I got an Andy Hui quote, whoohoo! Now all I need is one from Anand...
"Yes, AMD has mentioned dual core as an alternative to HT, although it would be expensive.
Heat would be doubled with two dies on the same package."
"Yes, AMD has mentioned dual core as an alternative to HT, although it would be expensive.
Heat would be doubled with two dies on the same package."
I think that AMD will do same thing with multithreading as they did with SSE and SSE2: they will wait 'till there's adequate support for it, then they will implement it. Right now, HyperThreading is +/-0. I mean, in some individual apps you can get big improvements, in most apps there is not tangible difference and is some apps it reduces performance.
Besides everyone! Remember IDF or Comdex last year? All the AMD Athlon 64 mobos had Hyperthreading support as one of their features, so I think a hyperthreading core is inevitable.
No, they didn't. One of the motherboard manufacturers made a typo. It was supposed to say "HyperTransport".Originally posted by: AtomicDude512
Besides everyone! Remember IDF or Comdex last year? All the AMD Athlon 64 mobos had Hyperthreading support as one of their features, so I think a hyperthreading core is inevitable.
It would be difficult at best for AMD to implement hyperthreading. Hyperthreading is dependant on the P4 architecture and would need a similar long pipeline to work as designed. AMD will have dual core processors "someday" though.
Originally posted by: AndyHui
No, they didn't. One of the motherboard manufacturers made a typo. It was supposed to say "HyperTransport".Originally posted by: AtomicDude512
Besides everyone! Remember IDF or Comdex last year? All the AMD Athlon 64 mobos had Hyperthreading support as one of their features, so I think a hyperthreading core is inevitable.
Oh darn...thanks for ruining it Andy.
Ummm, no. Multithreading has been around for a while, Hyperthreading is just Intels version of it. For example, Power5 from IBM will have something similar if I remember correctly. Multithreading is in no shape or form "dependant on the P4 architecture".
and then this...Besides everyone! Remember IDF or Comdex last year? All the AMD Athlon 64 mobos had Hyperthreading support as one of their features, so I think a hyperthreading core is inevitable.
AtomicDude512,
You seem pretty hell bent on twisting the AMD road map into something it's not. Not to sound like an a-hole, but let's face the facts:
1) No where in AMD's road map (published or non-published) do you see HT support. One motherboard oem makes a typo, and all of the sudden all the Athlon64 motherboards displayed at IDF support HT?
2) Read the very qoute that your passing around as gospel: "Yes, AMD has mentioned dual core as an alternative to HT, although it would be expensive. Heat would be doubled with two dies on the same package."... Microprocessor economics 101. The 2 items in bold are atop the list of no no's when making a mass market chip.
CH2
Good points but,
1)K6 was pre slot and used the super7 socket, im talking about p3's and new cllies and your comparing the k6 to them? thats like comparing PPro to T-bird. Last generation is somthing like tbird and duron for AMD, and last gen for intel is P3 and cellies
2) Why does older generation stuff still matter? because were talking about roadmaps, how can we compare road maps with out looking at older generations. Even in real life things like this matter, My friend can keep is Duron mobo and upgrade to an XP CPU with out changing anything accept for the CPU total cost, 60 bucks(xp1800). My girlfriend however cant upgrade her p3 733mhz with out buying a new mobo. newmobos dont take sd ram that the p3 took, so even on the low end of the spectrum, 200~ dollars for an comparable upgrade. Previously even more, only recently has DDR ram droped to a sane price.
3)Yes AMD had a slot->socket A->HAmmer SP/MP, but intel was slot->socket 370->split of 3 sockets (423,604,478). Not counting the hammer because it hasnt been released yet, AMD has changed the socket once, Intel has changed the socket twice/three/four times depending on how you want to count the split into p4 and xenon
4)Yes hammer was released late, probly shouldnt have even mentioned that
5) I dont care how you split it, AMD road map is much more upgradable and scaleable than intels roadmap. We will have to see how well the Hammer scales compared to the p4, so far p4 is doing very well.
Yeah, NO KIDDING! That's what I just said! If you had bothered to read my post (which you copied and pasted) you would see that I said that HYPERthreading is dependant on the P4 architecture NOT MULTIthreading. Geez....learn to read.
All Athlons are based on the same 7th generation CPU design from AMD. Tbird is no more a last generation CPU, than a Williamtte is last generation vs the Northwood. The P4 is Intel 7th gen CPU, Athlon is AMD's 7th gen , Pentium Pro/II/III is Intel's 6th gen, K6/2/3/Plus is AMD's 6th gen.
Again, you're comparing different generation CPU's. You can't compare upgrading 7th gen AMD to another 7th gen AMD and upgrading from 6th gen Intel to 7th gen.
Socket 370 is not a P4 socket, nor is 604. Socket 604 is for a completely different market and is irrelevent to the home user. With P4 Intel has gone from socket 423 to socket 478. For Athlon, AMD has gone from Slot A to Socket A. However, that is misleading, as just because the socket doesn't change doesn't mean the board is useable for all CPU's. Socket A has gone from 200Mhz bus to 266MHz bus to 333MHz bus. Good luck trying to get a CPU to work in a board that doesn't support the proper bus speed. If you get it to work at all, it often won't be running at the correct clock rate. I don't think there is a winner in the who has more incompatible formats competition.
