Wait? (for CPU upgrade)
Here's my specs:
Athlon XP 2400+ (12x166) - Epox 8k3a - 512MB Samsung PC2700 @Fastest 2.8v - Radeon 9700 Pro AIW - Santa Cruz sound - 80GB WD/80GB Maxtor - Toshiba 16x DVD - Lite-On 48x12x48 CDRW - Lian Li PC600 - 15" NEC Flat Panel
I get really good frame rates when gaming (about 55 fps in Unreal II with 4x AA, 6x AF @1024x768), but found that when I really crank things up the system starts to show it's bottleneck...
The 2400+ runs at default clock but 166mhz bus. Decent performance, but 2,000mhz isn't enough to take full advantage of the Radeon 9700 Pro.
My memory's going to stay the same for now, I've done the benchmarks myself that show that the difference between turbo and fast settings doesn't warranty the extra $$ required to get memory capable of running the most aggressive timings.
I'm probably going to find a 17" LCD down the line, but I really can't justify the expense now. CD burner will hold out until I get a DVD+/-R. Sound card may get upgraded, at this moment it still does everything I need... and I already have 80gb more storage than I need
Athlon XP 2400+ (12x166) - Epox 8k3a - 512MB Samsung PC2700 @Fastest 2.8v - Radeon 9700 Pro AIW - Santa Cruz sound - 80GB WD/80GB Maxtor - Toshiba 16x DVD - Lite-On 48x12x48 CDRW - Lian Li PC600 - 15" NEC Flat Panel
I get really good frame rates when gaming (about 55 fps in Unreal II with 4x AA, 6x AF @1024x768), but found that when I really crank things up the system starts to show it's bottleneck...
The 2400+ runs at default clock but 166mhz bus. Decent performance, but 2,000mhz isn't enough to take full advantage of the Radeon 9700 Pro.
My memory's going to stay the same for now, I've done the benchmarks myself that show that the difference between turbo and fast settings doesn't warranty the extra $$ required to get memory capable of running the most aggressive timings.
I'm probably going to find a 17" LCD down the line, but I really can't justify the expense now. CD burner will hold out until I get a DVD+/-R. Sound card may get upgraded, at this moment it still does everything I need... and I already have 80gb more storage than I need
0
Comments
Enisada
I would, but it's an AIUCB (poor overclocker/early revision) and runs pretty hot at default voltages and air.
Because if it is then when do they come down in price
At least 3 months after release, sometimes longer.
Craig
And their pricing is through the roof!
Enisada
I think I'll wait for the 3000/3200+ to come down in price, then grab one of them when they hit around $100. Thanks for the perspective.
So i think it will be better for me just to wait until 64 AMD's come down in price
Or get a £40 XP1700+ and some good cooling and Overclock it to 3200+ for less than half the price, only thing you'd be missing is the extra cache.
Craig
The Tbred 2.00ghz is a 2500+ @ FSB333. You system is fine. The only reasonable upgrade available to you is a NF2mobo /Barton 2500 (1.83ghz) combo.
I have one doing FSB333 2.00ghz on air (PR2700+) and FSB333 2.33ghz (PR3300+) on water
Or... you can do what us overclockers do, overclock!!! Overclock that videocard and cpu man! :P
I haven't met a single person thats is blowing away the fps on that game. From all the posts i have read you are getting some of the highest fps I have seen. The game was released way too early and quite frankly is just a pos.
At the time of release of that game you pretty much had the best gear to play it and still your not getting awsome fps. It's not your system it's there Game
I get an avg of 30fps in no action and it will drop down to 20- in high action scenes and i have a 9500pro over clocked to the max. My 3dmarks03 are 4100. Should be more than enough to play unreal2 well but it's not.
What is happening is this-- some people think the Athlon64 and the Opertons are the same. This is not quite true, though you will definitely see a commonality of performance between a 1-way Opteron and an Athlon64. The Opterons are more for high-end workstations and servers, one line derived from the "Sledgehammer" will be scalable 8 ways-- but will be a smaller form factor internally, it will be a 90 nm core implementation rather than the current 130 nm implementations.
