CougTek
Hairy Aussie
Given the small price difference between the 1.6GHz and the 1.8GHz, I don't think the first one will sell as well as the later.
Here's Opteron naked :
Additional informations at a hardware website near you.
[NEWS] - Opteron launches today.
- Thread starter CougTek
- Start date
Here's Opteron naked :
Additional informations at a hardware website near you.
Pradeep
Storage? I am Storage!
CougTek
Hairy Aussie
I have just read it and loggued in to link it. Opteron has every reasons to succeed.
Pradeep
Storage? I am Storage!
Unfortunately I don't see the 240 @ 1.4GHz being competitive with the current Xeons. I will prob wait till the clock speeds scale up closer to 2GHz at the low end before building an Opteron duallie.
AMD needs to improve performance: Alot. I expected it to overwhelm and the reviews it is getting are showing mixed results. Further they have the same marketing problem other AMD processors have (Bad MHz with good IPC's). Then there is the expense: They are too expensive which may be forgivable for a server processor but only if performance matches.
All-in-All I'm not impressed and I wanted to be: very badly. However, this is the first of its kind for AMD and I'll wait to see what happens. Maybe they will ramp up the MHz quickly and ramp down the cost simultanously; Then with some marketing they will have a hit like the Athlons.
All-in-All I'm not impressed and I wanted to be: very badly. However, this is the first of its kind for AMD and I'll wait to see what happens. Maybe they will ramp up the MHz quickly and ramp down the cost simultanously; Then with some marketing they will have a hit like the Athlons.
I am in two minds (only two? mwwaahhhaaahhhaaa)
This is a cut-and-paste from a post I made on RWT, link here
http://www.realworldtech.com/forums...PostNum=1354&Thread=3&entryID=16715&roomID=11
the responses should be well worth reading
..."Vincent Diepeveen (diep@xs4all.nl) on 4/22/03 wrote:
---------------------------
>I assume (correct me if i am wrong) that more power means that it is harder to clock
>a cpu higher. What are your expectations on the clockrates of the opteron in this
>respect. Is the 'projection' from AMD that they will reach 2.8Ghz by end of 2004 for opteron a very realistic clockrate?
>
>If not, then what clockrate do you predict?
>
Probably the wrong person to ask as I'm not a "chippie".
Ahh what the heck, a good debate never hurt anyone )
Against clock rate
Thermal design constraints,
new process (SOI), and,
the design rules.
My understanding is that an MPU design is targetted at a particular design width. x86 MPU's are usually released at a design rule larger than targetted - the target and one process generation smaller. It would be most likely that AMD targetted 90nm when laying down Opteron/Athlon 64 as that is where they will spend the bulk of their "life". Most recent x86 MPU's have acheived quite large gains @ their targetted process generation. The big "if" in all of this is that Opteron is late - very late, so either some of the design assumptions were "off" or some of the gains aren't what AMD expected, so this will put AMD behind in their targets.
The way Intel traditionally (v AMD) has maintained their ASP's is to be ahead on the performance front - either by superior design or via better/more advanced process tech (or both). The question then becomes, how aggressively can Intel scale XEON? More importantly, how aggressively will VAR's and end users tolerate aggressive frequency scaling. The last time Intel tried to "hurry along" XEON, customers revolted and forced a more measured pace. This lead to fewer releases with larger performance delta's - which is not necessarily a bad thing. This does leave open the possibility that Intel will have larger performance gaps between XEON models in order to acheive a performance lead. I expect Intel to choose this method of gaining an advantage.
Intel will be debuting 90nm Prescott tech (and derivatives) _this_ year, which promises big gains, but Intel is also running into thermal issues.
AMD is not getting into the server space to get into a performance race - that's not the point of server chips, but they must remain competitive vis a vis Intel. But they have to learn their new SOI baby _and_ they cannot afford to falter - this should make them cautious.
