Slashdot Mirror
VIA K8T800 Chipset Preview - Dual Opteron in Action
Mr.Tweak writes "It has been a long time coming but we are finally reaching the beginning stages of 64-bit mainstream computing. AMD has been the first to bring a 64-bit processor to the market with any true support in the Opteron. VIA is one of the key chipset companies supporting AMD64 and today TweakTown takes a preview look at their new K8T800 chipset with AMD Opteron 242 and 244 processors. 64-bit computing is boarding - don't miss the train!"
64-bit computing is boarding - don't miss the train!
I'll get on the train when the fare is a bit less.
I have over 70 freaks, do you?
As much as I'd love to board the train, the ticket price is a bit expensive, especially because I don't know exactly where the train goes.
I'm sure this is true for most other computer users as well. 64-bit on the x86 desktop is not really a 'mature' technology, even though it's been done before with Alpha and Itanium... I for one would hate to spend lots of money moving to an Opteron-based platform, to find out that one of my devices doesn't work or that one of my programs doesn't work. So I suspect that the risk involved (even though it's rather minimal, really) probably is going to keep a lot of people from moving to Opteron and co. for a while.
using namespace slashdot;
troll::post();
I can run this program without segfaulting?
main(){malloc(sizeof(cowboyneal)); }
I think the distinction between PC and RISC is quickly becoming obsolete, and so perhaps one should be including the UltraSparc IIe that's in the SunBlade 100/150, which retail for less than most new PC's.
By all accounts this is not the best of the current 64-bit chips, but I think it was the first to be offered in "PC-priced" systems.
Where have you been? I've had a 64-bit machine for almost 5 years now. ;) It's even been EOLed since July 2002.
My journal has hot
Just once, it would be nice to have a link to a motherboard review that wasn't split over 15 different sections.
'Without dragging on anymore, let's see what VIA's K8T800 chipset is capable of' - well, without dragging on any more than going to the next page after about two paragraphs...
-- Ed Avis ed@membled.com
The via board hinders the opteron. Instead of utilizing both on-chip memory controllers, it only uses one. You can see better results if you added a second set of memory banks for the second opteron to use.
There are more dual boards also including Tyans Thuder K8W. Rumor has it that nvidia will be coming out with a nforce chipset that will support dual opterons also.
Benchmarks
246 Benchmark
Overclocked 246
I'll stick to steam. Now only does it make a good chugga chugga noise like trains should but it also helps warm up the earth..
While it's slick that they had a dual proc board and all... none of the tests they used used the dual proc-ness of the system. They even indicate in their results that the second proc just threw overhead into the system.
;)
They've asked for help getting some dual proc benchmarking software. It would be great if someone could help them out. I'm really curious what that box is ACTUALLY capable of. IT's a goodly amount of horsepower with a reasonable amount of L2 cache with 64-bits of data-y goodness. It could make a heck of a "workgroup" size database server.... for a lot less than Sun's workgroup servers.
I think this next gen of procs (and their 64-bitness) is going to put another dent in Sun. First, lowend *NIX servers based on x86 put a huge dent in their pizza box market. Now, consumer grade 64-bit procs will probably start to eat heavily into their midrange market (like the 220R and that realm). In the big iron... well, that's contentious already. No need to to mess with them there
I'm down with that, as it were
...and now it works again, sometimes.
Because Dual Opteron along with a Pentium Gigatium makes the web fly! Even on 56K! Sorry, my fillings were picking up Intel Chip adverts from the space-year 2005.
There is a nice review in the latest german C'T too (issue 18 - page 36) of the MSI K8T Master2FAR board with dual opteron support, Via K8T800/VT8237 chipset and a lot of nice features (AGP, SATA, GBit Ethernet). Size constaints meant only a single memory-bus is implemented, a 5-10% speed bump on memory access. Another downside is no PCI-X slots.
The nicest thing though is the price: 280 euros. Sounds like a good workstation board.
Does anyone else think this site is a little weird? "PGA or Pin Grid Array can be separated into two parts: the first being Pin, which in this case is the small metal pins we see coming from the bottom of the processor and the other is the Grid Array." Well bless my stars! Who knew?!
