• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

AMD's Secret Weapon

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Personally I believe that they're likely to go ahead with this. It makes sense.

Most current (and obviously old) software is single threaded. Considering you can significantly speed up single threaded software by doing this then why not? Likewise, there's plenty of types of software which will likely see little performance increase by going dual core, therefore the developers will likely not code them to be dual core.

Also (I don't know, but I'm guessing), wouldn't coding multithreaded apps be more complicated than coding it for one core? I'm guessing that game developers etc etc are likely to stick with single threaded coding as the most important thing to them tends to be to push a title out as soon as they can to make money.

Likewise, if Intel & AMD keep adding additional cores then it's going to become harder and harder to create efficient code for these CPU's. Can anyone imagine creating code for a 8 core CPU sometime in the future?

I believe reverse HyperThreading to be the answer to this issue, therefore I expect Intel to go this route also. I wouldn't be in the least bit surprised if Intel doesn't already have this feature on the Conroe, just awaiting a driver update to enable it. I can't see any logical reason as to why they can't do this also.

Besides, any software that benefits greatly from multiple cores will likely be coded for multiple cores. You aren't likely to see the same performance with software coded for single cores vs the same piece of software coded for multitiple cores. The feature will help, but it's not the answer to all the multi core issues. As I said, software that benefits most from multi core will likely be coded for multi core.

]-[itman said:
Just some clarification for ya :)

Also, I wouldn't count AMD out just yet, quite a few tweaks are coming down the road, more than many are expecting and don't forget, we're still comparing 65nm intel chips to 90nm AMD chips and saying intel edges out AMD in power and has a decent bump in speed. Just imagine if AMD gets 65nm right...

Nothing new there. AMD has pretty much always been behind Intel on manufacturing process. I don't really expect this to ever change, unless Intel runs into problems with their process sometime in the future. AMD's process's and Intel's are suffiently different (SOI etc etc) that either one of them could have an issue down the track and fall behind because of it. I'd put my money on it being AMD that has the problem though. Intel can usually just throw money and resources at a problem till it gets resolved, AMD can't.

Just my two cents guys, and sorry if it was a bit long, I've a habit of doing that. I prefer doing one post to multiple posts.
 
Last edited:
i would like to see this done on the software side instead of the hardware side. and by that i mean give the algorithm that amd uses to split up the code to the software writers so that they can recomile their single threaded code and make to make it for dual cores then they could let out a dual core patch to now single threaded apps
 
]-[itman said:
Just some clarification for ya :)

Much thanks :)

However, a smaller production size only allows for higher clockability, and less power consumption. I don't see any clock-efficiency changes directly from moving to 65nm. Yes, it gives more die space, but just a smaller production size only helps so much.

Dan0512 said:
We can expect them as soon as Q1 07....

Well, first of all; Intel's Kentsfield, as Hitman corrected me, is still a Quad-Core Desktop processor that will appear as an "Extreme Edition." So now we have established that Intel has Quad Cores based on the Core 2 architecture as early as Q1 2007, ok?

Now, for AMD, I cite This article at Dailytech.com as saying that AMD's Quad Core Deerhound processor for servers (based on the K8 architecture) will not be coming out until H2 2007. That means that AMD isn't getting Quad Cores until Q3 or Q4 of 2007. Now, it doesn't take a genius to figure out that even with AMD's HyperTransport technology, a Core 2 "Quad" processor will dominate a K8 Quad Core processor from AMD. For AMD to have any chance at all, the Quad Core has to be K8L. Now, I cite a different article, but also at Dailytech as saying that AMD's Desktop K8L Quad-Core processor isn't coming out until Q1 2008.

So Intel gets a year to make changes to Kentsfield before any real challenge appears. Until then, The Kentsfield vs. Deerhound issue would be like Conroe vs. K8 Athlon 64 X2 is right now.

mjw21a said:
I believe reverse HyperThreading to be the answer to this issue, therefore I expect Intel to go this route also. I wouldn't be in the least bit surprised if Intel doesn't already have this feature on the Conroe, just awaiting a driver update to enable it. I can't see any logical reason as to why they can't do this also.

I loved, and totally agree with your entire post. But, to focus on this part; Intel has Core Multiplexing, pretty much the same thing as Reverse HyperThreading, also built into the Conroe processors. A link to a thread on XS was posted on the first page. So yes, you were correct on every point, most accurate post I've seen in a while, as your outlook on the software was also spot on, IMO.

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

Somebody have something I can read to better understand just how RHT and Core Multiplexing work? I know what they do, have a vague idea of how they do it, but still trying to learn more.

I have to explain it to someone, and all I know is that it works by sending every other instruction to the second core. According to said person, thats not possible.
 
Last edited:
xFlankerx said:
Much thanks :)

However, a smaller production size only allows for higher clockability, and less power consumption. I don't see any clock-efficiency changes directly from moving to 65nm. Yes, it gives more die space, but just a smaller production size only helps so much.