Up AMD's sleave is larger directly addressable memory, and
x86-64. Expect AMD to push this line (as a direct upgrade path - read this is soothing, conservative, no troubles, easy) for all it's worth. In the space targetted, this is important and is becoming an issue now - x86-64, sorry AMD64 is secondary. There are enough x86 applications out there that will benefit from more than 4GB now or soon to make a difference IMHO. What's the good of more MPU performance if that's not the bottleneck? I expect any business that will care enough about memory/performance limitations will already know what their limitations are and the solution will be a "no brainer". It will either be more horsepower (HT-XEON) or more addressable memory (Opteron). Don't underestimate the power of inertia in difficult fiscal times. Platform changes - SPARC, POWER, EPIC, are costly (cost overrun), time consumming and difficult. Despite what the chip fandom brigade would have you believe, I see Opteron and XEON as being solutions to two different problems with some overlap. The key is Microsoft. The sooner AMD has an AMD64 version of Windows server, the happier AMD will be. AMD also needs the Linux brigade SuSE, Red Hat as well, but Microsoft will not be happy about that at all. If Microsoft stalls or fails to deliver, expect heartburn in Sunnyvale,Austin and Dresden.
I expect Intel to have their desktops at 4.5/5GHz by end 2004. XEON maybe 4GHz. For AMD to be competitive, 2.8GHz is probably the target (maybe a little higher). Is it acheivable? On 90nm with no slip ups - yes. AMD's track record _isn't_ good in this area _but_ they have had 15 months so far to work on SOI and they have another 18 before the end of 2004, so if it is possible to acheive 2.8GHz with their 90nm process then they should acheive it.
Caveats:
A lot would depend upon how successful Opteron is in the market place and how well received A64 is. If Opteron gets only a luke warm reception and A64 is well received, then AMD's emphasis _will_ change, as above all, AMD needs to be profitable. My understanding is that good thermal characteristics are diametrically opposite(slight exaggeration) to performance, and A64 needs performance above all else. If part of the delay of A64/Opteron is related to not enough resources to go around (as some have suggested), then two streams of process/design development may not be possible at AMD - this is where Intel has the wood over all competitors - and so Opteron may be de-emphasised. Funny, but Athlon 64 could be Opteron's own worst enemy.
All in all, this is my crystal ball gazing from limited knowledge and not enough education. Take it as you will. I hope some more knowledgable than I make their thoughts known."...
After that, I read an interesting piece by Tim Sweeney (which piqued my interest because of an ongoing thread on RWT about the need for 64 bit PC's). I am a programmer and I respect Time Sweeney's (and John Carmacks) abilities a lot - without falling into fanboyisms. This is what he had to say
http://firingsquad.gamers.com/features/sweeney_interview/default.asp
This is the sort of things that will drive 64-bit, not the 64-bitness, but the memory addressability. I know 64-bit is slower than 32-bit (all being equal), but working for an organisation that has 26 000 (no typo) employees and we use OS/390, Solaris and Windows NT servers with a mix of DB/2, Oracle, Apache, SQLServer and SAP R/3, the 2GB (effective) RAM limit of 32-bit is troublesome now.
Anand give a great real life example here:
http://www.anandtech.com/cpu/showdoc.html?i=1816
Factor in more RAM and 64-bit OS and XEON is a no show performance wise.
For the type of code I write integer for the decision trees, float for the financials, Opteron/A64 is the answer. The users don't care - they just want apps/PC's that make their jobs easier. I know what will give it to them. This isn't a criticism of the Pentium 4/Intel. The P4 was just designed with a different job in mind - multimedia, and I write custom business apps. But blazing fast multimedia performance won't help squat when you run out of memory and have to cache/write to disk and load disk info into memory. Do home users need 64-bit? - most don't.
If AMD get a toe-hold - and IMO they have their best chance ever - then Intel _will_ release a 64-bit x86 chip. They have no other choice.
AMD need M$ to deliver the OS
AMD need to execute flawlessly - like they did with Athlon originally
Neither is a given.
0.05
This is a cut-and-paste from a post I made on RWT, link here
http://www.realworldtech.com/forums...PostNum=1354&Thread=3&entryID=16715&roomID=11
the responses should be well worth reading
..."Vincent Diepeveen (diep@xs4all.nl) on 4/22/03 wrote:
---------------------------
>I assume (correct me if i am wrong) that more power means that it is harder to clock
>a cpu higher. What are your expectations on the clockrates of the opteron in this
>respect. Is the 'projection' from AMD that they will reach 2.8Ghz by end of 2004 for opteron a very realistic clockrate?
