the version of clawhammer that they are benchmarking) is a very early rev, and should not even be considered when thinking about final benchmarks. Geez.
Well, one thing is for sure, xbit labs just blew their chances of ever getting their hands on another upcoming bleeding edge technology again.
true enough. i agree with you wholedheartedly. if it performs as well as a p4 running at 2.5-2.8gHz, then it should totally demolish the competition when its clock speed is bumped up.
but my concern is that there are a lot of factors that go into dictating the clock speed limitations of the chip. i mean-- the latest 32-bit athlons run pretty darn well compared to P4s, considering their clock speeds. but one gets the feeling that they're reaching the upper strata of how high they can be ramped up.
i think so, anyway. i would otherwise expect AMD to be matching or beating intel's latest and greatest with faster and faster iterations of the athlon. but it's not happening-- the performance crown remains with intel, while AMD gets the "sucker" price/performance crown. sorta.
anyway, i agree with you. i dig the fact that the chip does well on its own merits, and that it does so at comparatively lower clock speeds. maybe AMD will win more engineering/design accolades for their new processors.
Where is the 1.6 ghz P4 in this?
That's irrelevant. The proper way to square off chips is based on money. In other words a $200 dollar chip should go head to head with another $200 dollar chip, and an $800 chip goes against another $800 chip.
That's the only way to get fair results that are independant of implementation details. Clock rate the chip runs at is an implementation detail. It's not important. What's important is WORK per DOLLAR. That's the only thing that matters. Period.
think about this:
it's running at 1.6gHz!
And what is your point exactly ?
Take an Itanium2 at 1Ghz [*] and it will beat that 1.6 GHz processor to the ground. So are you going to sing the praise of the Itanium2 now ? No ? Ah sure it doesn't come from AMD but from Intel (and HP), so surely it must be a marketing trick.
You are a moron.
[*] soon to be replaced by a 1.5Ghz version
Or better, if you don't like paying more for less.Don't worry about english. Trying to get people to take you seriously has nothing to do with being able to create simple sentences. After all, why should a published article on a review site have better grammar than a post on slashdot?Not according to AMD's "performance rating." Does this site know something I don't?Ah. Silly me, thinking a transition through three or four completely new cores (tbird, tbred, palamino, barton) was significant.
Ah. And we all know how well Intel's lead in clockspeed translates into a lead in performance! It would be downright silly to compare a processor that doesn't break 2ghz, like, say, Intel's ACTUAL top-of-the-line offering, the Itanium, to a processing monster such as the P4!
I'll go back and read the full article, but so far I'm unconvinced...
>it's running at 1.6gHz!
...
Maybe they should have got their hands on one of those Liquid Nitrogen cooling units (or whatever) and just overclocked the hell out of it, see just how fast this bad boy will really go
I mean if you are going to dream, dream BIG.
Glonoinha the MebiByte Slayer
Hardware virtualization support would get the x86-64 a lot of traction in server environments, being able to run multiple virtual CPU's without needing complicated VMware-like software simulations.
Let's try running a plain old 32-bit x86 OS and apps on that Itanic. You can't? Guess what, you've already lost.
But here's another way to look at it - Itanium also has an x86 layer, but because it's really just an emulation, its performance sucks.
So I view this as a huge success. Why ? Because an Athlon-64 will be able to run "legacy" 32 apps at the same speed, while 64 apps will run faster.
You'd probably wonder why this is the case. Well, IMNSHO it's not because of the wider registers/ALUs, etc, but because of other improvements to the Instruction Set Architecture, like the 8 extra registers (16 total). Because you only have 8 registers on a regular x86, compilers can register-allocate very little. Adding 8 more registers means that you can keep more stuff in the register file, and you don't have to go to the stack (data cache) every single time.
The Raven
I guess they couldnt benchmark anything using 64 Bit Linux Distro's. I could swear Mandrake just released one and I'm sure with some begging a good site could get a beta of UT or something.
Man windoze must of made these d00dz blind.
i would be more interested in its price/performance because it better not be as expensive as an intel cpu.
I know you are psychotic, but please make an effort.
The chip will not be released to September. These numbers will be obsolete by then, because not only will the actual silicon have changed, but the common applications run will probably have changed quite a bit as well.
