Next Gen Hypertransport 3.0 is coming

W

Wrench00

2[H]4U
Joined
Sep 30, 2003
Messages
3,423
http://www.dailytech.com/article.aspx?newsid=1943

Wow AMD is really thinking about the future now. I remember watching a bit from ex-intel engineer and how he stated the p4 core was a piece of junk that would never scale that well because of IO. Any who Intel is hyping Conroe which doesn't even exist in the market. All intel !!!!!!s are Conroe this Conroe that, it will beat AMD this, it will smoke AMD that.

AMD is making some really brilliant moves. Opteron has been very succesful not because its ultimate CPU but because of the platform. I think AM2 will be short lived before they roll out next gen stuff.

As an ex-ibmer Server support guy this HT3.0 is going to take an even bigger chunk of Intel's server market. Desktops sales are good for AMD but no where nearly as good as server sales. I would love to see HT interface Vid cards screw PCIE, HT would smoke it anyday. Addons FPU's, PPU, GPU that directly interface with the CPU.

Intel can wave its flag and rant and rave and spend countless billions on PR but it chooses to stick with obsolete technology that boosts short term goals whilest AMD is investing its cash on R&D and future growth. 2007 is going to be awsome.
 
Intel has obsolete technology? Conroe smokes dual core AMD at the same clock speed by 20% or more in PCmark.. look on xs

And thought I'd point out that I see no difference with 600MHz htt or 1000MHz htt on my setup. It's already fine the way it is
 
repoman0 said:
Intel has obsolete technology? Conroe smokes dual core AMD at the same clock speed by 20% or more in PCmark.. look on xs

And thought I'd point out that I see no difference with 600MHz htt or 1000MHz htt on my setup. It's already fine the way it is
Click to expand...

Do you have a Conroe hell can you even buy one?

Obsolete Technology! Hell yes.. FSB has been Obsolete for a long time. Conroe would be even better if they coupled it to an IO system that didn't suck a big one.
 
And thought I'd point out that I see no difference with 600MHz htt or 1000MHz htt on my setup. It's already fine the way it is
Click to expand...

Which, ironically is a testament to just how absolete the FSB actually is.
 
repoman0 said:
No I don't have a Conroe. FCG does though, and came up with some awesome results.. 2.16ghz cpu running [email protected]. And it smokes anything in the current market :eek: Can't wait to get one!

http://www.xtremesystems.org/forums/showthread.php?t=97609

Who cares how old FSB is, if it performs like this then who gives a shit.
Click to expand...

Argh You missed the whole point of my post.

I was talking about IO. Who cares if Conroe beats a Single AMD CPU. HyperTransport is about massive SMP. If you haven't clued in most server chips are usually slower then the desktop parts. AMD and Intel make the best money on the server chip sales and not desktop sales since the margins are slimmer. Look at the prices of high end SMP products like XEON DP/MP and Opteron 8XX line. Did you ever read any SMP benchmarking articles. 4+ SMP configurations and such, AMD just whoops Intel like silly little girl. What HT 3.0 means is that AMD can capitilize on 8+ socket configuration chip sales.
 
And.. I don't care about that, I'm sure very few others here care either. So, great, AMD wins in the server market and they make money. And about 10 people on this forum will benefit from that.. the rest of us will be using conroe, or AMD's new stuff in 07 if we have patience
 
Wrench00 said:
Argh You missed the whole point of my post.

I was talking about IO. Who cares if Conroe beats a Single AMD CPU. HyperTransport is about massive SMP. If you haven't clued in most server chips are usually slower then the desktop parts. AMD and Intel make the best money on the server chip sales and not desktop sales since the margins are slimmer. Look at the prices of high end SMP products like XEON DP/MP and Opteron 8XX line. Did you ever read any SMP benchmarking articles. 4+ SMP configurations and such, AMD just whoops Intel like silly little girl. What HT 3.0 means is that AMD can capitilize on 8+ socket configuration chip sales.
Click to expand...
Maybe, the bulk of HardForum readership? :D Yes, the margins are higher on the server front but large voulmes also help out as well. Considering most servers shipped are the 2P kind it's no surprise. Considering the Xeons are still based on NetBurst technology, that is no surprise that AMD Opterons are extremely competitive against them.