>
>If not, then what clockrate do you predict?
>
Probably the wrong person to ask as I'm not a "chippie".
Ahh what the heck, a good debate never hurt anyone )Against clock rate
Thermal design constraints,
new process (SOI), and,
the design rules.
My understanding is that an MPU design is targetted at a particular design width. x86 MPU's are usually released at a design rule larger than targetted - the target and one process generation smaller. It would be most likely that AMD targetted 90nm when laying down Opteron/Athlon 64 as that is where they will spend the bulk of their "life". Most recent x86 MPU's have acheived quite large gains @ their targetted process generation. The big "if" in all of this is that Opteron is late - very late, so either some of the design assumptions were "off" or some of the gains aren't what AMD expected, so this will put AMD behind in their targets.
The way Intel traditionally (v AMD) has maintained their ASP's is to be ahead on the performance front - either by superior design or via better/more advanced process tech (or both). The question then becomes, how aggressively can Intel scale XEON? More importantly, how aggressively will VAR's and end users tolerate aggressive frequency scaling. The last time Intel tried to "hurry along" XEON, customers revolted and forced a more measured pace. This lead to fewer releases with larger performance delta's - which is not necessarily a bad thing. This does leave open the possibility that Intel will have larger performance gaps between XEON models in order to acheive a performance lead. I expect Intel to choose this method of gaining an advantage.
Intel will be debuting 90nm Prescott tech (and derivatives) _this_ year, which promises big gains, but Intel is also running into thermal issues.
AMD is not getting into the server space to get into a performance race - that's not the point of server chips, but they must remain competitive vis a vis Intel. But they have to learn their new SOI baby _and_ they cannot afford to falter - this should make them cautious.
Up AMD's sleave is larger directly addressable memory, and
x86-64. Expect AMD to push this line (as a direct upgrade path - read this is soothing, conservative, no troubles, easy) for all it's worth. In the space targetted, this is important and is becoming an issue now - x86-64, sorry AMD64 is secondary. There are enough x86 applications out there that will benefit from more than 4GB now or soon to make a difference IMHO. What's the good of more MPU performance if that's not the bottleneck? I expect any business that will care enough about memory/performance limitations will already know what their limitations are and the solution will be a "no brainer". It will either be more horsepower (HT-XEON) or more addressable memory (Opteron). Don't underestimate the power of inertia in difficult fiscal times. Platform changes - SPARC, POWER, EPIC, are costly (cost overrun), time consumming and difficult. Despite what the chip fandom brigade would have you believe, I see Opteron and XEON as being solutions to two different problems with some overlap. The key is Microsoft. The sooner AMD has an AMD64 version of Windows server, the happier AMD will be. AMD also needs the Linux brigade SuSE, Red Hat as well, but Microsoft will not be happy about that at all. If Microsoft stalls or fails to deliver, expect heartburn in Sunnyvale,Austin and Dresden.
I expect Intel to have their desktops at 4.5/5GHz by end 2004. XEON maybe 4GHz. For AMD to be competitive, 2.8GHz is probably the target (maybe a little higher). Is it acheivable? On 90nm with no slip ups - yes. AMD's track record _isn't_ good in this area _but_ they have had 15 months so far to work on SOI and they have another 18 before the end of 2004, so if it is possible to acheive 2.8GHz with their 90nm process then they should acheive it.
Caveats:
A lot would depend upon how successful Opteron is in the market place and how well received A64 is. If Opteron gets only a luke warm reception and A64 is well received, then AMD's emphasis _will_ change, as above all, AMD needs to be profitable. My understanding is that good thermal characteristics are diametrically opposite(slight exaggeration) to performance, and A64 needs performance above all else. If part of the delay of A64/Opteron is related to not enough resources to go around (as some have suggested), then two streams of process/design development may not be possible at AMD - this is where Intel has the wood over all competitors - and so Opteron may be de-emphasised. Funny, but Athlon 64 could be Opteron's own worst enemy.