True support... such as developing their own motherboard chipsets, like Intel, instead of farming the task out to VIA??
Don't get me wrong, AMD has done rather well relying on 3rd parties to develop mobo chipsets for them, but I'm not sure I'd call NOT having at least 1 in-house chipset solution "true support". If VIA and NVIDIA decided to dump AMD for some reason, they'd suddenly be in a pretty lousy position.
You see? You see? Your stupid minds! Stupid! Stupid!
... the majority of people who want to pay less for the same computers. Whenever a new super-duper computer comes up, a minority fringe with too much money (that I was once part of) blows ungodly amounts of money to get it, driving down the cost of perfectly serviceable, good, fast, but older computers for the rest of us.
So I say go AMD64 and go KIA, so I can buy those Athlons 2000+ based-machines I need even cheaper.
"A door is what a dog is perpetually on the wrong side of" - Ogden Nash
[Pruneau
#if _FP_W_TYPE_SIZE < 32 #error "Here's a nickle kid. Go buy yourself a real computer." #endif -- linux/arch/sparc64/double.h
So, the Slashdot Oracle is endorsing Sparc over Opteron? To hear is to obey, Master! ;)
UNIX? They're not even circumcised! Savages!
Sure, the benchmarks are impressive, but it would be nice to see more benchmarks using 64-bit-optimized software (eg. an optimized gromacs). 32 bit software just doesn't use all the power these chips have.
when you tried RENUMBER 10,10000
[second number is the line number interval]
There are places where the networks are not touching,and there are places where they are-Boeing's Lori Gunter
It has been a long time coming but we are finally reaching the beginning stages of 64-bit mainstream computing.
I guess you didn't hear about the Apple PowerMac G5 when it was announced months ago and began shipping last week, then....
I meant to say to accuse someone of theft when the "victim" wont tell you what they stole and the matter is still of legal dispute is irresponsible at best.
Warning: mysql_connect(): Too many connections in /home/httpd/vhosts/tweaktown.com/httpdocs/temp...
Welcome to the Slashdot effect. Maybe their servers need to be runnning dual Opertons?
If you outlaw the law, only criminals will have laws
The link says "AOL inks NFL broadcast deal". No mention of SCO anywhere.
Repeat after me: 64 bit processors aren't new. There's no new "computing fad" leaving the station. No new architectural wonder.
They aren't even new in desktop machines. I just threw out an Alphastation4 with a 64 bit 21064 from 1996 or something (nearly put my back out lugging the thing down the stairs. They built computers to last in those days). That was a competitor with the Pentium Pro IIRC. Many of the machines where I work were 64 bit ultrasparc before we started to go 32 bit x86.
That said, the new athlon does look pretty damn fast.
For those of you that wanted the grapics, I'll be taking an ASCII Art Course for the next one 8p
Introduction
We saw the release of the Opteron processor earlier this year which was powered by AMD's own 8131 Northbridge. With many months passed since and a lot of R&D invested, it was time to see what VIA were capable of when given the latest 64-bit processor to play with from AMD. The latest chipset to hit the market from the Taiwanese chipset manufacturer comes in the shape of the K8T800 and carries over everything you need when it comes to inserting those brand new processors into VIA-powered motherboards.
The K8T800 chipset carries over with it the latest Southbridge from VIA, the VT8237, which we were very impressed with when reviewing the PT800 only last month. It carries over the same features which we will look more into later on.
Due to the limitation in server hardware and server benchmarking software, we have chosen to write a preview on the chipset with a lot of benchmarks with different processor environments. We will help give people an understanding of the technology and the future it has.
Without dragging on anymore, let's see what VIA's K8T800 chipset is capable of when powered with some of the newest and most powerful processors to hit the market from Advanced Micro Devices.
The Opteron Processor
Before we get into talking about the chipset, we will take a moment to have a look at the processor which will be accompanying our test motherboard today.
The AMD Opteron is the latest processor to hit the market and carries the long awaited 64-bit architecture to the retail market. While we have seen Intel release a 64-bit processor to the market in the shape of the Itanium and Itanium 2 some time ago, it was one product that the end-user would not consider purchasing due to its high price tag and lack of chipset support.
