I duno what is up with the 400 counterclaim. many websites are discussing this as a real find, including the inquirer, anandtech and the register. they all hate toms too yet dont see it particularly flawed.
Ocworkbench.com is also considering it to be real, logic says that it must be so. AMD kept say the past year that they need not worry about performance compared to intels chips. They probably already have tested the intel chips and found they suck. COuldnt it be that AMD is just wanting more money? Gearing up for a cheaper OEM platform, a thing where the socket will remain static, the ram is cheaper, and wattages of chips form a larger menu to choose from? The 65 nm part of AMD is going to be on AM2, im guessing they wanted to make the socket they always wanted before doing so. I think it is all about cost, nothing about performance.
the other review you were talking about is on anandtech. they were given a crap cpu with a crap mem controller. Anand didnt feel it was appropriate to review it. If you notice in tomshardware, it says they have a fixed mem controller cpu, so they addressed those previous reviews.
that was discussed for the past 2 weeks.
I would like it nothing more in the world than for AM2 to absolutely do nothing for anyone, and for conroe/merom to be slower than a 3800 a64 single core for games. Complicated why I feel this way, but I would be satisfied. Would finally show how not for the user all of these companies are.
Some more info about AM2 processors
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
One:
AMD got no increase in bandwidth going from DDR-400 to DDR2-667. How can that be?
Three:
http://www.theinquirer.net/?article=29776
From the review:
Two:2x 512 MB DDR2-667 (333 MHz, CL3-2-2-8, 1T)
Running at CL-4-4-4-12 Timings - not changeable in test motherboard BIOS
AMD got no increase in bandwidth going from DDR-400 to DDR2-667. How can that be?
Three:
http://www.theinquirer.net/?article=29776
From the review:
As already mentioned, the memory is suspected to suffer from a performance bug, which may be true or simple a matter of the early product stage. The upcoming processor (revision F) will fix any of this
What a perverse attitude. Surely you want computer parts manufacturers (as a potential consumer of said parts) to produce better, more powerful products? That's what progress is all about, sure Intel have an unjustified dominance in the consumer market (surprised antitrust lawsuits have not been brought against them) but to actively want them to exploit this duopoly is bizarre.I would like it nothing more in the world than for AM2 to absolutely do nothing for anyone, and for conroe/merom to be slower than a 3800 a64 single core for games. Complicated why I feel this way, but I would be satisfied. Would finally show how not for the user all of these companies are.
-
~El~Jefe~
- Friend of SPCR
- Posts: 2887
- Joined: Mon Feb 28, 2005 4:21 pm
- Location: New York City zzzz
- Contact:
they are exploiting this currently, all of them. that is how you make money hand over fist, by exploitation.
pepsi coke amd intel
new concept?
First time I hope it is made so obvious that it is rather repulsive to consumers.
a 10% increase in speed is not progress. besides fora 10% increase in speed. Isnt it rather obvious that computing power is sluggishly moving along while other miniscule companies and engineering firms make massive leaps into the same technologies?
I love my dual core, even as it sits in its box. I consume it. however, I know my crack is cut with 80% crap.
pepsi coke amd intel
new concept?
First time I hope it is made so obvious that it is rather repulsive to consumers.
a 10% increase in speed is not progress. besides fora 10% increase in speed. Isnt it rather obvious that computing power is sluggishly moving along while other miniscule companies and engineering firms make massive leaps into the same technologies?
I love my dual core, even as it sits in its box. I consume it. however, I know my crack is cut with 80% crap.
-
BillyBuerger
- Patron of SPCR
- Posts: 857
- Joined: Fri Dec 27, 2002 1:49 pm
- Location: Somerset, WI - USA
- Contact:
Something seems fishy about their wattage claims. A X2 3800+ uses more power than the X2 4800+??? I have a feeling that they just grabbed different systems using different motherboards and possibly even different graphics cards and threw them in there. They don't really say what parts were used in each system.~El~Jefe~ wrote:click me but dont get burned
Look at this.
this is a major die shrink. 1/3 less just about.
check out the wattages. OUCH! dont let your kids near that. the wattages went down a small amount. they really didnt get much help. even tom's said the wattage pre-claim didnt hold up any.
