Today's Phenom News

Status
Not open for further replies.
That was a good find, and it verifies socket compatibility. Thanks for posting the link man.
 
That was a good find, and it verifies socket compatibility. Thanks for posting the link man.

The more we know, the more we can help our fellow [H]. ;)
(besides, I'm fishing for a job posting the news... Phfbt! As if!)
 
I still poo-poo them for throwing out 939, but then again I have the bad luck of always purchasing new computer hardware right as standards change. First I bought 939/AGP because at the time there weren't any 939/PCI-E, second I got this 939 Opteron thinking that it would overclock (it doesn't) and that I could get an FX CPU that was faster than the king of FX at the time (FX60), but they then killed off 939.

Of course, mostly my bad timing. But why suffer in silence when I can complain?

:p:D:cool:
 
That was a good find, and it verifies socket compatibility. Thanks for posting the link man.

2800, 3200 and 3600MHz are you kidding? Those seem a little faster than the current HT 2000MHz. I'm just wondering how many will want to cripple those things on old assed AMD2's? Now if the old AM2's can match their speeds, support the newer Cool N Quiet and etc.. pretend I didn't complain;)?

Man, talking about speed binnng to death.
 
2800, 3200 and 3600MHz are you kidding? Those seem a little faster than the current HT 2000MHz. I'm just wonder how many will want to cripple those things on old assed AMD2's? Now if the old AM2's can match their speeds, support the newer Cool N Quiet and etc.. pretend I didn't complain;)?

Man, talking about speed binnng to death.

HT currently runs at 1000mhz.. so that is a HUGE increase.
 
Yep,. And I might be misunderstanding it, but that threefold+ bump up equates to a lot more bandwidth... more than the Intel's have on slate. So, if they OC decently as well (and at 65nm, same as current Core2Duo's, why shouldn't they?) then they may indeed hold their own. Also, aren't the Stars supposed to offer independent memory control, as opposed to the AM2's where it's on die? If that were true, wow! If.
 
I have an idea of the implications for heavy I/O and server use, but will the extra bandwidth from HT 3.0 have any impact on desktop performance? I reduced the HT multiplier to 3x on my Opty 170, and I haven't seen any lower numbers now at 900mhz than before when it was at 1200.
 
Those Quad-core Phenoms with the 3600MHz HT look like they might be an Intel Q6600 on steroids! Ummm, bandwidth.

Any monkey can throw around big numbers. K8 has higher memory bandwidth, more bus speed, and much higher clock speeds than Conroe, and look where that ended up. :p I will believe it when I see [H]ardOCP's numbers (which are going to include x264 encoding results, right Kyle? ;)).
 
Monkey? Monkey?! I, Sir, am an ape! And of the highest order, at that. So, I'll be patient with you and illuminate that which escaped you: I pointed out those numbers as to identify the processor I was referring to. So, if you still insist on implying others are simply monkeys, do tuck up your own vestigial appendage first.

Oh, and 'Hoot!', too.
 
2800, 3200 and 3600MHz are you kidding? Those seem a little faster than the current HT 2000MHz. I'm just wonder how many will want to cripple those things on old assed AMD2's? Now if the old AM2's can match their speeds, support the newer Cool N Quiet and etc.. pretend I didn't complain;)?

Man, talking about speed binnng to death.

Ummmm Not really.

That is considerably faster then what we use today. In addition it isnt gonna have much of an affect on the desktop. It was designed to help scale past 8 sockets to 32 sockets.

The benefit that AM2+ brings to the table is not only faster HT3, but also a much more fine grained power management. They overhauled CnQ for the new socket. It is my opinion that for an enthusiast, you wont see much of a difference between the two. But if your going to be buuying a new chip, you should also buy a new board...

For those that cant, then you can keep the board you have and it should work.
 
2800, 3200 and 3600MHz are you kidding? Those seem a little faster than the current HT 2000MHz. I'm just wonder how many will want to cripple those things on old assed AMD2's? Now if the old AM2's can match their speeds, support the newer Cool N Quiet and etc.. pretend I didn't complain;)?

Man, talking about speed binnng to death.

Show me a case where a performance difference can be noted between 2000MHz HTT and 400MHz HTT on the desktop.