The Opteron comes with a massive 940 pin count which is the largest of any processor implementing the PGA format. PGA or Pin Grid Array can be separated into two parts: the first being Pin, which in this case is the small metal pins we see coming from the bottom of the processor and the other is the Grid Array. The Grid Array is the formation the pins are in, which here is a 940 pin formation. Together we have the highest pin count currently on the market for this kind of technology. The Opteron processor also carries a huge 1MB of L2 cache to help when it comes to processor-hungry applications. The Opteron also has three Hyper Transport buses which we will look into more in just a moment.
What else AMD has in store for 64-bit
While we are on the subject of AMD processors, we may as well have a quick look at what else AMD have in store for the 64-bit market.
The cheapest of AMD's 64-bit line will be known as the Athlon 64 and will most likely be launched next month in Taiwan at Computex 2003. The Athlon 64 will be the entry level 64-bit processor and the mainstream product with a pin count of 754 replacing Athlon XP. Athlon 64 has the smallest amount cache out of the group with only 512k and will only support Single Channel DDR and one Hyper Transport Bridge meaning it is not SMP-capable.
The other processor which will be making an introduction, which seems to be confusing a lot of people, is Athlon FX or Athlon 64 FX. The Athlon FX will be the enthusiast processor, according to speculation currently floating around. Consisting of 1MB cache and three Hyper Transport buses, the only difference between it and the Opteron is support for SMP. The Athlon FX will consist of 939 pins; the pin that is missing is the one that gives it the ability to run Dual Processor mode.
Hyper Transport Buses
As we just mentioned, the Opteron and the Athlon FX will consist of multiple HT (Hyper Transport) buses, while the cheaper mainstream Athlon 64 will only have one.
One of the buses goes to your peripherals (your AGP slot, PCI slot, etc). This is the only one on the Athlon 64. The s
It takes 40+ muscles to frown, but only four to extend your arm and bitchslap the motherfucker
Someone should have used mysql_pconnect() !
"It was hell!" recalls former child.
I've been drooling for months, saving for an AMD 64 bit system. I MUST READ!
Stop overloading this website NOW!
The power of my mighty Irken fist commands you. Obey the fist!
Yes, but ... Although the G5 is a 64 bit processor,
The OS is not truly 64bit, it only has "Hook's" in the OS where 64bit is / might be needed. There are a lot of reasons why apple have done this, and this is open for debate. In my book its not truely 64bit until the whole of the architecture is 64bit and that includes the OS.
Electronic Music Made Using Linux http://soundcloud.com/polyp
to Apple sales ....
I'd like to see if the Opteron 64 has sold 240,000 computers!! (or the Alpha) Apple's preorder total for the G5 is right at 240,000 and the numbers are expected to grow by about 15% once fully stocked in stores before the end of September.
And to those; "it's a 32 bit OS" naysayers, Panther will be partially 64 bit and multithreaded. Quit with the Apple bashing and read facts.
Yell & scream & rant & rave... it's no use... you need a shaaaave ~ Bugs Bunny
Ummmm yeah ...
Apple market share is around 10-13%
Apple quarterly sales are around 4%
I allready rode the 64bit train to DEC Alphaville and back.
Microsoft Windows runs on stress and frustration.
The benchmarking software that would give us the opportunity to test the SMP Opteron platform to its fullest extent costs many thousands of dollars (Ed. Note - If any company wishes to sponsor us with this type of software, you are most welcomed!).
Thousands of dollars? Can't they just run Apache or something on it and show transactions per second comparisons serving some complex dynamic page?
Only in Steve Jobs' distortion field is this statement comprehensible
"Even at only 3% (which is in reality 10%)"
3% != 10%
G5 (tricking out 2 months later) != Opterons shipping for almost 3 months.
Show me where Apple has anything near 10%, a 64bit OS to run their $3000 headless machine, etc.
I'm using a dual Opteron (model 240, 1.4 GHz) for intensive scientific calculations. Our program, on the Opteron system, works 1.5 to 2 times faster than the 32 bit version that runs on a double Athlon MP 1800+ (1.533 GHz), depending on the algorithm.
