For the "fun" of it, I decided to transform the list into a database. It's really interesting how many, for example, 3200+ CPUs there are, across sockets and cores. It's on my home PC, so I can't even find the best example for you, but I can see how you could be confused. I should do the same for Intel's, I think there's is more confusing. For example, a 520 and 518 differ in FSB speed -- the 520 has a 800MHz / 518 has 533MHz. It's a big difference but is only worth 2 points (assuming the 518 runs slower than the 520; if the 533MHz bus chip runs faster than a 520 but slower than 530 then it's a 522, if that exists, it's all theoretical since I'm not looking).
It took me a while, but I have found the answer myself.
From what I have seen from the Tyan Tomcat K8E S2865 it will play with 939 Opterons as well as any s939 CPU, though it doesn't feature the nF4 Pro.
Foxconn makes their NFPIK8AA-8EKRS for s940 CPUs NVIDIA nForce Pro 2200 + nForce Pro 2050 and nicy nice TI firewire controllers a+b. If they can swap the 2050 for a couple of PCI slots on a s939 version, that would be a nice board.
What you do is buy a socket-939 system now and a single core chip since they're pretty cheap these days. Then in maybe 6-12 months, pick up a dual-core chip when those prices have fallen a bunch so you're set for games that might finally arrivee that take advantage of the second core somehow.
I just replaced the Winchester 3000+ in my system with an X2 3800+. My compile times have sped up drastically. On the 3000+ compiling my source tree took about 4 minutes. With the 3800+ and serial make, it's 3:22 and the system shows 50% idle time. With parallel make -j2 it takes 1:50 and the system shows about 15% idle time. With -j3 it takes 1:45 and about 1% idle time. I'm pleased.
man those compile times are great but for noobs that are new to computers and only use it for surfing and games --- I think its still not really recommended that they get dual core
a good analogy i can think of is turbocharging in cars -- too often i see people with twin turbocharger cars that dont even rev their engines beyond the 3000rpm limit for the turbo's to kick in. What a waste!!! and how stupid (eg. always driving around with turbo lag)
I suppose; for noobs it would just mean they can spread viruses/worms twice as fast.
The turbo analogy is good, you've hit another topic near and dear to my heart. :P
(Of course with a properly sized turbo, "lag" is mostly a non-issue.)
lol, im assuming you mean dual core...your topic said dual but the post said single... : )
As of right now the big push for dual core is for those (like myself) that multitask alot. Yeah, its not a good idea to multitask while gaming and such, but it would be nice to be able to do whatever while im say, burning a dvd with dvdshrink. Or i suppose (depending on how powerful they are, as i dont own one) I could actually probably game while burning a dvd... who knows maybe someone with a dual core can chime in.
I'm sure that once the big name companies start pushing it then software companies will start coding for it.
Well, a lot of workstation applications already are SMP/SMT-aware, since most workstations are built off of Xeon or Opteron setups, which are usually dual or quad processor. Apps like Photoshop, CAD, and 3dStudioMAX all are optimized for dual-core. Here's a complete listing.
The problem with AMD (right now) is that they dont have a low speedgrade, and I'm sure this is a purely yield-related Issue. Intel can make cheap because they have much more capacity than AMD (something fab36 will hopefully help AMD deal with) and because they can just slice off a core if it turns out to be defective, effectively making their failed Dual-cores into single-core variants (this is because event hough the 2 cores are on the same die, they are simply side-by-side for the most part).
(Yes, I AM fudging on what each CPU "equals". The fact of the matter is that in most instances, the 3000+ is actually faster than a 3.0 GHz P4, and that applies to the others as well.)
So, 2x2.0GHz 3800+ is really equal to 2x3.2GHz Pentium D 840. There really is not Intel match for the 4200+ and above parts - the 4200+ is faster in almost every case than even the 840D Extreme Edition. (Tests under Windows XP with four CPU-intensive tasks will put the 840 EE ahead, but that's a Windows scheduler problem.)
No, the 2x2.0GHz 3800+ is closest to an 830 (2x3.0GHz). The 840 is faster in most benchmarks.
It's a little difficult to compare the two, really. The P4 was always faster than the Athlon (64) in certain apps, mainly encoding, while the Athlon was always faster in other things (like everything else, especially gaming). When you put two cores on one PCB, however, things can change. The Athlon X2's design is more advanced than Intel's. AMD really thought it out and made it backwards compatible with Socket 939, at the cost of memory bandwidth rather than invent a new socket. Intel instead requires a new motherboard that offers no real new features other than Pentium D support.