-
~El~Jefe~
- Friend of SPCR
- Posts: 2887
- Joined: Mon Feb 28, 2005 4:21 pm
- Location: New York City zzzz
- Contact:
hahah nice.
yeah i bet they were using different motherboards. I hope so! i would have to click around and read more about it, then read the forum entry commenting on it, but well, I am not goign to be bothered.
What intriques me is the new wattage choice for the am2 processors. I assume as well this is the most intriguing thing for the majority of us in this forum. However, here is a potentially nullifying question/problem:
My Questions and ideas:
Turions. they have some lower wattage claims. tdp's right? ok that's one concept.
Voltage. NOt really discussed in claims for tdp's. They dont say how it gets that way. a spd undervolt from normal shouldnt be considered in a lower wattage claim.
What new technology is there for the AM2 to get 2 different wattage claims? Undervoltabilty? Something magical? new transistors? I saw mention of less leakage and such, I guess this could be the deal.
Why would they then make a crappier version of the chip and a good version? they are starting out fresh for goodness sake, it is not like they are going through revisions at launch.
Could it just be that some are tested to work at lower voltages? Kinda strange how no mention of design or technology has seeped out at all with the first functional chip being reviewed. NDA's I guess, but inquirer and such dont care about them.
I will have to get out my tinfoil hat.
yeah i bet they were using different motherboards. I hope so! i would have to click around and read more about it, then read the forum entry commenting on it, but well, I am not goign to be bothered.
What intriques me is the new wattage choice for the am2 processors. I assume as well this is the most intriguing thing for the majority of us in this forum. However, here is a potentially nullifying question/problem:
My Questions and ideas:
Turions. they have some lower wattage claims. tdp's right? ok that's one concept.
Voltage. NOt really discussed in claims for tdp's. They dont say how it gets that way. a spd undervolt from normal shouldnt be considered in a lower wattage claim.
What new technology is there for the AM2 to get 2 different wattage claims? Undervoltabilty? Something magical? new transistors? I saw mention of less leakage and such, I guess this could be the deal.
Why would they then make a crappier version of the chip and a good version? they are starting out fresh for goodness sake, it is not like they are going through revisions at launch.
Could it just be that some are tested to work at lower voltages? Kinda strange how no mention of design or technology has seeped out at all with the first functional chip being reviewed. NDA's I guess, but inquirer and such dont care about them.
I will have to get out my tinfoil hat.
-
defaultluser
- Posts: 82
- Joined: Mon Feb 20, 2006 9:39 pm
You can't go wrong with DDR2
DDR2 on AM2 will perform better by the time the platform is ready for retail release. The processor core isn't completely debugged yet, and the motherboard support is almost non-existant.
Even if the performance improvement is not incredible, remember what AMD is getting out of DDR2:
1. Higher densities at a lower price. EVERYONE is moving to DDR2.
2. Lower power consumption by a sizeable amount. DDR2 has a much lower voltage (1.8v versus 2.5v), and those of you who undervolt know the value of dropping the voltage by such a large chunk. The power used by CMOS is roughly proportional to the square of the voltage * the frequency.
Another power consumption benefit: DDR1 uses memory "cells" with two buffers each, to deliver that dual data rate. DDR2 uses cells with four buffers, and thus gets twixe the thoroughput as DDR1 using the exact same internal frequency.
Thus, when you ignore the external control circuitry, DDR2 uses half the power of DDR1 at the same data rate, all for slightly increased latency. This is why the dual-channel DDR-667 chipset used by Yonah is kicking the pants off the Turion with DDR-400, even though the Turion (MT) processor actually uses the same or less power than Yonah. AMD knows they have to move to DDR2 if they hope to have the Turion taken seriously in the notebook market.
Even if the performance improvement is not incredible, remember what AMD is getting out of DDR2:
1. Higher densities at a lower price. EVERYONE is moving to DDR2.
2. Lower power consumption by a sizeable amount. DDR2 has a much lower voltage (1.8v versus 2.5v), and those of you who undervolt know the value of dropping the voltage by such a large chunk. The power used by CMOS is roughly proportional to the square of the voltage * the frequency.
Another power consumption benefit: DDR1 uses memory "cells" with two buffers each, to deliver that dual data rate. DDR2 uses cells with four buffers, and thus gets twixe the thoroughput as DDR1 using the exact same internal frequency.