All in all, this is my crystal ball gazing from limited knowledge and not enough education. Take it as you will. I hope some more knowledgable than I make their thoughts known."...
After that, I read an interesting piece by Tim Sweeney (which piqued my interest because of an ongoing thread on RWT about the need for 64 bit PC's). I am a programmer and I respect Time Sweeney's (and John Carmacks) abilities a lot - without falling into fanboyisms. This is what he had to say
http://firingsquad.gamers.com/features/sweeney_interview/default.asp
This is the sort of things that will drive 64-bit, not the 64-bitness, but the memory addressability. I know 64-bit is slower than 32-bit (all being equal), but working for an organisation that has 26 000 (no typo) employees and we use OS/390, Solaris and Windows NT servers with a mix of DB/2, Oracle, Apache, SQLServer and SAP R/3, the 2GB (effective) RAM limit of 32-bit is troublesome now.
Anand give a great real life example here:
http://www.anandtech.com/cpu/showdoc.html?i=1816
Factor in more RAM and 64-bit OS and XEON is a no show performance wise.
For the type of code I write integer for the decision trees, float for the financials, Opteron/A64 is the answer. The users don't care - they just want apps/PC's that make their jobs easier. I know what will give it to them. This isn't a criticism of the Pentium 4/Intel. The P4 was just designed with a different job in mind - multimedia, and I write custom business apps. But blazing fast multimedia performance won't help squat when you run out of memory and have to cache/write to disk and load disk info into memory. Do home users need 64-bit? - most don't.
If AMD get a toe-hold - and IMO they have their best chance ever - then Intel _will_ release a 64-bit x86 chip. They have no other choice.
AMD need M$ to deliver the OS
AMD need to execute flawlessly - like they did with Athlon originally
Neither is a given.
0.05
Fushigi
Storage Is My Life
But the real question is this: How fast will an Opteron be when it comes to Folding? :lol:
- Fushigi
- Fushigi
Dïscfärm
Learning Storage Performance
Fushigi said:
I'd be more interested in how fast the Opteron is at Bluffing. :erm:
Howell
Storage? I am Storage!
You gotta know when to hold 'em.......
Fushigi
Storage Is My Life
BTW, IBM will release an Opteron-based server later this year. Read about it. If this holds true, this could be the Tier-1 support AMD needs for legitimacy in the corporate world.
Funny it would come from the manufacturer of another 64-bit CPU.
- Fushigi
Funny it would come from the manufacturer of another 64-bit CPU.
- Fushigi
Fushigi
Storage Is My Life
Opterons announced today for 1-way and 8-way boxes. http://www.infoworld.com/article/03/06/30/HNnewopteron_1.html
- Fushigi
- Fushigi
blakerwry
Storage? I am Storage!
Fushigi said:
http://www.amdmb.com/article-display.php?ArticleID=240
Review of the opteron's performance (1 way and 4 way) at Seti crunching....
Santilli
Hairy Aussie
- Joined
- Jan 27, 2002
- Messages
- 5,273
Real question for me is:
How expensive is the MSFT OS going to be in 64 bit, and will it run on the Athlon Opertron?
In other words, is there going to be some factor that makes the entire question so expensive, much like duals/mobo have been for a while, that you are just financially wiser to buy a single 32 bit chip???
s
How expensive is the MSFT OS going to be in 64 bit, and will it run on the Athlon Opertron?
In other words, is there going to be some factor that makes the entire question so expensive, much like duals/mobo have been for a while, that you are just financially wiser to buy a single 32 bit chip???
s
Mercutio
Fatwah on Western Digital
The better question is. do you actually have an application in mind where investment in a 64-bit platform is a sound decision?
And if so, why is x86 compatibility important
And if x86 compatibility is important, are you fully aware that none of the normal reasons for sticking with x86 apply in this case (e.g. There's no MS Office for 64-bit x86)?
And if so, why is x86 compatibility important
And if x86 compatibility is important, are you fully aware that none of the normal reasons for sticking with x86 apply in this case (e.g. There's no MS Office for 64-bit x86)?
Ahh, not so Mr. Mercutio.