I consider these numbers completely meaningless. It's amazing that the chip can be competitive at all when it won't be released for another six months -- it just makes me more excited for its potential once it is finally released on the desktop.
the nice graph that was supposed to compare the chips developement? yeah, 2001 intel. very fair. also they said when the athlon the competition became much worse, gee i wonder who they were working for when they wrote the article. what the hell is intel advertising doing on slashdot? fucking corporate media
Another factor to consider is that the chipset is a critical factor: it took the Athlon almost 2 years before there was a solid chipset for it. Hopefully, AMD is already working on this.
No, I don't trust in god. He'll have to pay up front, like everybody else.
I would like to know if there is a good compiler planned for the x86-64 architecture. One of the reasons why I use Intel is because the Intel compiler is great. Now if they spent the time writting a compiler and they were giving it for free for non-commercial use just like Intel does (including support) I can very well see myself going AMD. Oh and it has to be a Fortran 90 and C/C++ with full support of OpenMP.
Wasn't that designed with 64-bit processors in mind?
Now this doesn't make much sense, because how can you run that in 64-bit mode even though you have a 64-bit processor, when the OS is running in 32-bit mode?
Or am I dreaming?
Huh?
Sure you can boot an x86 Os. Nobody sane will ever do so, but you can. The processor starts in x86 mode (or at least Merced did, Intel may have fixed that stupidity).
:-( ) you can run _native_ 32bits apps also. You don't _have_ to have 64bit apps.
As far as 32-bit x86 apps, they run, thank you. A bit on the slow side though.
With some of the OSes (not Linux
3DNow! is essentially a subset of SSE. Removing it would not save a signifigant number of transistors.
Until AMD brings out a 64 bit CPU that is SMP ready I think I'll just sit on the side lines for now. Besides, I've been running 64 bit all along with my Sun boxen. Now it's just time to sit back and let AMD and Intel battle it out to see who comes out with 64 bit SMP processor I can actually afford.
"I bow to no man" - Riddick
incase AMD's push for x86-64 falls through and they're stuck with supporting 32bit for a few more decades.
Non impediti ratione cogitationus.
lol, we (Zionist pigs) win...
you lose, sorry, you distugist, filthy stinkies....
If you think that's fun, try sex!
And what is your point exactly ?
.
Take an Itanium2 at 1Ghz [*] and it will beat that 1.6 GHz processor to the ground. So are you going to sing the praise of the Itanium2 now ? No ? Ah sure it doesn't come from AMD but from Intel (and HP), so surely it must be a marketing trick.
You are a moron.
yes, i will sing the holiest of holy praises for the i2 . it's nifty that it can perform so well at a lower clock speed . yes . and no
and finally,
You are presumptious. [*]
[*] your space bar seems to be malfunctioning , too . either that , or you ' re apoplectic .
Surely the real point of a 64 bit chip is its 64 bitness, not a pissing match with chips crippled with limited memory addressing.
It holds is own with 32bit chips, but the real benefit of this is when its used in an environment where more than 4GB of memory is an important parameter. Those other competitive chips won't stand a chance.
On my web site (under building PCs) I have some rather old data about ENTRY level machines and predictions of where these will be in the future. It looks as though we are getting 4 times the memory in about 3 years (possibly even less - this had come down to 3 years from 4 years in the 8 years prior) - meaning that by 2005 we will be in the 1GB range and by 2008 at 4GB. Remember that is entry level and that the top of the range machines use many multiples of this.
Classic hardware review article. Take every bit of old-news, re-hashed, "everyone's seen it before" information you can find to pad out your 2-page article to at least 10 pages. Insert lots of huge pictures to pad it out even further. After all, you get paid by the banner impression, right?
Oh, you're not stuck, you're just unable to let go of the onion rings.
>anyway, seriously speaking: what use is testing a processor touted as being a 32-bit compatible 64-bit chip, when _NO_ 64 bit apps were used in testing?!
Maybe because the majority of people who will be buying it will be running a 32bit Operating System, with 32 bit applications.
Unless you think Microsoft is gonna release a 64 bit XP Home anytime soon.
The Athlon64 is a successor to the Athlon XP, not an alternative, so it has to run 32 bit code as fast or faster than an XP, or no one will touch it.