The Athlon64 is being marketted toward replacing the Athlon, and the processor map and other info I have on Athlon64 say it is a year plus to 9 months away minimum-- for launch info, even. How fast it comes out will depend in part on P4 800 FSB cpu popularity, as it should be the AMD entry to go head to head with those.
Opteron will go against the newer Zeons and the next Intel server gens that are still "code named." It probably will end up against Itanium 2 in one of its lines also-- probably 2 and\or 4 way Opterons will go against higher-end Itanium 2's, depending on the clock rates for each and the relative pricings. Expect to see Itanium 2 boxes from Dell, HP, Compaq, and possibly some servers from IBM based on them also. Exepct multiple OS platforms for them also.
The Athlon64 Linux subversions WERE deved on one way Opteron-sample testbeds for the most part. I would expect Athlon64 to be very close to a one-way non-scalable Opteron, and the Opteron name then applied to multi-CPU scalable CPU lines possibly up to 8-ways. My understanding right now is that it will be a 90 nm core and the precise timing of the full scale fabbing of that size is being debated right now-- will have to wait and see, but expect Germany to know first as the fab facility most likely to be used is in Dresden, Germany.
John Danielson.
UT 2k3 was deved on boxes that used nVidia. Expect better frames for this game on nVidia boxes, and nforce2 nVidia boxes. Tihs will be true of many games deved on Linux render farms, and games that use graphics from movies or videos rendered on Linux render farms.
Every piece of software is biased toward what it was deved on and thus runs best on it. The people doing the dev know the fine details of what they dev on and will be a ble to tune that. Unless multiple versions are released for different hardware sets, the cross-compatibility stuff will be addon interpretive stuff that results in poorer performance on things the game was not deved for-- and ON.
John Danielson.
Currently all AMD CPUs are .13 micron including the 1xx, 2xx,& 8xx series Opterons. 90 um CPUs aren't due from AMD until 2004.
The Athlon 64 (754pins) and Opterons (940pins) have the same core, the difference being that only one HT link is operable on the Athlon 64 and it only supports single channel (64bit) DDR333 memory (at the moment).
Opterons have 2 or 3 HT links (for Dual or 4/8 way CPU operation) and support Dual channel (128bit) DDR333 ECC memory access.
You'll only see differences in performance when memory bandwidth is important.
IMO As currently configured the A64 won't live up to expectations. Get a Barton.
And if you overclock PC3500 or PC3700 might come in handy, especially if you overclock RAM bus rates. And he did spec DDR266 as what he had now, yes, so PC3200\DDR333 is base for what he wants to do.
So Opteron 4 ways will match newer Itanium 2s with P4 SMP cores, except will be 4 way.
I am not ripping on you, but you keep saying the same basic thing I just did....
I'm not ripping you either, I felt your explanation needed clarification, for example...
Let's reverse your sentence - "The Opterons are more for high-end workstations and servers, one line derived from the "Sledgehammer" will be scalable 8 ways, rather than the current 130 nm implementations, it will be a 90 nm core implementation.
As you can see the way you wrote the sentence it can be interpreted that 8xx series Opterons will be 90micron while 1xx & 2xx Opteron are 130 micron. As for Will Be, the Opteron also Will Be 65 nanometer and 50µ, but neither 90, 65, or 50 nm versions are currently being produced (Intel is making 90µ CPUs however).
AMD is notoriously late introducing new process technology to their CPU line. It took them 6 extra months to begin producing 13micron CPUs compared to Intel(Tbred-A Apr 10th). And another 6 months after that to work out the bugs in their design (Tbred-B paper launched in Aug, not available in quantity until Nov). I don't expect to see a 90µ CPU from AMD for at least another 12 months (Not from the Dresden Fab anyway,maybe IBM will make them for AMD instead).
Also the Athlon 64, Opteron 1xx, 2xx, & 8xx compete against the Pentium 4, P4 Xeon (2 way), and P4 Xeon MP (2-8way) processors. The Opteron does not compete against the Itanium Processor at all. Only because both CPU are 64 bit do some people think so. The P4 (Xeon), Opteron, and G5 all compete in the same space, were as the Itanium competes with the Alpha, IBM Power 5, HP PA-RISC, and SUN Sparc64 processors.
http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_608,00.html
The sledges give way to "Athens," and sledges are 4-8 ways with multiple subversions based on scaling capacity.