It's like they've made two clones of that Iraqi information minister
and one of them's a Mac zealot and the other's a PC zealot.
...64-bit computing is boarding - don't miss the train!
Miss the train? Miss the train? How could you possibly miss the train? As 64 bit computing comes into the market, 32 Bit will gradually move out (not unlike 8 and 16 bit have previously done). So when you need buy new systems, you're gonna eventually end up buying into the 64 bit thing.
It's all good, but it's not like you need to buy now, or forever miss the train!
You can laugh without eating a sandwhich, but you can do both if bring one.
one dual cpu test could be
make -j bzImage
on the linux kernel. -j spaws a new process when ever possible. It is possible to limit the number of processes, with a number right after the -j. This trick naturally works on other compilations than the kernel, so just find something big, and give it a test.
Actualy when i read Toms and others test of the Opteron back when it was released i kind of missed tests that used OS tools and programs to test the system, using real programs.
Source Compilation:
Maybe you dont recompile kernels all day, but if you are a developper you compile source code, so for you a compile test is worth while.
Gimp:
Apple usualy does alot of photoshop benchmarks, so why not use gimp, a known image, and a known script for gimp to benchmark the system with that? Some people use imagemanipulation for work, so lets test real work.
It might be an idea to test other programs, possibly some engineering tool, maybe blender, or some other tool that uses the floating point unit, rather than the integer part, and or possibly something that can use and test SMP, and/or tests that is cpu bound, memory bound, (or disk bound).
I believe that the expression is "gerry-rigged", where "gerry" is what the British called the Germans in the two World Wars, in the same way that Americans referred to the Viet Cong as "charlie" in the Vietnam war.
Juries don't usually rig things, unless you count the OJ trial...
Putting the various other 64-bit processors aside (and remember, there have been > 32-bit processors since the 1970s), don't forget about the 64-bit MIPS processors. There's one in each and every PlayStation 2. Yes, there's only 32MB in a PS2, but the processor is still truly 64-bit. Integer registers are all 64-bits wide. Actually, they're 128-bits wide, but there's a limited number of operations that can be performed on 128-bit values
Sun has had 64 bit desktop boxes since long before Apple was producing anything more than a 'cooperative multitasking' joke OS that people laughed at.
A Good Intro to NetBS
Are we about to enter 18 months to 2 years of mostly running 32-bit apps on 64-bit hardware? (Or even longer than that, if we recall that "32-bit Win95" was really 16-bit in sheep's clothing.) And what's a "generation" for hardware? 2 years?
There will probably be some nice bargains on 32-bit boxen later this year & next, I'm sure. And they're gonna run those old 32-bit apps just fine. Then, when they wear out in 2-3 years, you can upgrade to 64-bit hardware to actually run 64-bit apps.
In other words, that first train trip is going to take you to an amusement park that's still under construction.
"Obviously, I'm not an IBM computer any more than I'm an ashtray" (Bob Dylan)
It could be said that the sun was not designed for mainstream computing, but the G5s are shipping as Dual 2Ghz 64 bit PPCs and this years Apple WWDC keynote went into the technical aspects (very high memory, IO, and processor bandwidth, or has Jobs put it, it can transfer a DVD from memory to processor in one second).
The 64 bit AMD is also good but will still require a recompile. LinuxPPC-G5 or Linux-x86-Opteron are both different than the stock vanilla PC Linuicies.
Apple/IBM and AMD have a very healthy competition going on. Wither (should it be wither) Intel and whatever happened to the unsyncable Itanic?
Of course mine boarded years ago. I still have my trusty ultrasparc 5.
I have not seen so much excitement about 64-bit since IBM released their 64-bit OS/400 operating system for the AS/400. SPARC, MIPS, and PA-RISC got much less fanfare.
:)
Now with AMD bringing it mainstream, I feel ready to whip up a 64-bit compiler to make use of the Opteron instrucion set, recompile 2.4.22 for 64-bit, and recompile a bunch of apps.
I think I'll start with a DVD recorder...
Oh yeah, I can barely keep up with what I'm doing now. Anyone else up for some assembler to get the compiler done?
btw - anyone remember how many cycles a Zilog Z80 needed to get that big a number through?