You are aware that the Opteron and Athlon 64 will use different sockets just like the the Xenon and P4, correct? Different markets, different products, it's that simple.
No. No, hyperthreading is not Intel's version of Multithreading. Hyperthreading does not run two threads at the same time. It actually has two threads in process but it switches between them so fast that it appears to be running them simultaeously. Multithreading on the other hand, does run two threads at the same time on seperate cores entirely. IBM's two core implementation is MULTIthreading. Intel's HYPERthreading is just a way of using all of the processor's "units" more effeciently. It is not MULTIthreading.
Yes, Hyperthreading is Intel's brand name for their implementation of simultaneous multithreading, a hardware multithreading technique.
Yes, hyperthreading does run two threads simultaneously. SMT/Hyperthreading can fetch, issue, execute, and retire instructions from different threads in the same clock cycle. This is different from older multithreading techniques like course-grained multithreading (a fast thread-switching mechanism that begins fetching from a new thread after an event such as an L2 cache miss) and fine-grained multithreading (switching the thread from which the processor fetches and issues instructions each cycle). A Pentium 4-like design is not required for SMT; the important attributes are dynamic scheduling, register renaming, and high-issue rate, features that the P4 shares with most other high-performance microprocessors and designs that will have SMT, the IBM POWER5 and Sun UltraSPARC V. The now-defunct Alpha EV8 was also going to feature a very impressive 4-way SMT implementation.
The POWER4 is a chip-level multiprocessor (CMP), which is not a version of multithreading.
* not speaking for Intel Corp. *
Originally posted by: paralazarguer
Lol...right from Intel's mouth...
I'll take it right from the FAQ's thanks.
Nope, right from my mouth. I'm a grad student studying computer architecture, I've read nearly every simultaneous multithreading publication from academia and the industry in conjunction with my research, and I've done work on an in-house (at UW) functional-level execution-driven SMT simulator. I have just a little bit of insight into the topic.
Or, if you would like, you can take it straight from the horse's mouth.
* not speaking for Intel Corp. *
Originally posted by: Sohcan
Originally posted by: paralazarguer
Lol...right from Intel's mouth...
I'll take it right from the FAQ's thanks.
Nope, right from my mouth. I'm a grad student studying computer architecture, I've read nearly every simultaneous multithreading publication from academia and the industry in conjunction with my research, and I've done work on an in-house (at UW) functional-level execution-driven SMT simulator. I have just a little bit of insight into the topic.
Or, if you would like, you can take it straight from the horse's mouth.
* not speaking for Intel Corp. *
Looks like the FAQs need to be updated
Chiz
Originally posted by: paralazarguer
Yeah, NO KIDDING! That's what I just said! If you had bothered to read my post (which you copied and pasted) you would see that I said that HYPERthreading is dependant on the P4 architecture NOT MULTIthreading. Geez....learn to read.
Oh for crying out loud.... HyperThreading is Symmetric Multithreading. SMT has been around for a while. Intel started using it recently and called it HyperThreading. HyperThreading == Symmetric MultiThreading. And SMT is in no way dependant on P4 architecture.
So HYPERthreading is not dependant on P4 architecture since it's exactly same as Symmetric Multithreading (HyperThreading is just Intels brand-name for it) and SMT has been around for a while and it's NOT sometning that is exclusive to just Intel and P4!
Originally posted by: NaughtyusMaximus
It is if you want to call it HyperThreading.
Of course AMD couldn't use Intel trademarks (which I assume HyperThreading is). It would be same if Intel renamed L2-cache in to "megaspeed-technology". AMD could still use L2-cache just fine, they just couldn't call it "megaspeed". AMD could use SMT just fine, they just couldn't call it "HyperThreading". Trademark is tied to the CPU (or more presicely, to the company), but the tech is not.
Originally posted by: Nemesis77
Originally posted by: NaughtyusMaximus
It is if you want to call it HyperThreading.
Of course AMD couldn't use Intel trademarks (which I assume HyperThreading is). It would be same if Intel renamed L2-cache in to "megaspeed-technology". AMD could still use L2-cache just fine, they just couldn't call it "megaspeed". AMD could use SMT just fine, they just couldn't call it "HyperThreading". Trademark is tied to the CPU (or more presicely, to the company), but the tech is not.
yes, the name is a trademark, but intel can patent technology, and call amd on it if they think amd is copying outright their products.
Originally posted by: Zugzwang152
Originally posted by: Nemesis77
Originally posted by: NaughtyusMaximus
It is if you want to call it HyperThreading.
Of course AMD couldn't use Intel trademarks (which I assume HyperThreading is). It would be same if Intel renamed L2-cache in to "megaspeed-technology". AMD could still use L2-cache just fine, they just couldn't call it "megaspeed". AMD could use SMT just fine, they just couldn't call it "HyperThreading". Trademark is tied to the CPU (or more presicely, to the company), but the tech is not.
yes, the name is a trademark, but intel can patent technology, and call amd on it if they think amd is copying outright their products.
Symmetric MultiThreading is not patented by Intel. They might have some details of their implementation patented, but the concept and technology itself is something that AMD could use. Like I said, IBM is planning to use it as well, and I have heard that Sun is going to do so too (with their SPARC-series).
Originally posted by: NaughtyusMaximus
It is if you want to call it HyperThreading.
They should call it HyperZ Threading MaX
Bfonnes
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter