REVERSE Hyperthreading for AM2!

Status
Not open for further replies.
Nice edit. But I already read what you wrote before.

apokalipse said:
no, I did mean silicon-germanium transistors. IBM and AMD are developing them together.
http://news.zdnet.com/2100-9584_22-5982887.html

although yes, AMD may use Nickel Silicide aswell (which I admit I didn't know before).

Good enough. From that article alone, it wouldn't have justified your argument, but the link below says that it is an option for AMD so I'll admit that its a possibility.

That said, I do applaud you for admitting that you did not know about Nickel Silicide.

apokalipse said:
less power consumption. that's what I mean by efficiency.
plus there's the lower production costs.

Also admitted.

apokalipse said:
can you say that for certain?
yes, current AM2 CPU's are not that great. but they are AMD's first chips to use DDR2. and their memory controller really needs improving.

Well, you say below that you can't say for certain whether AMD will be using SiGe in K8L, but that its "pretty likely." Similarly, I am also pretty sure that the performance will not increase. Please don't blame it on the memory controller, from the same source as your SiGe information came from, it also reports that AMD included, in AMD, "a surprisingly robust DDR-2 controller with an improved Crossbar and Arbitrator." If you wish me to take that website as your source on SiGe, then you should admit this also.

Plus, unlike SiGe, we have seem the performance of DDR2 on AM2. It has not been impressive. To say that "oh well, whatever, they'll improve it later on" is a sad excuse. You people wanted to wait to see how DDR2 performed on the "Final :rolleyes: Edition" processors, and when it didn't improve, this excuse of the faulty memory controller was an easy escape. But back on topic, we have seen the performance of DDR2. As has been said repeatedly before, Athlon 64 processors do not benefit from the additional bandwidth that DDR2 provides. And since even the fastest DDR2 performance (nothing to do with the processor) is not much better than DDR performance, there won't be any improvements on that front either.

apokalipse said:
I don't know exactly how, since I don't work for AMD. but it will involve changing the architecture to process more instructions/clock cycle, like they did from K7 to K8. K8 chips don't need as high frequencies to get the same performance as K7 chips.

K7 to K8. And K8 to K8L. See a discrepancy? K8L is an evolution, not a revolution. You don't have to work for AMD to know that changes in K8L. Even I know about them, and I don't care much for K8L. From what I've seen, I can assure you that other than the additional complex decoder, there will be nothing more to help AMD on clock efficiency.

apokalipse said:
I can't say for certain they will be used on K8L, but I'd say it's pretty likely IMO.

http://www.hi-techreviews.com/modules.php?name=News&file=article&sid=7156

I wouldn't go so far as "pretty likely" since Nickel Silicide is a much easier alternative, but a possibility, yes.
 
As for you RHT guys. Here ya go. Apparently, RHT is real, it is incorporated into AM2, and it will kick Conroe's arse. That also means that the rumours that the Original Poster was talking about came true. A strange thread, indeed.

AMD Socket AM2 has a Secret Weapon. Today's article.

Yes, Opterons are also coming for AM2 and will also have RHT.

Opteron info. Same day stuff.

--------------------------------

Now lets break down the AM2 article.

It seems that all AM2 CPUs were outfitted with a support for Reverse-HyperThreading, an architectural change which enables software to think that it is working on a single-core alone. By combining two cores, the company has been able to produce the six IPC "core" that will go head to head against four IPC "core" from Conroe/Merom/WoodCrest combo.

It seems that in certain cases, even an old AMD Athlon 64 3800+ can wipe the floor with Core 2 Duo E6300 CPU.

Seems right. Normally, it would only take a 1.54Ghz Conroe to match the X2 3800+. Also true is the fact that the Conroe is not twice as fast as the X2. If you can indeed create a 4.0Ghz processor, that would be impressive. Now, keep in mind that this would be a 4.0Ghz single-core Athlon 64 vs. a 1.86Ghz Dual Core Conroe proecssor. In a Dual Core environment, the Conroe would kick the A64's arse. However, in a single-core environment, such as most of today's applications - the Conroe would only be performing like a 2.4Ghz single-core Athlon 64. A X2 3800+ w/RHT in a single core environment would indeed by wiping the floor with the Conroe. I hope that this paragraph has showed you that I am not an Intel/AMD fanboy, only a performance fanboy.

As we all know - the results from E6700 and X6800 against FX-62 will be nice, but the real fight with AMD is the one for the Conroe with 2MB of L2 cache. The system memory avoidance technology is working flawlessly on a 4MB cache model, but the case is reversed in the two Meg cache variant, especially in cache-hit sensitive apps, such as games.

This, however, is incorrect. It is true that Intel did not include an Integrated Memory Controller because they believed that more cache would bypass the need for an IMC (so the Conroe got more cache in place of an IMC). They were correct. However, the 2MB cache on the E6300 and E6400 has, in no way, inhibited the Conroe's performance. A 2.13Ghz Conroe E6400 still beats a FX-60 easy, and challenges a FX-62.

In single-treaded apps, Core 2 Duo is expected to struggle against Reverse HyperThreading CPUs, which work at higher clock frequencies and produce higher instruction per clock ratios (IPC).

True indeed. As was said before in the article, this would enable AMD to effectively have a 6-issue core to Conroe's 4-issue core. IPC is just another way of saying "X-issue core." Scary stuff for Intel, but keep in mind - only in single-threaded applications.

AMDs Reverse-HT is a dynamic technology, and with Microsoft's Windows update and a new processor driver, the driver will copy the graphics drivers of today's 3D accelerators. The driver will detect the app, see if it is multithreaded or not and turn the ReverseHT on, or leave it off.

That, Nitestick, is the answer to your question as to how the cores will handle switching. And that also confirms that only a Windows and BIOS update would be required to enable it.

-----------------------------

You didn't expect it to be that easy, did you?

Intel is working on something simlar. Except they call it Core Multiplexing.

Conroe may have a "Secret Weapon" embedded into it, just like AM2 did.

If true, AMD's RHT just went to waste, as a 1.86Ghz Conroe can Multiplex to a 3.72Ghz single-core Conroe (4.84Ghz Athlon 64).

So with both processors RHT'ing or Multiplexing - a 4.84Ghz Athlon 64 (The Conroe) vs. a 4.0Ghz Athlon 64 (The A64). Conroe still wins.

-----------------------------

Oh what a mess we've made.
 
I'm making seperate posts because they're on seperate topics.

I was incorrect in my last post when I said X2 3800+ @ 2.0Ghz + RHT = 4.0Ghz A64. Some of the smarter people on XS have said that the increase is not 100%, so it won't be 4.0Ghz. It seems correct, as we have seen Dual Core CPUs in Dual-Core supporting games not perform like they were twice their clockspeed. A X2 3800+ was benchmarked in Quake 4, and I would guess that it was performing like a 3.2 to 3.4Ghz Athlon 64. So a 66% (2/3) performance increase, maybe?

People are also saying that nothing is official, so some people (AMD people as well) are going to wait until AMD says something about it themselves. I agree that it is not official, and anything could happen, but even I think its likely to be true. Then again, I don't trust the person that posted the article on The Inq.
 
Take it as a grain of salt, the inquirer is a very unreliable source and no doubt all these websites posting this story are simply borrowing the russian site as the source...notice how they say that it is still considered both a rumour and that AMD has still not commented on it

Why do I still consider this a rumour and nothing more? When socket 939 was first released is was known almost a year in advance that the socket would be capable of supporting multicore processors with a BIOS update straight from AMD as a source. I have no doubt that if AMD were to include this easter egg in AM2 they would have made it known almost right away
 
AMD is supposed to announce this on the day after Conroe. However, it wasn't AMD that said that :)

And the Intel "BadAxe" motherboard has Multiplexing as an option in it's BIOS. Thats as good a confirmation as I need.
 
well with conroe being just around the corner i suppose we can find all the answers. cheers for the info infomatic or flanker or whatever i'm supposed to call you ;). that does explain the switching but i was thinking more of the multitasking side of things. when you are running multiple applications and one of them is using RHT what would happen if you were trying to run a couple of background apps on seperate cores in the background. personally i'm not a big fan of the RHT concept anyway, it seems the main point of it is to improve games performance which isn't really a problem on a decent dual core anyway. i prefer to use the second core for multitasking while i am gaming rather than waste them both on gaming.