Thus, when you ignore the external control circuitry, DDR2 uses half the power of DDR1 at the same data rate, all for slightly increased latency. This is why the dual-channel DDR-667 chipset used by Yonah is kicking the pants off the Turion with DDR-400, even though the Turion (MT) processor actually uses the same or less power than Yonah. AMD knows they have to move to DDR2 if they hope to have the Turion taken seriously in the notebook market.
-
Devonavar
- SPCR Reviewer
- Posts: 1850
- Joined: Sun Sep 21, 2003 11:23 am
- Location: Vancouver, BC, Canada
@defaultluser:
Even DDR doesn't eat up much power. The difference between one and two sticks of DDR is ~3W; for DDR2 the difference is ~2W. Yes, DDR2 does seem to be less power hungry, but RAM doesn't suck up much power anyway.
And, why are you bringing up the Turion? M2 supports AMD's desktop chips; the Turion will have a different socket.
Even DDR doesn't eat up much power. The difference between one and two sticks of DDR is ~3W; for DDR2 the difference is ~2W. Yes, DDR2 does seem to be less power hungry, but RAM doesn't suck up much power anyway.
And, why are you bringing up the Turion? M2 supports AMD's desktop chips; the Turion will have a different socket.
-
~El~Jefe~
- Friend of SPCR
- Posts: 2887
- Joined: Mon Feb 28, 2005 4:21 pm
- Location: New York City zzzz
- Contact:
yes the 35 vs 65watt, the 89/110 guys I got.rpsgc wrote:I assume you're talking about the 35 and 65W CPUs right? Those are "Energy efficient" processors and are not the "mainstream" processors.
If you are talking about the reduction of TDP from 110 to 89W then that's explained on this thread.
anyone know what that means? I know they are not "mainstream". That used to mean they are the same thing just tested better. If that's the case, mildly useless, if they are built different, I imagine a slightly higher price and a boon to spcr people. This definitely would seal up the coffin for intel. too bad too.
VIA~El~Jefe~ wrote:they have less than 80% currently because of amd and that other company i always forget, SIS? or something.
I think they are meant for the HTPC market (under the Live! brand)~El~Jefe~ wrote:yes the 35 vs 65watt, the 89/110 guys I got.
anyone know what that means? I know they are not "mainstream". That used to mean they are the same thing just tested better. If that's the case, mildly useless, if they are built different, I imagine a slightly higher price and a boon to spcr people.
-
~El~Jefe~
- Friend of SPCR
- Posts: 2887
- Joined: Mon Feb 28, 2005 4:21 pm
- Location: New York City zzzz
- Contact:
HEy! i just found out that quad cores will not be made for socket AM2!
YEAH!
so useful that socket will be!
er
yeah
i mean that makes it absolutely worthless as an upgrade from 939.
even from 754. whoa, i know big statement. Well here: you can move your ram and vid card to an asrock dualsataII 939 and just get a new chip. or throw it all out and get am2.
jeez. im glad i did what i did this week and bit the bullet for a 4200
YEAH!
so useful that socket will be!
er
yeah
i mean that makes it absolutely worthless as an upgrade from 939.
even from 754. whoa, i know big statement. Well here: you can move your ram and vid card to an asrock dualsataII 939 and just get a new chip. or throw it all out and get am2.
jeez. im glad i did what i did this week and bit the bullet for a 4200
In principle, AMD quad-cores could run on socket AM2. But the first quad-cores are destined to the server market (they will be very expensive at first), where they need the extra HT connections.
The problem with DDR2-667 over DDR1-400 is that the performance coming from the increased bandwidth is lost due to the increased latency.
And as the old single-channel Athlon64 showed, latency is more important than bandwidth, at least for games.
If most software would properly implement the prefetch instructions, then bandwidth would be most important.
Once AMD will increase the cache size, using the new technology they aquired (which works only with the SOI process), we'll benefit more from the extra bandwidth provided by DDR2.
The problem with DDR2-667 over DDR1-400 is that the performance coming from the increased bandwidth is lost due to the increased latency.
And as the old single-channel Athlon64 showed, latency is more important than bandwidth, at least for games.
If most software would properly implement the prefetch instructions, then bandwidth would be most important.
Once AMD will increase the cache size, using the new technology they aquired (which works only with the SOI process), we'll benefit more from the extra bandwidth provided by DDR2.