It's not surprising as Xeon MP, tend to have slower FSB then regular DP line and with bandwidth hungry NetBurst that is, the performance would be bound to suffer.

Intel's as well doesn't seem to be really concentrating their efforts on this segment 4P+ anyway they are more concerned with the larger 2P Server segement with the Woodcrest introduction with Dual Independent Buses of 1.33GHZ each as well as processors of up to 3.0GHZ.

Xeon MP is looking to be the last to move to Core Architecture, as it seems to be the lower priority at Intel.
 
repoman0 said:
And.. I don't care about that, I'm sure very few others here care either. So, great, AMD wins in the server market and they make money. And about 10 people on this forum will benefit from that.. the rest of us will be using conroe, or AMD's new stuff in 07 if we have patience
Click to expand...


UHM you are a trully huge an ignoranus. About 10 people with benefit!?!$#@..

WTF?

All of us will benefit, servers will be faster, faster forums, queries, DNS, queries. Hell google searches will be faster and more accurate. Most search engines will benefit greatly, databases will be more robust, datacenters will be able to store more. Faster interconects, better Internet speeds etc.. Hell grid computing that does groundbreaking simulations on diseases, mathematics, physics will benefit greatly from HT3.0 Even entertainment will benefit from HT.

You must be young because your scope of your vision is very narrow.
 
repoman0 said:
And.. I don't care about that, I'm sure very few others here care either. So, great, AMD wins in the server market and they make money. And about 10 people on this forum will benefit from that.. the rest of us will be using conroe, or AMD's new stuff in 07 if we have patience
Click to expand...

One thing to note though, is this whole new world of coprocessors. Sure you can make an argument that only specialized software can use it. But the same thing was true of the GPU years ago.

I think this is the beginning of the revolution. aka the integration race. And AMD is in the lead.
 
coldpower27 said:
Maybe, the bulk of HardForum readership? :D Yes, the margins are higher on the server front but large voulmes also help out as well. Considering most servers shipped are the 2P kind it's no surprise. Considering the Xeons are still based on NetBurst technology, that is no surprise that AMD Opterons are extremely competitive against them.

It's not surprising as Xeon MP, tend to have slower FSB then regular DP line and with bandwidth hungry NetBurst that is, the performance would be bound to suffer.

Intel's as well doesn't seem to be really concentrating their efforts on this segment 4P+ anyway they are more concerned with the larger 2P Server segement with the Woodcrest introduction with Dual Independent Buses of 1.33GHZ each as well as processors of up to 3.0GHZ.

Xeon MP is looking to be the last to move to Core Architecture, as it seems to be the lower priority at Intel.
Click to expand...

Because if they made XEON MP not suck they would kill the ITANIC. Did you see the prices of those things? AMD has no supercomputing type product so its not undercutting itself in anyway.

But you still have to admit that even with 2p boxes and 2 cores per socket that 1333FSB isn't going to supply no where the IO what the conroe corewill need. Wait till late 2007 when quad cores and up will start hiting the market.
 
Wrench00 said:
Because if they made XEON MP not suck they would kill the ITANIC. Did you see the prices of those things? AMD has no supercomputing type product so its not undercutting itself in anyway.

But you still have to admit that even with 2p boxes and 2 cores per socket that 1333FSB isn't going to supply no where the IO what the conroe corewill need. Wait till late 2007 when quad cores and up will start hiting the market.
Click to expand...
Admit, we'll see iif 2x 1333FSB remember one for each Socket won't be enough for Woodcrest, I still expect a performance lead from Intel thanks to high IPC and high clockspeed.
 
Wrench00 said:
UHM you are a trully huge an ignoranus. About 10 people with benefit!?!$#@..

WTF?