I was incorrect in my last post when I said X2 3800+ @ 2.0Ghz + RHT = 4.0Ghz A64. Some of the smarter people on XS have said that the increase is not 100%, so it won't be 4.0Ghz. It seems correct, as we have seen Dual Core CPUs in Dual-Core supporting games not perform like they were twice their clockspeed. A X2 3800+ was benchmarked in Quake 4, and I would guess that it was performing like a 3.2 to 3.4Ghz Athlon 64. So a 66% (2/3) performance increase, maybe?

60% sounds like a good ballpark figure but i'm not sure if quake 4 is the best to try and guage it because it has it's FPS seemingly capped.
 
Ooh, Good question. So if there are two applications, one multithreaded and one single-threaded, then what would happen, right? Yea, thats an interesting one.

For Quake 4, I remember seeing the FPS go higher with the more powerful Dual Cores. Is it possible to unlock the FPS cap in that game, like you can do in BF2?
 
not sure. i have heard some benchmarkers complain about the cap and how it makes their results inaccurate so i guess if there was an unlock they would have found out. personally i haven't looked into it.

i was thinking they could have some sort of a monitoring system like they have for cool n' quiet that polls at like 1000hz to switch between modes as required at that instant. don't think that would work......somehow i just see a lot of BSOD's coming from that approach
 
Infomatic said:
Well, you say below that you can't say for certain whether AMD will be using SiGe in K8L, but that its "pretty likely." Similarly, I am also pretty sure that the performance will not increase. Please don't blame it on the memory controller, from the same source as your SiGe information came from, it also reports that AMD included, in AMD, "a surprisingly robust DDR-2 controller with an improved Crossbar and Arbitrator." If you wish me to take that website as your source on SiGe, then you should admit this also.
yeah, I do disagree about the DDR-2 memory controller being "surprisingly robust" but I posted the article to show you that I'm not the only one who thinks SiGe will probabbly be used by AMD. the DDR2 memory controller really needs improving.

Infomatic said:
Plus, unlike SiGe, we have seem the performance of DDR2 on AM2. It has not been impressive. To say that "oh well, whatever, they'll improve it later on" is a sad excuse. You people wanted to wait to see how DDR2 performed on the "Final :rolleyes: Edition" processors, and when it didn't improve, this excuse of the faulty memory controller was an easy escape. But back on topic, we have seen the performance of DDR2. As has been said repeatedly before, Athlon 64 processors do not benefit from the additional bandwidth that DDR2 provides. And since even the fastest DDR2 performance (nothing to do with the processor) is not much better than DDR performance, there won't be any improvements on that front either.
well I did not say that AMD will or won't benefit from DDR2 even with an improved DDR2 memory controller (which they do need regardless), maybe it will benefit, maybe it won't. we'll just have to see. my point was that you can't say anything for certain.



K7 to K8. And K8 to K8L. See a discrepancy? K8L is an evolution, not a revolution. You don't have to work for AMD to know that changes in K8L. Even I know about them, and I don't care much for K8L. From what I've seen, I can assure you that other than the additional complex decoder, there will be nothing more to help AMD on clock efficiency.
we can know the basic changes
when I say basic, I mean from AMD's point of view. adding a complex decoder, yes that's a change.
but I mean AMD doesn't tell us exactly what the architecture of the CPU is (if they did, Intel would be all over it also) I mean like, the actual blueprints with all the architectures, which transistor goes where, and so on...

AFAIK, AMD will be improving their architecture. I don't know how, AMD does. yes, K8L is an evolution of K8. that doesn't mean they can't change their architecture in some ways, and it still be similar to K8. like I said, we can't know these things. only AMD can.

though I would say it is 99% certain that AMD is changing their architecture. to what degree is another matter.

Infomatic said:
I wouldn't go so far as "pretty likely" since Nickel Silicide is a much easier alternative, but a possibility, yes.
well, AMD is co-developing them with IBM. and we have seen a big success from IBM while using them (remember that 500GHZ chip?)
 
Status
Not open for further replies.
Back
Top Bottom