I certainly agree that progress is good. Fast progress is even better (even if it means that 12 months from now we'll see CPUs two times faster than an FX60 in real world application).That's what progress is all about, sure Intel have an unjustified dominance in the consumer market (surprised antitrust lawsuits have not been brought against them) but to actively want them to exploit this duopoly is bizarre.
Oh, about lawsuit, where have you been:
http://www.theregister.co.uk/2005/06/28/amd_suit_intel/
http://www.anandtech.com/weblog/default.aspx?bid=228
We won't see much action for a while though.
As much as I am an AMD fan, Intel is a saint next to Creative in terms of providing fair competition in this market market. True Intel have done things to block AMDs market growth. But at least it didn't go around buying off AMD, Cyrix, VIA etc. as soon as it found out that they are making competing chips (and I am pretty sure it had plenty of opportunities).
AMD limited growth can't be entirely be pointed towards Intel either. They have historically had more production capacity issues than Intel. This has improved, but it has hurted them.
I like this article, and mostly agree with it: http://www.firingsquad.com/features/intel_core_new_low/
Up till the P2/CeleronA period, I would buy no others than Intel. Okay, did look at AMD/Cyrix as well, but up to that point, I do believe that they earned their position. Now well, I've moved to the AMD camp for reasons stated in the article. Basically, whoever provides the best product for my needs in my price range gets my money. Sadly it also mean that in the mean time, Creative is getting my £££ (I consider Creative to be far more 'evil' than Intel).
Thanks for bringing that article to my attention. I mostly agree with it as well.I like this article, and mostly agree with it: http://www.firingsquad.com/features/intel_core_new_low/
What I find particularly interesting is that Intel's Chief Technology Officer predicted as early as 2000 that this continuous clock scaling with continuous heat increase was unsustainable, but did nothing about it.
They foresaw the disadvantages of the Prescott, but they made it anyway.
Only after the negative backlash from Prescott did they make power-efficiency and performance per watt their number 1 priority.
TomZ: How is that done? I've done some programming, and I don't know anything about using prefetch instructions.
The AMD Software Optimisation Guide for AMD64 (25112.pdf) has good info on prefetch instructions, in chapter 5.6 (or you could try the Intel dox, IA32 Architecture ... Instruction set reference, 25366718.pdf).
From my experience, once the algorithm is well established in C, I replace parts of it with inline ASM, including MMX / SSE. When I get no performance improvements, there's likely a memory latency limitation, so it's time to use some prefetch instructions. Those don't give you more bandwidth, but to some extent they hide the actual latency and your code makes better use of the existing bandwidth.
The AMD Software Optimisation Guide for AMD64 (25112.pdf) has good info on prefetch instructions, in chapter 5.6 (or you could try the Intel dox, IA32 Architecture ... Instruction set reference, 25366718.pdf).
From my experience, once the algorithm is well established in C, I replace parts of it with inline ASM, including MMX / SSE. When I get no performance improvements, there's likely a memory latency limitation, so it's time to use some prefetch instructions. Those don't give you more bandwidth, but to some extent they hide the actual latency and your code makes better use of the existing bandwidth.
I think the problem is that they realised it a little too late. The P4, designed especially to ramp up the hz (we all remember the up to 10Ghz claims), was already a completed project by 2000 (testers had their hand on it by the end of the year, although I imagine would've been completed before that).jaganath wrote:What I find particularly interesting is that Intel's Chief Technology Officer predicted as early as 2000 that this continuous clock scaling with continuous heat increase was unsustainable, but did nothing about it.
They foresaw the disadvantages of the Prescott, but they made it anyway.
They couldn't kill the product before launch, since they have at the time discarded the P3 and wouldn't have anything else to offer. In terms of pure performance, the P4 still had the edge over the earlier Centrino's, other I am pretty sure those were already more efficient as a whole. But then, I guess there is no real point in porting the Centrino technology (back) to the desktop when P4 still sells... pretty well I am sure. Once we overlook the enthusiast community, which is still rather small. I have friends, computing science student (but not computer building enthusiast) who swears by Intel, despite my attempt to sway them. Oh well, their choice.
AMD has been nimbling Intel's market share, but it's still just nimbling... Which I reckon is the reason they stuck with the P4 design for so long (I am sure they were looking for a certain amount of profit after covering the R&D cost).
Just my guesses.