All of us will benefit, servers will be faster, faster forums, queries, DNS, queries. Hell google searches will be faster and more accurate. Most search engines will benefit greatly, databases will be more robust, datacenters will be able to store more. Faster interconects, better Internet speeds etc.. Hell grid computing that does groundbreaking simulations on diseases, mathematics, physics will benefit greatly from HT3.0 Even entertainment will benefit from HT.

You must be young because your scope of your vision is very narrow.
Click to expand...

lol, please stfu. You kids get way too overworked about this stuff *wishes there was a laugh guy once again*

And I'm sure every forum will upgrade to HT3.0 cpu's asap.. we haven't even seen htt2 yet. I highly doubt google will be upgrading, if you keep up on the news you'll see they just invested a whole lot of money in a HUGE opteron array.. maybe they have htt3 early :rolleyes: How will datacenters be able to store more? Faster CPU's magically give them hdd space.. I see

The goddamn point is, hardly anyone here will see the difference that htt3 will bring. Most of us use simple desktop computers, the internet is already hella fast (bottleneck sure as hell isn't in the CPU's, it's the type of connection.. cable can only go so fast (40mbps/channel, same with dsl (8mbps/connection), t1 (1.5), fiber, whatever).

Explain to me again how entertainment will benefit as well? You have no idea what the fuck you're talking about, you're pulling stuff out of your ass.
 
repoman0 said:
UHM you are a trully huge an ignoranus. About 10 people with benefit!?!$#@.. WTF? All of us will benefit... You must be young because your scope of your vision is very narrow.
Click to expand...
repoman0 said:
... hardly anyone here will see the difference ... You have no idea what the fuck you're talking about, you're pulling stuff out of your ass.
Click to expand...

No, repoman0, he's not. He called you out on your lack of vision and your small, selfish world-view....

Then you made another short-sighted post and proved him even more correct.

It's your own fault you look like an idiot; don't get mad at him for your own shortcommings.
 
repoman0 said:
lol, please stfu. You kids get way too overworked about this stuff *wishes there was a laugh guy once again*

And I'm sure every forum will upgrade to HT3.0 cpu's asap.. we haven't even seen htt2 yet. I highly doubt google will be upgrading, if you keep up on the news you'll see they just invested a whole lot of money in a HUGE opteron array.. maybe they have htt3 early :rolleyes: How will datacenters be able to store more? Faster CPU's magically give them hdd space.. I see.
Click to expand...
UHM Iwill and other server mobo manufacturers already have HTX implementations that enable large computing grids through Infiband. Cray, HP, are already utilizing these features in their blades and servers. Faster IO means you can add more Hard drives and other Storage mediums with out completly taxing the bus. You can add co-processors that increase FPU power.
Look at your SIG you have an AMD system your already using HT 2.0.
repoman0 said:
The goddamn point is, hardly anyone here will see the difference that htt3 will bring. Most of us use simple desktop computers, the internet is already hella fast (bottleneck sure as hell isn't in the CPU's, it's the type of connection.. cable can only go so fast (40mbps/channel, same with dsl (8mbps/connection), t1 (1.5), fiber, whatever).
Click to expand...
Actually HTT2.0 was adopted fairly quickly after ratification so you will most likely see new motherboards using HTT3.0 by the end of this year. With AMD plans of quad cores for socket F they will need all the IO the platform can deliver.
Internet is not fast enough for future delivery of digital content to your living space. That includes movies, games, music. As far as your internet speeds are concerned you are very unfortunate.I live in SK and internet is no where as limited as it is in North America.. I have VDSL so its much faster the ADSL . You might want to look up DSL tech first, before jumping to conclusions.

repoman0 said:
Explain to me again how entertainment will benefit as well? You have no idea what the fuck you're talking about, you're pulling stuff out of your ass.
Click to expand...
Uhm faster render farms for CG, larger datacenters that can handle larger queries that can stream music, movies and games to your protible devices. ETC,

The only person that is overworked here is you.

I will not even bother answering your questions. Because you obviously didn't read the HT3.0 PDF or even the news article. I was in server support long enough to see direct results of IO improvements.

