![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
AM2 Reverse-HyperThreading
- Thread starter jarman
- Start date
- Status
there was a blurb and quite a large thread on slashdot about this. Naturally can't find it at the moment.
In any rate, there were quite a few knowledgable posts on the subject, both for and against. The theories I recall:
- 2nd core could be used to run ahead speculatively, with the aim of pre-filling a shared cache for the 1st core (or perhaps just do heavy speculative processing like the P4 does)
- cores could share execution units
- redundant execution a la Tandem which compares execution results or stages to check for differences (increasing reliability)
- instructions analyzed for dependency and instruction "strands" distributed to a different core when there was no dependency
Keep in mind that hammer has been in development for quite a few years, and it was the original intention to have a multicore CPU. Surely they would have recognized the large tradeoffs of parallelization and considered a solution, even way back then.
PS: Alot of those engineers had been working on the DEC AlphaProcessor project. The alpha, as i recall, had a feature where they kept several sets of the same CPU registers synchronized together so that multiple execution units could have concurrent 'local' access to them. That still leaves problems of locking/dependency, but it's interesting what they were doing way back then that resembles this very problem (1995?)
In any rate, there were quite a few knowledgable posts on the subject, both for and against. The theories I recall:
- 2nd core could be used to run ahead speculatively, with the aim of pre-filling a shared cache for the 1st core (or perhaps just do heavy speculative processing like the P4 does)
- cores could share execution units
- redundant execution a la Tandem which compares execution results or stages to check for differences (increasing reliability)
- instructions analyzed for dependency and instruction "strands" distributed to a different core when there was no dependency
Keep in mind that hammer has been in development for quite a few years, and it was the original intention to have a multicore CPU. Surely they would have recognized the large tradeoffs of parallelization and considered a solution, even way back then.
PS: Alot of those engineers had been working on the DEC AlphaProcessor project. The alpha, as i recall, had a feature where they kept several sets of the same CPU registers synchronized together so that multiple execution units could have concurrent 'local' access to them. That still leaves problems of locking/dependency, but it's interesting what they were doing way back then that resembles this very problem (1995?)
dandragonrage
[H]F Junkie
- Joined
- Jun 5, 2004
- Messages
- 8,295
Where the hell did the name Reverse HyperThreading come from? That is the worst name ever for it. Something like "Core Combining" would be much better, and I'm sure someone could think of something even better than that.
Rishi M. said:
The blurb asserts that the technology is already in shipping products, not speculating on technology that's "in development". Contrast this with the "announcement" that AMD had licensed ZRAM, which any blogger can publish without fear of being proven wrong. This is quite a limb TheInq is going out on.
On the other hand, it could still be anti-conroe FUD planted by AMD.
venm11 said:
Exactly what it is LOL! But you mean Anti-Conroe BS or Conroe FUD. Anti Conroe FUD would mean they were trying to clear FUD, not spread it. Hell I think? They have to save face for saying Hyperthreading was a Gimmick and never retracting the statement. They even have a few folks here who still believe everything they say.
AMD hates Hyperthreading because it was another of Intel innovations that made Intel Billions while AMD was still in the red.
jarman said:I happened to skim past this article on the Inq. Does anyone happen to have a link to a good write-up on this technology?
Scratches head
How in the hell can AMD add this feature to AM2 and then have it like magic appear in an Old 3800+?"It seems that in certain cases, even an old AMD Athlon 64 3800+ can wipe the floor with Core 2 Duo E6300 CPU."
In your dreams Fantoys! Some one, anyone, please explain this "old 3800+" thing?
Sabrewulf165
2[H]4U
- Joined
- Jun 23, 2004
- Messages
- 2,978
Donnie27 said:Scratches head
"It seems that in certain cases, even an old AMD Athlon 64 3800+ can wipe the floor with Core 2 Duo E6300 CPU."
In your dreams Fantoys! Some one, anyone, please explain this "old 3800+" thing?
If ya weren't busying yourself with being the embodiment of an overzealous anti-fan boy, you might make an educated guess that they were simply referring to the AM2 X2 3800+. The SKU "X2 3800+" is relatively old after all.
Granted that part of the article is quite vague, and the whole thing is probably bunk anyway, but if you're going to go after it based on semantics, at least put a little effort into it.
Sabrewulf165
2[H]4U
- Joined
- Jun 23, 2004
- Messages
- 2,978
venm11 said:
Well one of their writers was "sure" that AMD and ATI were going to merge... haven't heard much about that lately have you?
It would be very nice if this RHT thing turned out to be true and a real killer... it would certainly give people options, but I just have a gut feeling this will turn out to be bunk. One thing's for sure, if it *is* true and AM2 X2 procs have performance-effective RHT, we should hear about it sometime soon after the official Core 2 launch. Personally I've already sold my 939 setup and am just waiting until next month to see what my new DDR2 platform will be based on.
Sabrewulf165 said:
I'm not the one saying Webmasters as FOS for trying to add features AMD doesn't have to try and FUD Conroe. The only thing I questioned is how an OLD 3800+ and all of a sudden Morph a Feature from out of Thin Air? Now one said old X2 had the tech, they said AM2's. OLD 3800+ is not AM2 3800+. Sure it is more than likely Bunk. But trying to add the tech too even older tech foolishly lame LOL!
gwai lo said:One thing about this on the Inq and it's suddenly huge news and taken as truth by so many people...
I've already seen posts (on other forums) about "oh, but wait for RHT from AMD".
wtf? Rumor on Inq = truth?