~8^]
-Just refresh a few times... that's what worked for me...
:
Oh man, sounds just like the chicks in accounting at the last company I worked for doing IS/IT work. I was working on a machine under the desk when I heard this beautiful conversation
G1 : I hit the print button but it isn't printing.
G2 : Sometimes they get lost on the way to the printer.
G1 : What do I do?
G2 : Print it again. Print it five times. Just keep hitting the print key until one of them finds its way to this printer.
Meanwhile I am under the table just losing it, totally unable to maintain composure. Somewhere in that office, I am guessing, there was a printer working doubletime just spewing copy after copy of that accounting report.
Glonoinha the MebiByte Slayer
VIA again? The steady rain of Linux problem reports tracked back to odd behavior from VIA chipsets has kept me from trying out any of AMD's newer designs, since the only boards I can find for them are always lumbered with VIA parts.
Stop it. Just stop it.
I got a dual Opteron w/2 gigs of ram and 80 gigs hard drive for $1,700, shipped.
:)
Not that much.
Instead of utilizing both on-chip memory controllers, it only uses one.
Can you recommend any chipsets/boards that use both controllers? Does the Tyan board use both controllers?
Thanks!
A benchmark toold does cost a lot of money.
I'm not really sure what exists for Linux benchmarking, but you have to remember those people are doing public (as in low level/everyone knows what I speak about)
So first we need 3DMark200X Linux version, then you must be able to install and actually play Quake 8)
So the problem is not just installing an SMP capable OS (Win2K pro does that just fine...) but to play with 2 cpu...
Most people I know do tests using photoshop, as it is optimized for SMP systems and the filters can be quite heavy.
A good old Divx Encode could have been deemed reasonnable too.
But basic usuers do understand Sandra and 3Dmark better.
Sad, isn't it ?
It takes 40+ muscles to frown, but only four to extend your arm and bitchslap the motherfucker
OK, so it does add some load to the server, but the MySQL database seems to be working on-and-off under the /.ing, so refreshing may get you in... BTW, at my school, MANY people are that stupid (the getting lost on the way to the printer crap)... and I have to delete the print jobs - on POS inkjets that are overloaded with documents before they get printed 20 times - not counting what the chicks print out multiple times... which usually has a dark background, so I have to replace the ink...
I think Nvidia should bring to the market an NForce chipset not only supporting dual Athlon64 chips, but also the AthlonXP/MP series. With 64bit WindowsXP delayed until next year, they [Nvidia] could really score with gamers with the dual XP chipset rig...perhaps AMD should give them a call...And please, Nvidia, go Apple's route and drop the legacy support. Drop the floppy port, the serial/PS2 ports, the parallel port, etc. The market needs USB2, Firewire, Wifi, Ethernet (some might wish for Gigabit)Bluetooth, SATA (and with RAID), 4 DIMM slots (not 3), and 5 or 6 PCI slots...thank you...
"Right now, somewhere in this world, Scott Baio is plowing a woman he doesn't love," - Peter Griffin, *Family Guy*
It's not going to come from 64 vs. 32 bit as much as from the 8 extra general purpose registers (GPRs). With 8 registers (x86 architecture) you can do very little register allocation, and consequently have to go to memory almost every time you do an operation. +8 registers changes this significantly. The register file is very fast to access and also has a fixed cost, whereas a memory operation can take a couple of hundred cycles (!!) if it misses in the L1, L2 and has to go to main memory.
The Raven
The train hasn't even left the station. Or are you attempting to tell us that 64-bit computing will be bug free from its inception?
I'm going to wait until the second generation of this technology emerges before I put my money on anything.
Great, so now we'll have two 64-Bit processors that won't run most of your apps because VIA has put hardware DRM on the motherboard!
I'm going with SiS for now, or maybe Intel.
#include <sig.h>
Sure, Opteron can do arithmetic with 64 bit integers (as can the Pentium 4 using SSE2 instructions). What makes a processor 64-bit is whether it can deal with 64-bit addresses. The Opteron's TLB supports only a 48-bit virtual address space and a 40-bit physical address space. Read the fine print here. Competing 64-bit processors such as the Itanium 2 support a full 64-bit virtual address space. It would be very difficult for the x86-64 platform to use 64-bit virtual addresses because of the heirarchical page table structure which it uses (based on the ia-32 page table structure which was designed for a 32-bit virtual address space).