AM2 won't cripple AM2+ chips for I/O bandwidth. Even synthetic benchmarks won't pick it up unless its merely measuring the raw throughput. Disk access, audio, PCI-E, etc will all have plenty of bandwidth.

I actually forgot about this article:
http://eclipseoc.com/index.php?id=6,47,0,0,1,0
Read up on it and learn something about hypertransport, pretty cool system bus.
 
2800, 3200 and 3600MHz are you kidding? Those seem a little faster than the current HT 2000MHz. I'm just wonder how many will want to cripple those things on old assed AMD2's? Now if the old AM2's can match their speeds, support the newer Cool N Quiet and etc.. pretend I didn't complain;)?

Man, talking about speed binnng to death.

I've said it before and I'll say it again... The increased HT means nothing, not even with 4 cores. This isn't Intel we're talking about, it is AMD. Their multi core design doesn't even use the HT link to communicate like Intel uses the FSB. If we were talking about 4 SOCKETS, then yes, it would utilize HT, but the native quad core architecture does not use HT as the interconnect. The only thing AM2 boards will cripple is the improved power saving features that AM2+ offers.
 
I really want to stay with AMD for the time being because I'm a big fanboy! :D
 
I absolutely love my X2 4600+ and would love for my next upgrade to be another AMD processor, I'm going to follow this and see what these new processors can really do.
 
K8 wasn't powerful enough to use the HT2 links up until 4 sockets, then you get your money's worth in scaling. HT3 will be doing 8 sockets and 4 cores per socket. So yeah big difference. However K10 is also 70% more powerful then K8 and will use the HT3 links totally different btw because whatever K8 was is 90% not there in K10 if you didn't know this isn't a reworked core but a totally diff desine all together. The only thing thats the same is the memory controller most likely because the rest of the cpu isn't even remotely similar to k8.

FSB is the bottleneck and 40 years opsolete unlike HTL that is double the speed and twice the bandwidth upto 24gb/s vs intels 9 to 12Gb FSB. No wonder K8 can't be stressed on a HTL loaded with extra bandwidth. AMD sure eleminated that bottleneck forever even at 400mhz HTL speeds. Good going AMD got Intel fallowing in your footsteps with CSI = HT1.

That AFX looks nice to me for AM2. Oh and like I said K10's HT3 speed is dependent on core speed of the clock. I proved that long ago. My clams still hold correct. huh I was right all along. I told you so. I can't wait to see when I'm right about K10 in its performance too all along. I can't wait to see how K10 OC's. Even if its only 3ghz to 3.5ghz, it will still be faster then a C2Q at 4.5ghz and 5ghz because AMD's 50% over a 3ghz C2Q agenst a 2.5ghz K10 alone.
 
K8 wasn't powerful enough to use the HT2 links up until 4 sockets, then you get your money's worth in scaling. HT3 will be doing 8 sockets and 4 cores per socket. So yeah big difference..

So unless someone is planning to build a PC with >4 sockets, AM2 should suit them just fine if they've already got an AM2 board.
 
So unless someone is planning to build a PC with >4 sockets, AM2 should suit them just fine if they've already got an AM2 board.

Exactly, however you'll be missing out on some really kick ass power saving features. If your building a HTPC, or an SFF, then you'll want to upgrade to an AM2+ board.
 
Show me a case where a performance difference can be noted between 2000MHz HTT and 400MHz HTT on the desktop.

AM2 won't cripple AM2+ chips for I/O bandwidth. Even synthetic benchmarks won't pick it up unless its merely measuring the raw throughput. Disk access, audio, PCI-E, etc will all have plenty of bandwidth.

I actually forgot about this article:
http://eclipseoc.com/index.php?id=6,47,0,0,1,0
Read up on it and learn something about hypertransport, pretty cool system bus.

Again, I pointed that Quad Core is the reason for the upgrade, not Dual Core. AM2 will Cripple Kuma to a lesser extent than Agena not if you can link me to Barkie, the point will be easily proven:)

The problem is not 2 socket or more. The problem is ONE Quad Core feeding off ONE memory controller. AMD current uses two cycles to do Intel's SSE the newer processors will do it in one cycle, it gain more Floating Point and Integer so it should or need to Take in more bandwidth data wise, process it and send more back. If AM2 is all that it needs, I say again, AMD is PHUCKED! This would be like saying all Kentsfield needs is i875 and 800MHz FSB and DDR1 PC3200 instead of 1333MHz FSB and DDR2 800 or 1066.