Those that want to run it on a server will buy an Opteron instead.
Advanced users are users too!
Yeah. Gotta love those "chicken sinks".
The memory controller in the chip they tested handily beat out the dual-channel Nforce platform, and while it didn't beat the dual-channel DDR 400 memory of the P4, it wasn't too far behind.
Here's the cool part: That was an Athlon64. A desktop-oriented chip with a single-channel memory controller. The chips that are coming out on Tuesday are the "Hammer" multi-processing chips, targetted at servers, with *dual-channel* memory interfaces. Look at the memory numbers on the single-channel, and come close to doubling them. Now THAT will be impressive!
Here's hoping that the Hammers will really live up to the hype. Not only would they make good, solid servers, they'd definitely step up the competition with Intel, which means faster, cheaper chips for us.
steve
Oh, you're not stuck, you're just unable to let go of the onion rings.
In an interesting review they would have managed to work out how to use Linux and run some 64 bit apps instead of "Look mom, I can run Windoze, I'm so 1337, and look, I can download some 32 bit benchmarks and take screenshots too!"
libguestfs - tools for accessing and modifying virtual machine disk images
Euhm, to the poster: low latency is better than high latency. It is an indication for the time you have to wait before the first bit of data is available. The P4 has a faster memory throughput, but higher latency. This means you'll have to wait longer for the data to become available, but once it's there, you'll get it faster.
The other enhancements to the chip more than make up for it. In fact, it should typically perform anywhere from 5-25% faster than an equally clocked AthlonXP.
The big downside to a longer pipeline is that it increase the performance penalty of flushing the pipeline after a mispredicted branch. However, AMD did two things here. First the Athlon64 has a better branch predictor than the AthlonXP, which should reduce the number of mispredicted branches. The second thing that AMD did was to change the pipeline slightly so that the extra pipeline stages don't need to be flushed on a missed branch.
FWIW AMD is kind of following Intel's lead in this regard. While the P4 has a VERY long pipeline (twice the length of the Athlons), it doesn't take all that much of a performance hit for a missed branch because they don't need to flush all of their pipeline. The P4 also has excellent branch predictors (second only to the old AMD K6 in the x86 world from what I've seen). That being said, the P4 still has a very long pipeline and very small L1 caches, as well as a number of odd quirks as to what instructions can be where in the pipeline, so the end result tends to be a reduction in the average IPC of the chip vs. the AthlonXP/64.
As for clock speeds, my understanding is that AMD is aiming for the 2.0GHz range when they bring the chip to market.
They always price their chips heaps lower.
To the point that an AMD chip can sometimes be cheaper by 3 figures over a Intel chip of the same performance (that's performance, not bus speed).
Gez I remember when Duron 600s were heaps cheaper than the P6 Celeron 600 (Cu'mine), even though they trounced 'em performance wise. & still outperformed them even when one plays arround with the Celeron's bus speed & then unlocked the Duron & played arround with its bus speed & multipler, so that both chips were at the same speed & had the same memory speeds - Celeron 600@900 (FSB @ 100 rather than 66) vs Duron 600@900 (9x rather than 6x FSB muliplyer, with the memory bus multiplyer at 1 rather than 4/3, remember its only the socket A's CPU to northbridge bus that's DDR)
... that doesn't reside in a space heater? Sure, I would like a faster computer than my current Athlon 500, but the high energy consumption, which leads to excess heat which leads to noisy fans xor fried chips (which may be good in England but not here in the States), is getting out-of-hand crazy.
I feel sorry for all the gamers out there that must buy all these $1000 space heaters every year or two just to continue their gaming fix. Outside of gaming, are even 1% of these systems even close to fully utilized? I seriously doubt it.
I understand that my needs are different from the average desktop owner. I use my computer to play DVDs, encode/decode OGGs, download and watch movie trailers and other video, and compile bucket loads of free software packages via linux from scratch. I even try a little development myself. However, for these tasks, I have found that the optimum computer remains a CPU from about 1998 or 1999. And if you have a few hundred dollars, maxing the memory of an old Athlon or P3 and replacing the IDE drives with SCSI ones leads to a surprisingly fast system.
