NEWS AM2 Info

Here is an interview with an AMD guy talking about the upcoming DDR2 platform.

Quote:

---

TG Daily: AMD recently has been very tight-lipped on its plans for 2006. What exactly will ensure that you can keep your current pace, given the fact that your competition appears to be getting up to speed?
Allen: First, we have the transition to DDR2 in 2006, which presents new opportunities for OEMs and large Opteron platforms. In addition, the blade market continues to grow and this segment has selected Opteron because of its performance-per-watt advantage. Those are two factors that are driving the deepening commitment to Opteron in 2006.

---

Anyone notice the key word Opteron

Forget about nn_step.

No matter what you say or what prove you have. You can even have Hector Ruiz send him an email on the matter telling him it's going to be DDR2, and he would still find excuses not to believe him.

Just ignore his comments on the matter.

Socket F - FB-DImms with DDR2
Sound familiar to anyone else?

Quote:

---

Originally Posted by nn_step

Anyone notice the key word Opteron

---

yea that AMD guy is talking about servers, not desktops.

Actually you gotta remember, AMD probably waited for DDR2-667 as it has to use the worse case scenario, and 533 wasn't enough of an increase to change the entire platform.

AMD has to support the JEDEC timings for DDR2-667 which are 5-5-5-15, remember JEDEC for DDR400 is 3-4-4-8, so stop comparing it to the DDR400 2-2-2-5. If you want to compare to that DDR2-667 is availalble with 3-2-2-8 as well.

Socket AM2, will be DDR2 enabled, and will launch in H1 2006 from what I have read, it would be suicide for AMD to offer AM2 with DDR3 now as DDR3 is far from readily available.

From my guess this is what will happen AMD will shift to AM2 in 2006, with 90nm processor debuing on there across the entire spectrum, Athlon 64x2, Athlno 64, Sempron 64.

Then at the end of the year if the 65nm process is ready we will see Athlno 64x2 shift to 65nm, but probably at the lower clockspeeds first as AMD seems to be playing it safe. It would follow their current mantra, of take your time, do it right.

Conroe is launching in Early H2 2006 from what I see.

The thread was a good read until all the arguments started. Just let him have his own opinion and ignore him please. Everyone is entitled to believe what they want...now go back to talking about interesting stuff so I can learn :).

Actually Kobalt if you need help learning the theory I can help

Considering it's plastered all over AMD's website that they are using DDR2, I don't know what this arguement is about. Might as well pull out the tinfoil hat's at this rate...

The arguement is over which socket is coming out first and is going to use DDR2
They say M2
I say Socket F

Quote:

---

Originally Posted by nn_step

Actually Kobalt if you need help learning the theory I can help

---

There's a "theory"? I'm just learning about what to expect in the coming months so I can time my upgrade properly.

Quote:

---

Originally Posted by Kobalt

There's a "theory"? I'm just learning about what to expect in the coming months so I can time my upgrade properly.

---

Yeah theory behind Processor design
What you thought it was a simple as they say it is?

IMHO AM2 will be massacred(by Conroe) if, nothing changes beside the mem controller, and higher clock.

Yep...once Intel gets the link between the L2's (AMD has it between the L1's AFAIK) AMD is going to have a tough time competing. I hope AMD has something up its sleeve...

Meow....you haven't heard of the quad core K9 ? (I'm not kidding ;) )

Though Conroe should level the playing field, no doubt about it.

Perkam

Quote:

---

Originally Posted by perkam

Meow....you haven't heard of the quad core K9 ? (I'm not kidding ;) )

Though Conroe should level the playing field, no doubt about it.

Perkam

---

Aren't they suppose to come out in 2007?

Yup...don't wanna derail the thread though. AMD will need to have mem controllers that can significantly improve ddr2 results over Intel...maybe cas 2.5 ddr2 ?? I dont know....but 90nm on ddr2 has been done by Intel...which has now itself moved onto 65nm...

Perkam

Quote:

---

Originally Posted by perkam

Yup...don't wanna derail the thread though. AMD will need to have mem controllers that can significantly improve ddr2 results over Intel...maybe cas 2.5 ddr2 ?? I dont know....but 90nm on ddr2 has been done by Intel...which has now itself moved onto 65nm...

Perkam

---

ddr2 @ 2-2-2-5 would be hilarious for the launch :)

@400mhz that is :fact: :woot:

i think it wont be K9 they are going to call it k10

I thought k10 would be more of an architectural change, what would have been the k9 was scrapped, and dual k8 cores took its place. Dual cores are kind of k8 mark II, and quad cores will be k8 Mark III, then we'll be moving straight onto the next generation that was once named k10. Whether multicore k8 can be called k9, or whether what was called k10 becomes k9 I'll leave up to you....

DDR2-667 Will not be the limit. Like our current DDR400 systems. They can't only take DDR400. Socket 754's 939's and 940's can take any DDR made, DDR600 and so on. Same with socket M2, S1, and F. They will beable to take DDR2 800 or whatever. Even tho the specs don't say they can. Thats really not true. There is no limit as long as its DDR2 it will take any speed just like DDR. *Shrugs*Yet, for a good wile, the current sockets will be better then the new ones. Mostly the OCed ones that are generations ahead of anything they will make in stock to try to replace 939's and older sockets.

At the level in technology we are at now. You will hardly see a difference ever anymore for a good wile. Things are just way too advanced to keep up with. It will not be worth a upgrade maybe when M2's are atleast 2 years old. But now its better to wait a year for 939 stuff to get real cheap then get a FX-60 for 300 bucks then be happy for the next 5 years with a OCed chip at 3.5ghz with dual cores. You can't get anything better then that until what the FX-70. =P When you reach 3.5ghz on a amd. Your way ahead for awile.

In most cases DDR is much better then DDR2. For now if they can't get the timmings with DDR2 together. Its like a downgrade. Bandwidths are crazy enough as it is. With current systems and even more so OC'ed. Your not going to see memory bandwidth at 10,000mb/s anytime soon at stock. How big is the biggest system bandwidth OC... like 60Gb/s 100Gb/s with current systems. 3000MT/s 3500MT/s hyper transport? Still better then stock M2's at 2500MT/s hyper transport.

The point I'm trying to make also is that the newer sockets arn't all they are cracked upto be. They may use newer crap, but its worthless. When they both will OC the same and with what... a 10% diff... is it really worth a upgrade. When the most advanced stuff on 939's or even my 754 can compete with it that much. I mean hell if I can get a nice htt speed from the cpu then damn I could blow alot of cpus away on 939's and M2's even. Just like a 939 can. Only 5% diff with 754's and 939's. They would perform the same if it wasn't for there better cpu line and dual cores. Thats the only 2 things that set 754's apart from 939's a dual channel memory controler and the cpu line. Even tho they OC the same if not much off at all from the FX's. Limit of FX's is 4ghz. Moble 3700+ or 4000+ can get upto 3.5ghz. Only a 500mhz diff. Couldn't tell anyways. And that load of bull with the hyper transport being 400mhz slower in 754's. I OCed mine hyper transport to allmost 3ghz MT/s. Made my mobo something better then it should be even M2's. At stock anyways. Been running my OC like this since like september.

No limits yet. I can't wait to get a moble 4000+ Then really kick some ass. I want to see how high I can make the hyper transport stable. I reached 2800MT/s stable, I mean very stable. 24/7 and I hope to get to 3ghz with a venice 3000+ at the end of the week when I upgrade. For the next year or 2 I'll stay with this system upgrading to a 4000+ Newark when its alot cheaper. Then go 939 in around 2007 maybe if I feel the need for dual cores. I'll just go ahead and get a fx60. The best for the 939 will ever get.

Maybe what the 4000+ will be enough. We will always feel the need to upgrade, to think we have the best. When really thats further from the truth. Minor advances when you think about it, really won't show enough of a difference thats even worth the change. Maybe 5 years from now sure if you can afford the exteras. But for all you know your current system could verywell out perform the so called better sockets. When 3 cores or quad cores come out. Then I'll be intrested alot. But for now, the tech of even dual cores are useless as 64 bit until everything is writen for it. As why I got 64 bit windows. But I can't even really use it. No damn drivers for my cable modem out for it yet. So I play games on it is all for the extera power. Yet not very many games are 64 bit yet and not even close to multi core supported yet. I don't really multi task so having a 2nd core is useless to me right now.

I only prepare enough ahead to be ahead of the game. Some just feel they need to go overboard. Goes out and buys the newest thing feeling like they have to have it and really beleaves they are ahead by that much. When they really arn't just like the ppl with 939's some really beleave they have a better system then somebody with a 754. They can't prove it, cus they are really pritty much the same except for the pin count and memory controlers. The construction is the same.

For example my 754 board was just made literally in september of 2005 brand new. With the newest chipset and such. That can really kick ass, they sure made sure these systems would go out with a bang thats for sure. The performance I just get out of a sempron really is surprising to me. Yet I want to hit that 3ghz mark with a real athlon 64. A sempron has limits abit too much for my liking. But most on 754's with kick ass cpus can really bench some serious numbers. My cpu ranks about as good as a FX-55 despite my speed and my lack of L2. Doesn't matter when you have the htt of the cpu at 350.