Basically, if you already have a Socket 939 board, it would make more sense to grab an X2 as an upgrade rather than jump ship and go Pentium D. You have to add the cost of a motherboard, something that everyone buying a Pentium D needs, so Intel can offer the chip at a lower cost somewhat, because they are also selling a Northbridge/Southbridge (whatever they decided to call it, that's what they really are). They also put next to no tech design into the Pentium D, little research money comparitivley. They literally glued two cores together and changed the northbridge/BIOS to support the chips. The architecture map shows it, the two cores go through the FSB to talk to each other, just like in a dual Xeon system which has two independent sockets. AMD's cores don't hit the FSB to talk, which can help out in many situations. Unfortunately for AMD, Intel's 800MHz FSB is still fast enough to run two 3.0GHz chips fast enough to compete. If Intel only had DDR FSB (effectively a 400MHz bus) I don't think they could have come out with the Pentium D at the speeds they did. Fortunately for Intel, in their quest to produce the 3.8GHz chips, some fall short, way short, like 2.8 or 3.0, and when you pair them together, glue them together (we saw the IDF pic of a chip w/o it's heat speader) and call it a Pentium D, it works really well for them.
If you had a P4 now, though, it probably makes more sense to get a Pentium D, unless you were looking for leading edge, in which case the 4600+/4800+ (if you just have to have 1MB L2 for each core) is your target.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
16 Comments
Back to Article
ElFenix - Monday, September 5, 2005 - link
these PR numbers don't tell me anything =(andrewln - Tuesday, September 6, 2005 - link
mhz does not mean anything tooElFenix - Sunday, September 11, 2005 - link
mhz is very useful for comparing between the same families of processors.ideally they're be model/pr number, core mhz, bus mhz, and l2 cache size in these tables, i think.
joex444 - Tuesday, September 6, 2005 - link
http://en.wikipedia.org/wiki/List_of_AMD_Athlon_64..."> has a list of all the Athlon 64s, and below it has links to other types of CPUs where you can get a list like the one above from the link in See Also.For the "fun" of it, I decided to transform the list into a database. It's really interesting how many, for example, 3200+ CPUs there are, across sockets and cores. It's on my home PC, so I can't even find the best example for you, but I can see how you could be confused. I should do the same for Intel's, I think there's is more confusing. For example, a 520 and 518 differ in FSB speed -- the 520 has a 800MHz / 518 has 533MHz. It's a big difference but is only worth 2 points (assuming the 518 runs slower than the 520; if the 533MHz bus chip runs faster than a 520 but slower than 530 then it's a 522, if that exists, it's all theoretical since I'm not looking).
ceefka - Monday, September 5, 2005 - link
Would these boards require a 939 Opteron or do they also play with Venice, Toledo or Machester (X2)?ceefka - Monday, September 5, 2005 - link
It took me a while, but I have found the answer myself.From what I have seen from the Tyan Tomcat K8E S2865 it will play with 939 Opterons as well as any s939 CPU, though it doesn't feature the nF4 Pro.
Foxconn makes their NFPIK8AA-8EKRS for s940 CPUs NVIDIA nForce Pro 2200 + nForce Pro 2050 and nicy nice TI firewire controllers a+b. If they can swap the 2050 for a couple of PCI slots on a s939 version, that would be a nice board.
Tides - Sunday, September 4, 2005 - link
Is there really any gain from single cores atm? When will we see apps/games/programs take advantage of this.yacoub - Tuesday, September 6, 2005 - link
What you do is buy a socket-939 system now and a single core chip since they're pretty cheap these days. Then in maybe 6-12 months, pick up a dual-core chip when those prices have fallen a bunch so you're set for games that might finally arrivee that take advantage of the second core somehow.highlandsun - Sunday, September 4, 2005 - link
I just replaced the Winchester 3000+ in my system with an X2 3800+. My compile times have sped up drastically. On the 3000+ compiling my source tree took about 4 minutes. With the 3800+ and serial make, it's 3:22 and the system shows 50% idle time. With parallel make -j2 it takes 1:50 and the system shows about 15% idle time. With -j3 it takes 1:45 and about 1% idle time. I'm pleased.xsilver - Monday, September 5, 2005 - link
man those compile times are great but for noobs that are new to computers and only use it for surfing and games --- I think its still not really recommended that they get dual corea good analogy i can think of is turbocharging in cars -- too often i see people with twin turbocharger cars that dont even rev their engines beyond the 3000rpm limit for the turbo's to kick in. What a waste!!! and how stupid (eg. always driving around with turbo lag)
highlandsun - Tuesday, September 6, 2005 - link
I suppose; for noobs it would just mean they can spread viruses/worms twice as fast.The turbo analogy is good, you've hit another topic near and dear to my heart. :P
(Of course with a properly sized turbo, "lag" is mostly a non-issue.)
Chapbass - Sunday, September 4, 2005 - link
lol, im assuming you mean dual core...your topic said dual but the post said single... : )As of right now the big push for dual core is for those (like myself) that multitask alot. Yeah, its not a good idea to multitask while gaming and such, but it would be nice to be able to do whatever while im say, burning a dvd with dvdshrink. Or i suppose (depending on how powerful they are, as i dont own one) I could actually probably game while burning a dvd... who knows maybe someone with a dual core can chime in.
I'm sure that once the big name companies start pushing it then software companies will start coding for it.
Bona Fide - Sunday, September 4, 2005 - link
Well, a lot of workstation applications already are SMP/SMT-aware, since most workstations are built off of Xeon or Opteron setups, which are usually dual or quad processor. Apps like Photoshop, CAD, and 3dStudioMAX all are optimized for dual-core. Here's a complete listing.http://www.denniskarlsson.com/smp/">http://www.denniskarlsson.com/smp/
Does anyone think that AMD's dual-core offering will match up with Intel's any time soon, in terms of price?
Athlon 64 X2 3800+ = $350
Pentium D 820 2.8 = $250
Athlon 64 X2 4200+ = $475
Pentium D 830 3.0 = $350
Athlon 64 X2 4600+ = $690
Pentium D 840 3.4 = $540
Furen - Sunday, September 4, 2005 - link
AMD's 3800+ is a match for the D 830 performance-wise, so the chart should be shifted like this:Pentium D 820 2.8 = $250
Pentium D 830 3.0 = $350 Athlon 64 X2 3800+ = $350
Pentium D 840 3.4 = $540 Athlon 64 X2 4200+ = $475
Athlon 64 X2 4600+ = $690
The problem with AMD (right now) is that they dont have a low speedgrade, and I'm sure this is a purely yield-related Issue. Intel can make cheap because they have much more capacity than AMD (something fab36 will hopefully help AMD deal with) and because they can just slice off a core if it turns out to be defective, effectively making their failed Dual-cores into single-core variants (this is because event hough the 2 cores are on the same die, they are simply side-by-side for the most part).
JarredWalton - Monday, September 5, 2005 - link
Actually, you still don't have the price table correct.1.8 GHz 3000+ ~= Pentium 3.0 GHz
2.0 GHz 3200+ ~= Pentium 3.2 GHz
2.2 GHz 3500+ ~= Pentium 3.4 GHz
2.2 GHz 1MB 3700+ ~= Pentium 3.6 GHz
2.4 GHz 3800+ ~= Pentium 3.8 GHz
(Yes, I AM fudging on what each CPU "equals". The fact of the matter is that in most instances, the 3000+ is actually faster than a 3.0 GHz P4, and that applies to the others as well.)
So, 2x2.0GHz 3800+ is really equal to 2x3.2GHz Pentium D 840. There really is not Intel match for the 4200+ and above parts - the 4200+ is faster in almost every case than even the 840D Extreme Edition. (Tests under Windows XP with four CPU-intensive tasks will put the 840 EE ahead, but that's a Windows scheduler problem.)
joex444 - Tuesday, September 6, 2005 - link
No, the 2x2.0GHz 3800+ is closest to an 830 (2x3.0GHz). The 840 is faster in most benchmarks.It's a little difficult to compare the two, really. The P4 was always faster than the Athlon (64) in certain apps, mainly encoding, while the Athlon was always faster in other things (like everything else, especially gaming). When you put two cores on one PCB, however, things can change. The Athlon X2's design is more advanced than Intel's. AMD really thought it out and made it backwards compatible with Socket 939, at the cost of memory bandwidth rather than invent a new socket. Intel instead requires a new motherboard that offers no real new features other than Pentium D support.
Basically, if you already have a Socket 939 board, it would make more sense to grab an X2 as an upgrade rather than jump ship and go Pentium D. You have to add the cost of a motherboard, something that everyone buying a Pentium D needs, so Intel can offer the chip at a lower cost somewhat, because they are also selling a Northbridge/Southbridge (whatever they decided to call it, that's what they really are). They also put next to no tech design into the Pentium D, little research money comparitivley. They literally glued two cores together and changed the northbridge/BIOS to support the chips. The architecture map shows it, the two cores go through the FSB to talk to each other, just like in a dual Xeon system which has two independent sockets. AMD's cores don't hit the FSB to talk, which can help out in many situations. Unfortunately for AMD, Intel's 800MHz FSB is still fast enough to run two 3.0GHz chips fast enough to compete. If Intel only had DDR FSB (effectively a 400MHz bus) I don't think they could have come out with the Pentium D at the speeds they did. Fortunately for Intel, in their quest to produce the 3.8GHz chips, some fall short, way short, like 2.8 or 3.0, and when you pair them together, glue them together (we saw the IDF pic of a chip w/o it's heat speader) and call it a Pentium D, it works really well for them.
If you had a P4 now, though, it probably makes more sense to get a Pentium D, unless you were looking for leading edge, in which case the 4600+/4800+ (if you just have to have 1MB L2 for each core) is your target.
IDK where I'm going here, so...yea.