I also live in one of the most wired nations on Earth. My Cell has TV, my DSL has WAY over 8mbps limit mainly because I have VDSL. You are truly deprived of knowledge.
 
duby229 said:
Which, ironically is a testament to just how absolete the FSB actually is.
Click to expand...

The Athlon 64 series doesn't HAVE a frontside bus. The memory controller is built right into the chip, which is why it doesn't matter if its 600MHz or 1000MHz. Either way, the bandwidth is more than enough. Hell, it's overkill at anything over 400MHz.
 
Bona Fide said:
The Athlon 64 series doesn't HAVE a frontside bus. The memory controller is built right into the chip, which is why it doesn't matter if its 600MHz or 1000MHz. Either way, the bandwidth is more than enough. Hell, it's overkill at anything over 400MHz.
Click to expand...

Which is exactly my point. Those are the very reasons why it is better than a FSB. The very reasons why a FSB is absolete......

Ironic isnt it?
 
Wrench00 said:
As an ex-ibmer Server support guy this HT3.0 is going to take an even bigger chunk of Intel's server market. Desktops sales are good for AMD but no where nearly as good as server sales. I would love to see HT interface Vid cards screw PCIE, HT would smoke it anyday. Addons FPU's, PPU, GPU that directly interface with the CPU.

Intel can wave its flag and rant and rave and spend countless billions on PR but it chooses to stick with obsolete technology that boosts short term goals whilest AMD is investing its cash on R&D and future growth. 2007 is going to be awsome.
Click to expand...
What exactly is the scope of your argument? HyperTransport as a peripheral bus (HyperTransport vs. PCI Express vs. PCI-X 2.0), or HyperTransport as a system bus (HyperTransport vs. Common Serial Interconnect vs. other system bus schemes)? First qualify your argument.

I would like to know from you which technologies from Intel you consider to be obsolete. I also would like to know when we can expect wide availability of HT videocards, physics cards, and so on.
 
xonik said:
What exactly is the scope of your argument? HyperTransport as a peripheral bus (HyperTransport vs. PCI Express vs. PCI-X 2.0), or HyperTransport as a system bus (HyperTransport vs. Common Serial Interconnect vs. other system bus schemes)? First qualify your argument.

I would like to know from you which technologies from Intel you consider to be obsolete. I also would like to know when we can expect wide availability of HT videocards, physics cards, and so on.
Click to expand...

There are already several socket 940 coprocessors being worked on right now.

Plus I'm sure there will be socket F coprocessors when that socket goes live.
 
does no one else find it kind of funny that people are arguing about the "real world benefit" of a whitepaper.

Doesnt matter how good a technology is, implementation is key. Seeing as we have nothing to benchmark or compare to, i find it really interesting that some of you can draw such strong conclusions at this point.

From the sounds of it, it could be very interesting, but its all going to come down to AMD properly making use of it, and if im not mistaken intel is going to be moving away from the FSB late 2007/2008. HT3.0 looks very appealing on paper, for both server and personal use imo, but as I said earlier, implementation is key.
 
xonik said:
What exactly is the scope of your argument? HyperTransport as a peripheral bus (HyperTransport vs. PCI Express vs. PCI-X 2.0), or HyperTransport as a system bus (HyperTransport vs. Common Serial Interconnect vs. other system bus schemes)? First qualify your argument.

I would like to know from you which technologies from Intel you consider to be obsolete. I also would like to know when we can expect wide availability of HT videocards, physics cards, and so on.
Click to expand...


FSB is obsolete. AMD dumped it because it was such a huge bottleneck for IO.Even tho PCIE is nice direct HTX interface would be better. Self configuring HT would enable any slot to set its own speed accordingly to needs.
 
Wrench00 said:
FSB is obsolete.
Click to expand...
Obsolesence is not really a sure sign of inferiority. One must just look at AMD's current DDR platform and 90 process. AMD is not standing still, as we all know, but neither is Intel. I haven't seen any indications from Intel which say that they will hold onto the FSB concept much longer than its useful life. As it stands now, however, a FSB isn't really holding Intel back from meeting performance goals.
 
You must log in or register to reply here.
Back
Top