Now M2's will they say run at about 300htt. Well then that means at normal speed of a say athlon 64 4000+ for a 939 at 200htt would be slower then a athlon of the same model and speed but at 300htt the data rate is about 50% better. Atleast I have a change to beat that by 50htt with my sempron cpu wise. And others can get better. Like best I seen was a cpu at 430htt or fsb whatever you want to start or end calling it.

M2's will OC about the same but not much. They are only another technology that is slightly better then before. Not really worth the upgrade until years after launch. Maybe not because of the limits of silicon. Speeds are close to the limits. For all we know in 5 years we will be into diminds by then. Every 2 years cpu speed double. At around 2010 we will be at the limit of silicon about 10ghz AMD wise mesures the speed correcting. For intels it will prob be about 20ghz... and watch AMD's still kick intels ass because my sempron rates faster then a 3.8ghz intel. Thats almost a 50% diff. Intel really is sad and needs to get it together. The got a advantage with those pentium M's. Being able to rate almost at true AMD speeds. But they don't apply the same arcitecture into there faster cpus. Its really sad my 2.4ghz OC can kick the ass of a 3.8ghz cpu siposidly anyways. I think intel lies about speed. Why AMD's can beat a faster clocking cpu is beyond some ppl. AMD's are made too damn good, they always have something up there sleeve and arn't about to let intel take the lead. They will always be 2nd best. There fsb's are horrible 1066 compared to my 2800fsb. Come on intel get it together. As well as so many other factors clock speed doesn't matter. A X2 4800+ still benches better then the newest 65nm intels EE cpus or pentium 5's as some ppl call them. The FX60's really blow the new intels EE's out of the water period. And dun a good 1ghz lower then them. Rate at about a intel going with dual cores at 4ghz. Amd's really get going when they reach 3.5 like a FX60 dual cores. Rates like a intel EE OCed at around 5ghz.

AMD's arn't laguing behind. They are so ahead they don't need to worry, they just like to keep intel on there toes. Think about why eachtime intel gets out a cpu, AMD amedeitly launches one better. Hell they prob aready have FX70's in proto-types just in case just to blow intel out of the water every time. =P

Intels L2 and L3 really doesn't seem to matter or effect performance agenst AMD's. Superior L2 AMD's have most likly once again becaue of better construction. AMD got this far and can still bench better with just 1mb in each core wile intels have 2mb's of L2 in each core. AMD still seems to be doing everything right, and holding alot back thats for sure. Mostly cus notice every CPU OC's too damn good most of the time with lower models getting 1ghz and 1.5ghz OC's even 100% OC's. Some ppl got 100% OC's on my model. >_> Damn cpu has a bad memory controler.

Most of the points are that AMD is just better and always will be in every way despite 90nm tech vs 65nm from intel. Prob is 90mn has been researched by AMD more then intel did, they just rushed it. Like with 65nm with crappy construction. When AMD brings it out it will be noticably better. Just like there 90nm tech can compete still with the intels 65nm technology. Its apperent what is better and what is not. :l Damn I like to talk too much. :l Damn damn need more jigawatts. :l I'll just shut up with my load of crab now... but you know its true. :l If you can read all that your crazy and have All the time in the world. :l Me... I don't do anything on sunday. =P :l :l :l :l

serge84, long post but little content there. Whilst i can agree with the sentiment that socket am2 is unlikely to be noticably better than 939 until at least the 65nm shift, the reasons you give are generally flawed and you cant compare the effects of cache on intel and amd architectures, the main benefit of the cache on intels part is to offset the lack of i/o and memory bandwidth which AMd's hypertransport and integrated memory controller mostly solves. Adding L3 cache to opterons is liable to only help in a small number of specialised tasks, as the IO i already pretty good. Intels 90nm process was not a failure, its a very good (and cheap) process, thats given us dothan as well as prescott. Prescott was a bad design for the process however, and attacking intel is rather pointless as the competition between the two companies drives the improvements. At the minute, intel have the better process technology, the better internal architecture (in yonah and soon enough conroe) and more inovations in the pipeline. Amd however have better chip IO and scalability due to HT and IMC. Socket AM2 will further increase bandwidth (which isn't needed) at a negligable hit to latency - but more importantly allow the cheaper and higher capacity ddr2 to be used.

you cant compare a OCed plataform with something stock and that doesnt exist right now is SO USELESS.

what you need to do is wait and see you dont know how they will OC and how the 65nm parts are going to OC in fact you dont know nothing unless you work for AMD.

if you see now 754 and 939 has a hughe difference not just 10% as a plataform 754 is old and useless same will happen with 939 with time it will be old and useless.