Social Contract? I don't remember signing any Social Contract!
just a theory :-)
Why mod things as insightful when they are cleary wrong? Bigger pointers means bigger executables, which in turn means more cache misses. AMD expect an increase of between 0 and 10% (usually around 5%) in program size / cache misses. It takes about 10 seconds to find this information on Google. That's 9 seconds more than the person who wrote this message though about it, and about 11 seconds longer than the person who modded it as "insightful"...
Actually, the great part is the extra registers. That, along with the integrated memory controller and the improved branch prediction should help make it significantly faster than the Athlon XP, even in 32-bit code (it requires recompilation but no rewriting). The bigger cache is also nice, but I suspect most Athlon 64 models will have 512 KB L2, like the Barton XP.
64-bit registers and bigger-than-32-bit addressing (40-bit for now, I think, with higher models using 48-bit) are there mainly to give AMD a foothold in the server market. For most of us, they will not have an immediate impact.
RMN
~~~
Another problem with supporting such a large instruction set is the debugging time. It is the primary reason that the Athlon 64 was delayed so long. In modern chip design, the fun part (i. e. the actual design) is about 20% of the development time. The other 80% is the debugging time. If the number of transistors increases by, say, 10%, then the deugging time may increase by 30%.
3DNow! had its use as a marketing tool to show that AMD can take leadership. Beyond that, 3DNow! is no longer justified since SSE2 and SSE can do whatever 3DNow! does.
I'm the Iraqi information minister, and I tell you this: you are a liar! You are all liars! There is no Athlon 64. Never! AMD is commiting suicide. Intel has been wipded out. Now we ambush Transmeta and we destroy it, with the help of God.
It's palomino, not "palamino", you illiterate cunt.
For the masses, 64-bit is a marketing gimmick. It's useless and will continue to be useless for several months (programs need rewriting, retesting, etc., assuming the authors bother to do it in the first place). This is for the 64-bit registers.
The 64-bit memory controller (which doesn't actually support 64-bit addressing, BTW) is even less relevant for normal users. Most people don't have any use for more than 512 MB. Even workstations can work perfectly with 2 GB or so (unless you're working with floating-point / HDR HD streams, or massive prints, anyway - I do sometimes, and in that situation memory is worth its weight in plutonium).
While "the real point of a 64 bit chip" is its 64-bitness, the real point of the Athlon 64 is not that it's a 64-bit chip.
Much more important to its success in the mass market (and the Athlon 64, unlike the Opteron, is aimed at the mass market) is its performance with generic, existing applications. And for those, the extra registers, the improved branch prediction, the integrated memory controller and (hopefully) the higher clock speed matter at least as much (in fact a lot more) than the fact that the registers are 64 bits wide.
The chip used in this "review" is very odd. From its low clock speed and big cache, it looks more like an Opteron than like an Athlon 64. The Athlon 64 was expected to launch at "3200+" (around 2 GHz) and have 256 or 512 KB of (L2) cache. With the Athlon 64 scheduled to come out in September, I wouldn't be very surprised if this "engineering sample" was in fact an Opteron 242 packaged as an Athlon 64 (socket 754 instead of 940), sent to motherboard / chipset makers so they could test their Athlon 64 boards (remember, the socket is different, but apart from SMP support and the respective HT links, the Athlon 64 and the Opteron are almost exactly the same).
With AMD's recent manufacturing problems (hopefully solved with the IBM deal) and with the Opterons launch approaching, I doubt they'd be wasting time and fab space making "real" Athlon 64s.
This is pure speculation, of course, but I think it makes some sense. I expect the final Athlon 64s to run at a higher clock speed and have less cache (meaning they'll perform much better than this "sample" in most tasks, and slightly worse in some - namely the compression test, which relies heavily on the big cache).
RMN
~~~
Look at recent versions of GCC. A lot of work has gone into optimizing it and adding support for X86-64. A lot of improvements have gone in to versions 3.2 and many more in the upcoming version 3.3.
This post is encrypted twice with ROT-13. Documenting or attempting to crack this encryption is illegal.
My apologies for the "moron" bit, I was indeed mistaken. Too much FUD from AMD fanboys...
I guess a good cure for what ails you is kitchen soup...
Enby in Waltham
I am an REAL AMD employee.
This chip must have snuck out of the recent ones we sent out for testing. This is a current rev.