The big selling factor is even if there is no 64-bit x86 binary, that the 32-bit binary will run just fine as is, and as fast as the 32-bit binaries will on a 32-bit x86 machine.
In the past, you could run x86 binaries via software emulation FX!32 (Alpha) or via hardware emulation (Itanium) at the speed of an 800MHz Pentium !!!. Neither are an ideal solution.
x86 compatibility is important because just about everything useful is available on x86.
There is too much software available on x86 for it to fade away in the next few years.
The big selling factor is even if there is no 64-bit x86 binary, that the 32-bit binary will run just fine as is, and as fast as the 32-bit binaries will on a 32-bit x86 machine.
In the past, you could run x86 binaries via software emulation FX!32 (Alpha) or via hardware emulation (Itanium) at the speed of an 800MHz Pentium !!!. Neither are an ideal solution.
x86 compatibility is important because just about everything useful is available on x86.
There is too much software available on x86 for it to fade away in the next few years.
Mercutio
Fatwah on Western Digital
Blah blah blah compatibility blah blah blah alpha blah blah x86 blah blah.
There aren't many reasons to run 64-bit programs on the desktop. That will change someday, I know, but Santilli asked about compelling reasons to purchase a 64bit x86 platform today.
Yup, you can run 32-bit Win2000/XP on one. No particular benefit to doing that, unless you call "costing more money" a benefit. The benchmarks in some cases are better than Athlon or P4, but... not so much that 64-bitness is worth the money right now.
Your mainstream apps aren't available in 64-bit versions. No advantage there. You can run the 32-bit version, sure, but you could do that no matter what you bought.
In short, I don't think what we're looking at is particularly compelling right now.
There aren't many reasons to run 64-bit programs on the desktop. That will change someday, I know, but Santilli asked about compelling reasons to purchase a 64bit x86 platform today.
Yup, you can run 32-bit Win2000/XP on one. No particular benefit to doing that, unless you call "costing more money" a benefit. The benchmarks in some cases are better than Athlon or P4, but... not so much that 64-bitness is worth the money right now.
Your mainstream apps aren't available in 64-bit versions. No advantage there. You can run the 32-bit version, sure, but you could do that no matter what you bought.
In short, I don't think what we're looking at is particularly compelling right now.
Fushigi
Storage Is My Life
Keep in mind that Opterons are being positioned solely as a server chip. You'll see x86-64 apps in the channel this fall, I'm fairly certain. Several are ready now. Anyway, buy it as a Xeon alternative today that'll run even better tomorrow.
On the desktop, yes, there's nothing current that Opteron will shine in. You'll need 64-bit Photoshop, Autocad, or some other psuedo-workstation app to justify the Opteron premium.
<AS/400 bias>
Welcome to the game, x86-land. You're only slightly behind where IBM was in 1995.
</AS/400 bias>
- Fushigi
On the desktop, yes, there's nothing current that Opteron will shine in. You'll need 64-bit Photoshop, Autocad, or some other psuedo-workstation app to justify the Opteron premium.
<AS/400 bias>
Welcome to the game, x86-land. You're only slightly behind where IBM was in 1995.
</AS/400 bias>
- Fushigi
Maybe not everyone needs 64-bit x86 Merc, but some do, or think they do
http://firingsquad.gamers.com/features/sweeney_interview/default.asp
x86-32 is limited to about 2GB of RAM for data, the other 2GB is taken up by the OS - though you can make this 3:1 in favour of data with some trade-offs.
64-bit x86 might be compelling for content generation. Maybe. I don't know. I don't know enough to comment, but Tim Sweeney does I'd imagine.
http://firingsquad.gamers.com/features/sweeney_interview/default.asp
x86-32 is limited to about 2GB of RAM for data, the other 2GB is taken up by the OS - though you can make this 3:1 in favour of data with some trade-offs.
64-bit x86 might be compelling for content generation. Maybe. I don't know. I don't know enough to comment, but Tim Sweeney does I'd imagine.
Howell
Storage? I am Storage!
honold said:
http://www.ansys.com/services/hardware_support/parallel/
plus
Datamining, ERP and CRM software would all benefit I suspect.