I call it Hit Phishing!
Van Smith was made famous by it.
It wouldn't surprise me if they did. On the other hand, I just poked around and the first article I could find on the subject at Inq was this:
http://www.theinquirer.net/Default.aspx?article=32077
and it has very different language; the author isn't asserting anything other than that someone else speculated that there migth be a sale (link thereto included).
So... we need another example.
Sabrewulf165 said:
Well, to be fair everyone jumped on that bandwagon after the Forbes article began the speculation . Kyle even posted a comment from an Intel rep on the feasibility of the merger.
dandragonrage
[H]F Junkie
- Joined
- Jun 5, 2004
- Messages
- 8,295
Listen guys. As I mentioned before about the name. Just no. AMD would never consider calling it that since they have nothing to do with hyper threading, so the info is definitely not from AMD. IMO, this thread should be locked.
maverick777
Gawd
- Joined
- Jun 5, 2006
- Messages
- 540
People call it "reverse hyperthreading" for the following reason; Hyperthreading on Intel simply means tricking a single core CPU into behaving like a dual core CPU. AMD's "reverse hyperthreading" does the exact opposite. It makes dual core CPU's behave like single core CPU's. Optimizing X2 CPU's in single threaded applications. All of this is in theory and speculation of course. Although I wouldn't mind this "feature" in my 3800+ .
.dandragonrage
[H]F Junkie
- Joined
- Jun 5, 2004
- Messages
- 8,295
It is still the worst damn name anyone could have thought of. Sure, let's take some feature with end-user result A and compare it to another feature with an opposite end-result B. The former is Hyper-Threading, and B is Reverse Hyper-Threading? Sure, sounds like not too bad logic, but when implemented, these two things are completely different, and NOT opposites. Only the end results are basically opposites, and Hyper-Threading being Intel's marketing term for the technology, Reverse Hyper-Threading would NEVER be used on an AMD processor unless AMD got rights to use Hyper-Threading. And even then, it would STILL be a bad name.
Thus, to me at least, any article using this stupid name is an invalid article.
Thus, to me at least, any article using this stupid name is an invalid article.
VVVVVVVVVVVVVVVI
Supreme [H]ardness
- Joined
- Aug 7, 2000
- Messages
- 5,919
Donnie27 said:
Intel didn't invent Hyperthreading, they just renamed SMT.
Still, AMD is pretty much the only major manufacturer that doesn't use some form of SMT.
VVVVVVVVVVVVVVVI said:
Intel didn't invent SMT LOL! Just like AMD didn't invent the integrated Memory controller or Alpha's EV6 type bus LOL! Hyperthreading came From Alpha just like those techs did. Look up Jackson Tech. Oh hell I'll do it for you!
http://techreport.com/onearticle.x/1947
Chris_Morley
Former [H] Consumer Managing Ed.
- Joined
- Jul 5, 2000
- Messages
- 3,609
Actually, I'd like to know who at AMD has told you they hate HyperThreading...Donnie27 said:
Chris_Morley
Former [H] Consumer Managing Ed.
- Joined
- Jul 5, 2000
- Messages
- 3,609
And who at AMD did you have this discussion with?Donnie27 said:
Chris_Morley
Former [H] Consumer Managing Ed.
- Joined
- Jul 5, 2000
- Messages
- 3,609
I actually talk to a lot of AMD people on a regular basis, and I've never heard any of them say it. But I'll humor you. How about a VP or two, Directors of this or that, Product Managers, field sales engineers, and even a machinegun-toting Ph.D that invented pre-execute or "run ahead". None of them ever told me they hated HyperThreading.Donnie27 said:
I just know I wouldn't make sweeping generalizations about things I didn't have first hand knowledge of. That's why I asked my question. Are you somebody I should know of?
dandragonrage
[H]F Junkie
- Joined
- Jun 5, 2004
- Messages
- 8,295
jarman said:Your sense of logic is awe inspiring...
What it means is that none of this info is from AMD in any way, thus it's completely invalid. Anyone repeating it is just an AMD fan boy that can't admit that CONROE IS BETTER THAN AMD'S CURRENT STUFF.
Sabrewulf165
2[H]4U
- Joined
- Jun 23, 2004
- Messages
- 2,978
Chris_Morley said:
Haha, somebody got put in their place.
Chris_Morley
Former [H] Consumer Managing Ed.
- Joined
- Jul 5, 2000
- Messages
- 3,609
OK, I'll bite. Bang asked for a roadmap showing quadcore before 2008. I gave him one. He questioned its authenticity, I backed it up with pictures I personally took. Was I arguing its voracity? No, just that it existed and that it came straight from AMD.Donnie27 said:
But thanks for playing!
Chris_Morley said:
I'll bite back!
http://www.amdboard.com/amdroadmap.html
Got Socket F? Things do change and maps against real hardware is laughable at best. Anand admited that that he was bond by AMD to NOT compare AM2 to Conroe. are you guys under the same thumb?
Again, all Road Maps say they are subject to change without notice and are a just guides. That goes for Intel and AMD. Bangs map did show the first 65nm AMD Quad processors as K8 based and AMD said as much. The next new one doesn't ship until 2008, might be stampped and tested in 2007.
I loved this line;
http://www.digit-life.com/articles2/cpu/intel-conroe-2-13-ghz.html
Conroe seems too
for the AMD K8 core, despite our recent optimistic forecasts...- Status