HTT runs at 1GHz and the newer ones will run at 1.4, 1.6, and 1.8GHz. I don't see this as marketing or hype and I'm thinking AMD is doing it for a damned good reason. Shame on you guys for bashing AMD, wait a minute, [sound=Steve Earkel]Did IIII say thatttt[/sound].:p Whenever it does ship, someone please save this thread?
 
Yep,. And I might be misunderstanding it, but that threefold+ bump up equates to a lot more bandwidth... more than the Intel's have on slate. So, if they OC decently as well (and at 65nm, same as current Core2Duo's, why shouldn't they?) then they may indeed hold their own. Also, aren't the Stars supposed to offer independent memory control, as opposed to the AM2's where it's on die? If that were true, wow! If.

Not really, they have different ways of ending up at similar points. Currently Intel has more bandwidth internal and AMD external. One of the reasons AMD increased this on the new processors. Then they had to do some work on Smart Cach and Smart Memory access too, just like Intel.
 
K8 wasn't powerful enough to use the HT2 links up until 4 sockets, then you get your money's worth in scaling. HT3 will be doing 8 sockets and 4 cores per socket. So yeah big difference. However K10 is also 70% more powerful then K8 and will use the HT3 links totally different btw because whatever K8 was is 90% not there in K10 if you didn't know this isn't a reworked core but a totally diff desine all together. The only thing thats the same is the memory controller most likely because the rest of the cpu isn't even remotely similar to k8.

FSB is the bottleneck and 40 years opsolete unlike HTL that is double the speed and twice the bandwidth upto 24gb/s vs intels 9 to 12Gb FSB. No wonder K8 can't be stressed on a HTL loaded with extra bandwidth. AMD sure eleminated that bottleneck forever even at 400mhz HTL speeds. Good going AMD got Intel fallowing in your footsteps with CSI = HT1.

That AFX looks nice to me for AM2. Oh and like I said K10's HT3 speed is dependent on core speed of the clock. I proved that long ago. My clams still hold correct. huh I was right all along. I told you so. I can't wait to see when I'm right about K10 in its performance too all along. I can't wait to see how K10 OC's. Even if its only 3ghz to 3.5ghz, it will still be faster then a C2Q at 4.5ghz and 5ghz because AMD's 50% over a 3ghz C2Q agenst a 2.5ghz K10 alone.

And you know all of those numbers how?
 
So unless someone is planning to build a PC with >4 sockets, AM2 should suit them just fine if they've already got an AM2 board.

He's wrong as hell as usual. If AMD didn't have to update the motherboard, IMHO they wouldn't have. It is a waste of money and time. AMD could be working on something else they desparately needed finished or tweaked. Again, Dual core might not be hurt as bad on AM2 but Quad core will.

Please note? I tried to expain to Duby and Serge why 4 X 4 wasn't going to be a Kentsfield killer LONG before the first tests showed up. Like Dan here at HardOCP, I love motherboards and know a Processor ain't shit without it.
 
He's wrong as hell as usual. If AMD didn't have to update the motherboard, IMHO they wouldn't have. It is a waste of money and time. AMD could be working on something else they desparately needed finished or tweaked. Again, Dual core might not be hurt as bad on AM2 but Quad core will.

they updated it for the extra power saving stuff. stuff that many people would benefit from. not us. ;)


The problem is not 2 socket or more. The problem is ONE Quad Core feeding off ONE memory controller.

yeah, but that has nothing to do with the HT link. there are interconnects within the cpu to handle that traffic :p
but that's what intel is doing now... 4 cores, one memory controller. and it has to go through the FSB first to get to that memory controller. i fail to see where the problem is with amd's implentation, nor how HT link has anything to do with it. i understand how much you love intel, but please, understand how the HT link works and what it's there for before trying to bash it
 
The problem is not 2 socket or more. The problem is ONE Quad Core feeding off ONE memory controller. AMD current uses two cycles to do Intel's SSE the newer processors will do it in one cycle, it gain more Floating Point and Integer so it should or need to Take in more bandwidth data wise, process it and send more back. If AM2 is all that it needs, I say again, AMD is PHUCKED! This would be like saying all Kentsfield needs is i875 and 800MHz FSB and DDR1 PC3200 instead of 1333MHz FSB and DDR2 800 or 1066.

Ummmm Not exactly.....

I'm not going to get into another FSB vs IMC debate again. Except to say That AMD's IMC has more bandwidth the Intels FSB.... The aggregate bandwidth between HT3 and IMC is --MUCH-- more bandwidth then Intels FSB.

In addition K10 will officially support a DDR2-1066 DIMMS. Which it will have full bandwidth to due to the IMC. Intel's FS wont even come close even at a 1600mhz FSB...
 
AMD current uses two cycles to do Intel's SSE the newer processors will do it in one cycle, it gain more Floating Point and Integer so it should or need to Take in more bandwidth data wise, process it and send more back.

but that's what intel is doing now... 4 cores, one memory controller. and it has to go through the FSB first to get to that memory controller.

FSB is the bottleneck and 40 years opsolete unlike HTL that is double the speed and twice the bandwidth upto 24gb/s vs intels 9 to 12Gb FSB.

The amount of bullshit on display in this thread is staggering :rolleyes:
 
1000MHz and 2000MHz effective=P

Again, I pointed that Quad Core is the reason for the upgrade, not Dual Core. AM2 will Cripple Kuma to a lesser extent than Agena not if you can link me to Barkie, the point will be easily proven:)

The problem is not 2 socket or more. The problem is ONE Quad Core feeding off ONE memory controller. AMD current uses two cycles to do Intel's SSE the newer processors will do it in one cycle, it gain more Floating Point and Integer so it should or need to Take in more bandwidth data wise, process it and send more back. If AM2 is all that it needs, I say again, AMD is PHUCKED! This would be like saying all Kentsfield needs is i875 and 800MHz FSB and DDR1 PC3200 instead of 1333MHz FSB and DDR2 800 or 1066.

HTT runs at 1GHz and the newer ones will run at 1.4, 1.6, and 1.8GHz. I don't see this as marketing or hype and I'm thinking AMD is doing it for a damned good reason. Shame on you guys for bashing AMD, wait a minute, [sound=Steve Earkel]Did IIII say thatttt[/sound].:p Whenever it does ship, someone please save this thread?

Not really, they have different ways of ending up at similar points. Currently Intel has more bandwidth internal and AMD external. One of the reasons AMD increased this on the new processors. Then they had to do some work on Smart Cach and Smart Memory access too, just like Intel.

And you know all of those numbers how?

He's wrong as hell as usual. If AMD didn't have to update the motherboard, IMHO they wouldn't have. It is a waste of money and time. AMD could be working on something else they desparately needed finished or tweaked. Again, Dual core might not be hurt as bad on AM2 but Quad core will.

Please note? I tried to expain to Duby and Serge why 4 X 4 wasn't going to be a Kentsfield killer LONG before the first tests showed up. Like Dan here at HardOCP, I love motherboards and know a Processor ain't shit without it.

Wow, 5 replies one after another. :rolleyes:
 
2800, 3200 and 3600MHz are you kidding? Those seem a little faster than the current HT 2000MHz.
I'm thinking AMD will do something like the 800MHz/1066MHz P4s Intel had a while ago. For example, depending on the socket type a 2.4GHz Kuma will initialize at either a 12x multipler (200MHz DDR HT x 5) on an AM2 socket or with a lower multipler in an AM2+ socket.

That is unless HT 3.0 just uses really high multipliers (i.e. 9 x for 3.6GHz effective speed). Then the CPU multiplier just stays at 12x in either case. I wish I could find the HT 3.0 base frequency.
 
Not really, they have different ways of ending up at similar points. Currently Intel has more bandwidth internal and AMD external. One of the reasons AMD increased this on the new processors. Then they had to do some work on Smart Cach and Smart Memory access too, just like Intel.

Donnie27, thanks for the clarification! ;)

The amount of bullshit on display in this thread is staggering...

Xeth, no. We don't do that here in the jungle. Let's all be civil and have a conversation on a topic that interests us, and inform each other, instead of using it is a place to wave one's insubstantial ePenis.
 
He's wrong as hell as usual. If AMD didn't have to update the motherboard, IMHO they wouldn't have. It is a waste of money and time. AMD could be working on something else they desparately needed finished or tweaked. Again, Dual core might not be hurt as bad on AM2 but Quad core will.

Please note? I tried to expain to Duby and Serge why 4 X 4 wasn't going to be a Kentsfield killer LONG before the first tests showed up. Like Dan here at HardOCP, I love motherboards and know a Processor ain't shit without it.

He's not wrong, he is absolutly 100% correct (about the HT issue anyway)... I don't know why some people seem to be missing this poing, but the cores are NOT using Hypertransport as ther interlink. The reason they are "wasting" their time on new boards is becuase people think they'll need them and somthing that says 3600MHz hypertransport sounds a lot better than something that says 2000MHz hyper transport and people will buy it. It is also very obvious to me that very few people know what hypertransport is used for, again, quad cores do NOT use HT for their interconnects... Quad Sockets would, quad cores do NOT, which means that weather you run single core, dual core, quad core, or 8 cores, if they are on a single socket, you will not saturate the bandwidth of a 2000MHz HT bus.
 
Thank you.

Most people dont seem to realize that AMD's SRQ switch is a dedicated partial mesh switch (tongue twister, say that ten times fast :D)

When comparing Intel's FSB, the only thing you can do is look at the IMC bandwidth, AND HT bandidth....

Some people will ask why.... Well, becouse in an Intel system, both IO and memory need to share the same bandwidth. In an AMD system, IO and Memory have separate data paths. The only way to fairly compare them is as an aggregate bandwidth.

That will of course change with CSI, but until then...
 
they updated it for the extra power saving stuff. stuff that many people would benefit from. not us. ;)

yeah, but that has nothing to do with the HT link. there are interconnects within the cpu to handle that traffic :p
but that's what intel is doing now... 4 cores, one memory controller. and it has to go through the FSB first to get to that memory controller. i fail to see where the problem is with amd's implentation, nor how HT link has anything to do with it. i understand how much you love intel, but please, understand how the HT link works and what it's there for before trying to bash it

Well, I already made the jump to X2 6000 and AM2, so I'm not exactly impartial. ;)... But I think I'm going to have to agree with Eclipse on this one. Assuming you're right then I look forward to slapping a quad core into this board when they're released.

If not, then I will be an unhappy AMD customer. *crosses fingers for next sixth months*
 
... Most people dont seem to realize that AMD's SRQ switch is a dedicated partial mesh switch (tongue twister, say that ten times fast :D)

When comparing Intel's FSB, the only thing you can do is look at the IMC bandwidth, AND HT bandidth....

Some people will ask why.... Well, becouse in an Intel system, both IO and memory need to share the same bandwidth. In an AMD system, IO and Memory have separate data paths. The only way to fairly compare them is as an aggregate bandwidth.

That will of course change with CSI, but until then...

Thanks Duby. At the risk of repeating myself; Ummmmm... bandwidth! :D
 
I'm thinking AMD will do something like the 800MHz/1066MHz P4s Intel had a while ago. For example, depending on the socket type a 2.4GHz Kuma will initialize at either a 12x multipler (200MHz DDR HT x 5) on an AM2 socket or with a lower multipler in an AM2+ socket.

That is unless HT 3.0 just uses really high multipliers (i.e. 9 x for 3.6GHz effective speed). Then the CPU multiplier just stays at 12x in either case. I wish I could find the HT 3.0 base frequency.

Assuming the reference clock stays at 200MHz like it is with sockets 754, 939, 940, and AM2, which it likely will be since AM2+ is backwards compatible and a change in reference clock would also requre a change in multiplier for the CPU as well when using in AM2 versus AM2+, then it has a multiplier of 9 and a base frequency of 1.8GHz. 9x200=1800Mhz, double pumped gives you 3600MHz effective.
 
Status
Not open for further replies.
Back
Top