There are (see map for why I say ARE) two variants of Athlon64's, the desktop version and the mobile Athlon64. Both will first be released as will be what was coded "Clawhammer," and claw was a 1-2 way code. The desktop "claw" will give way to SanDiego, the mobile will give way to "Odessa." Odessa is likely to speed step, as a competitive move against the P4-Mobile, which does NOW, and might also be a lower voltage core so as to gen less heat.
The May map says that they will be 90nm cores, ALL of them, sometime in about the first half of 2004. I am thinking that this might get accellerated.
I also do not know that AMD might not make a dual-cored desktop processor in the Athlon64 line.
That would fit engineerings def of "Clawhammer."
I also do not know if the 1-2 way Opterons will not simply be renamed as Athlon64's(In so far as the core used), and possibly the one-way depiped as you say(If it was in fact ever piped, AFAIK 1-way was only a single pipe out, the 2-way claw was talked about as being one or the other (1 or 2 out), and the Sledges were talked of as 2-3 output pipes.
So although I was a tib vague, it was because I heard contrary things over the passage of a year's time from reputable journalists who were interviewing AMD high level folks. But at present, AMD, in their online virtual pressroom, is indicating that all the opterons and the Athlon64's will be 90 nm cores during 2004 at the latest.
One reason I say at the latest is that Linux devs have folks in AMD engineering helping them, and those folks said to dev on Opteron bases and helped get the Linux folks sample platforms to dev on. And that relates with the timing of the Athlon64 subversion of Linux by two distros.
I would almost bet my remaining eye tooth that the cores of the Opteron 1 and 2 ways will get transplanted onto packages that fit the new forthcoming socket that does not take the 4-8 ways and that those CPUs will get called Athlon64's--and that this will happen about the time the cores are transitioned to 90 nm production.
Similar to the way the new Celerons have essentially the old P4 core, and similar to the way someone stuck a 1.8 GHz Willamette core(product code BX80531NK180GSL5UK-- from a saved box end panel label) on a 478 package P4 die to go with Brookdale chipsets(Soyo Fire Dragon), and similar to the way that the newest Durons use the old ATHLON core. History has this habit of repeating itself,and Iexpect a "do it again"pattern.
I would expect the Opteron 8-way implementations to indeed compete in the market with the Itanium 2's. The 2 ways cannot. The 4-ways would probably lose. Since it is too soon to tell exactly what will be decided on before it happens,and since things like Opteron are not engineering's names but are marketting's line name for what Engineering considers multiple cores, I try not to get too specific too early.
Just amplifying a bit.
John Danielson.
<a href="http://www.short-media.com/forum/showthread.php?s=&threadid=1180" target=_blank>Forum post</a>
Dual Operation requires 2 HT links, more than 2 CPUs requires 3 HT links. All Opterons 1xx, 2xx and 8xx series use Socket 940. To my knowledge no other Opteron socket (other than a 939 pin Socket for Athlon 64 version) is planned.
<a href="http://www.aceshardware.com/read.jsp?id=50000247" target=_blank>Aces Hardware: AMD 0.13um 2-way Opteron systems</a>
John Danielson.
4 way Opteron pic (to get 8 just add CPUs..)
Itaniums use 130watts
Current Opteron only use 40-60 watts of power. An 8 way Opteron system is easily produceable
Sledge was speced as 4-8 way, and that is scaled on die. I think the rumored heat has to do with running 4x the number of circuits in same or close to same space with a process that is 3\4 the size of the old one(much less the same nm density), and that there are no problems with the smaller-number-of-circuit per cubic inch or cubic mm claws.
In respect to heat, If you can do a Blade implementation, you can do any number.
Fujitsu Opteron Blade system pic
That means to get 2 4's on one die either you have a die overhang the flat socket (not too good), or you stakc two layers of 4 each. If you stack,bottom layer stays hotter by a bunch. Power times two for 8 versus 4 plus cooling problems.
It is not so much what happens when you add CPUS, it is what happens when you try to get twice as much in very close to same cubic space and try to deal with the resulting heat complexity that you run into things that current computers have problems calculating for you. But you will definitely have a heat management problem when going to 8 on-die unless you go to a 60-70nm density-- and that is not doable right now.
John Danielson.