I want them to be neck and Neck for two reasons
1) Causes Extreme improvements
2) Keeps everything Cheap

They both have their advantages and disadvantages
When I Game I go AMD but when I play with Music editing I always grab the fastest core around..
To say one is superior to the Other in General is a HUGE LIE
Learn to accept it

nn good point, just look at ati an Nvidia, seems like graphics stood still when ati was on top with the 9800, now that nvidia is brining something good to the table, we have seen a product release and a refresh all in about two months (x1800->x1900)

If intel beats amd with conroe, amd will be forced to bring us something amazing, or intel will hit the gas and bury them. if amd is still ahead when conroe is released, they will just keep the same process of not revealing anything new or innovative, with maybe one or two high end chips per year.

on the other hand, if amd gets spanked this year, we will see intel just hanging out and not giving us anymore speed.

frankly i am tired of the AMD lineup, we get maybe a 200mhz speedbump every 6 months but that doesnt really increase the ceiling for us overlockers much anyways..

It time for amd to show me something new. and frankly i expect to be wowed with M2. If its just a jump to 2.8ghz dual cores and ddr2-667 ill be dissapointed.

There are two things that i want from amd, and they both start with 3. I want a 3.0ghz DC and a 3.2 ghz single core.

Give it to me AMD. :)

The only thing coming down the pike of interest is a feature; virtualization. All AMD plans to do performance wise is switch to DDR2, which will possibly hurt performance(bad latency). The first m2 chips will be just what's out now but with the m2 interface + features(there's also DRM crap but that's just a screw to the consumer, not a feature). AMD's move to 65nm supposedly third or fourth quarter. That's purportedly their roadmap. As bad as that is, they're also usually late(so.. 2007?). I can only hope Intel gets up to speed to force AMD to do more, provided they're able to respond. Oh, I've read one account of 65nm coming out second quarter(that'd be nice, but AMD being AMD, optimistic).

The tweaked 90nm process might be a boon for am2 as well if the rumours about it prove to be accurate.

socket-f.... DDR2...intergrated pci-e...LGA... opty 2xx,4xx,8xx

nnstep..pls...

http://www.xtremesystems.org/forums/...1&d=1138046013
http://www.xtremesystems.org/forums/...1&d=1138046013
http://www.xtremesystems.org/forums/...1&d=1138046013

DDR2 FB-DImms
No arguement on Socket F being DDR2

uh yeha, so where's the DDR3 gonna go?you realize that FB dimm means that the memory controller is on the DIMM itself, with less interconnects between the chipset and the memory, which, although it makes motherboard design simpler(69 pins for DDR2, in comparison to the 268 currently used), it also add latency to the current K8 tech as the memory control is not really on-die anymore, as well as allowing manufacturer's to use 2-layers of pcb interconnects over 3 for a single channel, and could thereby introduce greater crosstalk as OEM's try to make the boards as cheap as possible. of course, they have implemented the same sort of routing as netwrok cables to deal wit hthis problem, with the data being inverted and passed along another path to ensure data integrity,and we all know the problems that can arise from such situations.


Now, ok, FBdimms can handle DDR3 as well as DDR2, but the only real advantages are greater capacity, brought about by the fact that you can easily route more channels in the same space, which is also what brings the greater bandwidth. Because the interconnection is serial, like pci-e, you can see why they chose fbdimms to go with intergated pci-e control to the cpu, but you don't seem to realize that this tech is only really benficial in real-world work in a highly threaded environment, and not the computing environments that almost every user is used to now...in other words, it's only practical for servers that need large ram capacity, spread accros multiple channels.

the only thing new about FBdimms is that they break up the communication between the controller and module from travelling the same paths(which really isn't new at all), as current DDR tech uses, and routes each signal on a different trace. this means it might be possible to squeeze more info on that trace, but with the greater capacity adding simply by adding additional channels, and not higher-bandwidth connections, there is no real advantage to FBdimms over standard dimms, at least to us.


Maybe when new OS'es pring out that are multithreaded, but i definately isn't gonna help a windows machine.

Bingo.. You hit it on the head..
EVERY SINGLE major analyst says Thread Level parrelization is the Future of Processing..
ANd look FB-Dimms are made Specifically for that... Hmm isn't that surprising...
FB-Dimms bring a Brand new level to the bandwidth equation..
Right now there isn't a Memory Setup that would provide enough Bandwidth in a cost effective manner for an Eight Core system...FB-Dimm make that possible..
First it will go to the servers and then it will change and reach the rest of us