4GB isn't very much once you start running a number of virtual machines for production use.
I also wonder if AMD has talked with the vmware and virtual pc people regarding what sort of instructions would be helpful and actually implemented them.
So... that sounds roughly what a dual Athlon MP 2800+ would perform at.
How does the cost compare to a dual Athlon MP 2800+
Bah! who cares - I am still waiting for a dual AGP slot mobo.
These benchmarks were done in 32 bit mode. Win XP Professional Edition Service Pack 4 doesn't run in 64 bit mode. There are operating systems which can run in 64 bit mode but they're not official Microsoft software.
...as long as XMame gives me a few more FPS in Hard Drivin' :o)
I've noticed that everyone who is for abortion has already been born - Ronald Reagan
There are at least two significant factors that would tend to offset the speedup obtained by adding AMD64's 8 extra general purpose registers: higher average instruction length (contributing to higher instruction read I/O), and 64-bit cache pollution.
Higher average instruction length is a factor because each 64-bit instruction (requiring use of 64-bit registers, or simply using any of the new "extended" general purpose registers) must be prefaced by a "64-bit mode" byte. That adds 8 bits to every 64-bit instruction. Increasing average instruction length has several consequences. First, the increased absolute size of code generates more instruction fetch traffic, using more memory bandwidth. Second, larger average instruction length decreases the effectiveness of the instruction cache because more space is needed to store the same number of total instructions (decreasing locality).
64-bit cache pollution is similar to the problem with the instruction cache, but generalized to include data. Using 64-bit data types and pointers increases (often unnecessarily) the size of each piece of data brought into a cache line. Once again, this increased size of each unit of data decreases data locality and increases the likelihood of capacity cache misses. The primary way to account for this is to simply increase the size of the data cache.
So in other words, it is not altogether obvious that the extra GPRs alone will compensate for these two issues that plague 64-bit computing. (For those who don't know why extra GPR's would speed things up: they offer the compiler some extra temporary storage space when performing computations, thereby reducing the number of data load/stores to main memory which are orders of magnitude slower than keeping things in registers.)
Now, don't get me wrong. I'm excited about these new AMD64 systems too. But I'm waiting for the legitimate 3rd party performance tests to come out so that I can verify the truth of these claims.
So, if I don't buy a VIA board today or in the next few days, I'll miss 64 bit computing? Then do all of the subsequent trains only take me to 32 bits, or do I go to 128 bits at that point? It's all so confusing.
2004 will be the true start of 64bit in the Desktop-World. 64Bit will be available for 100 extra in high-end consumer-products and 64bit will prepare itself to become the standard from 2005 onwards like 32bit and UDMA100 are standards today.
I am not in a hurry. My next system could show up 2004 or 2005 and it will have 64bit. Until then my good'old spring-2001 system will serve me well after I replaced the CPU with a Palomino2100+ and the GPU with a Geforce4200 for 250 in late 2002.
64 bit processors are all fine and dandy, but what I'd like to see are more 64 bit, 66/100/133Mhz PCI bus slots on motherboards.
Yes, Id like to put my gigabit NIC, my IEEE1394b storage silo and my Ultra320 RAID scratch volume on my 64 bit machine please. The KT800 chipset doesn't seem to cut it in this regard.
Kudos to Apple for doing this with the (higher end) PowerMac G5s. Now that's a machine I drool over.
Krishna
--- I'd love to go out with you, but I have to study for a Turing test.
And exactly at the same time, Sun had a brain-dead GUI while Apple had the GUI everyone else copied from (I'll spare you the links on this one).
Follow the link and look for yourself. There's no AGP on that board.
Cause until 2004 there wont be any
AMD64 Windows.
Now a MP 2600+ (the fastest I can find) it's 300 euro.I think the price of the Opteron 240 was around 400 euro, 3 months ago. So I think that now the performance/price ratio should be slightly higher for the Opteron... and it just can get bigger in the future. I'm looking forward to my next 4/8 processor Opteron machine!