The move to 65nm (rev. G) isn't just a die shrink, the core will be tweaked at the same time to allow for "higher IPC" (AMD's words). AMD probably saw what core2 was doing and decided to throw in a couple of features of K8L into rev. G to help close the gap. Even without it, if what theinquirer.net believes is true (sorry, it's late, I'm too tired to find the exact link), then AMD is expecting up to a 50% clock increase with 65nm. More reasonably, let's make it 35%-40% and you have a 3.8ghz stock AMD processor, which should atleast tie a 3.33ghz Conroe cpu in most cases with a few wins without any tweaks to the core. Of course, this is all somewhat speculation at this point, however, my point is that AMD will have a chance to come back swinging and I wouldn't count them out yet until we see what they actually have waiting in the pipeline.
 
xFlankerx said:
I loved, and totally agree with your entire post. But, to focus on this part; Intel has Core Multiplexing, pretty much the same thing as Reverse HyperThreading, also built into the Conroe processors. A link to a thread on XS was posted on the first page. So yes, you were correct on every point, most accurate post I've seen in a while, as your outlook on the software was also spot on, IMO.

Thanks mate. I've a habit of frequenting the rumour mill over at the Inq + the front page here. A good source of information when you sift fact from fiction.

Beyond that though, nothing special in my post I'm afraid. Just logic based on info at hand. Your comment is appreciated though. I tend to read much more than I post.

BTW: Nice to see Intel already have their equivalent to reverse HyperThreading. If that's the case then it should allow AMD to close the performance gap somewhat. I like competition, it keeps the prices low.
 
The amount of speedup you get would be determined by how much of the program is able to use from RHT. If the program is purely sequential, it's hard to imagine it benefiting from RHT. If RHT takes up additional resources, it could in fact cause slow down in a purely sequential program.

Though, it's all speculative.
 
Honestly I dont know why you guys are poopooing this its a great invention in the way of dual core technology, you're telling me you'd rather leave it up to the C++ programmer to divide his application into multiple threads rather then your CPU?! You've got to be kidding me this is how dual cores should have been done in the first place anyway...

gooutside said:
The amount of speedup you get would be determined by how much of the program is able to use from RHT. If the program is purely sequential, it's hard to imagine it benefiting from RHT. If RHT takes up additional resources, it could in fact cause slow down in a purely sequential program.

Though, it's all speculative.

I'm sure RHT would be completely under the users control however so it shouldnt be that much of an issue. Honestly we have no idea how this works or if its even true I guess so its hard to be speculating on technical aspects.
 
xFlankerx said:
Much thanks :)

However, a smaller production size only allows for higher clockability, and less power consumption. I don't see any clock-efficiency changes directly from moving to 65nm. Yes, it gives more die space, but just a smaller production size only helps so much.
IT helps alot. Optimizations, less heat, generally faster (less distance for data to dravel), etc. It helped with the Core 1s alot, and will with AMD.


Well, first of all; Intel's Kentsfield, as Hitman corrected me, is still a Quad-Core Desktop processor that will appear as an "Extreme Edition." So now we have established that Intel has Quad Cores based on the Core 2 architecture as early as Q1 2007, ok?

Now, for AMD, I cite This article at Dailytech.com as saying that AMD's Quad Core Deerhound processor for servers (based on the K8 architecture) will not be coming out until H2 2007. That means that AMD isn't getting Quad Cores until Q3 or Q4 of 2007. Now, it doesn't take a genius to figure out that even with AMD's HyperTransport technology, a Core 2 "Quad" processor will dominate a K8 Quad Core processor from AMD. For AMD to have any chance at all, the Quad Core has to be K8L. Now, I cite a different article, but also at Dailytech as saying that AMD's Desktop K8L Quad-Core processor isn't coming out until Q1 2008.
K8L will be sometime in the Q2 of 2007. This has been stated many times in this section, meaning that quad core from AMD will be out in the first half, not the second. socket 1207 quad cores might take longer since the socket isn't out yet.


Somebody have something I can read to better understand just how RHT and Core Multiplexing work? I know what they do, have a vague idea of how they do it, but still trying to learn more.

I have to explain it to someone, and all I know is that it works by sending every other instruction to the second core. According to said person, thats not possible.

In the article linked at the begining of this thread, it indicates that it is driver driven, loadbalencing instructions between cores with the help of some onboard hardware on the chip. we don't have too many details on this however, nor on this multi-plexing (I haven't heard of it).

This is a serious advantage to any platform. Yes we know that the future will probably yield plenty of dueal core games, however, what about HL2? Doom 3? Fear? Games that aren't dual core. What about many windows applications out there, applications frmo companies who dont want ot write it in dual core? What about spressing both cores easily? Crunching numbers? Server applications? The uses for this technology are ENDLESS. How can anyone see it any other way?
 
I can see this being a good thing on the server side from a licensing perspective also, where licensing is sometimes worked out per CPU core.

If a multi core CPU is seen by software as a single core then in some instances it will make per CPU licensing much cheaper. Some major developers still price their software per core rather than per CPU socket. This could drastically reduce the overall cost of a server, and I'm not talking about hardware.

Hardware is often the cheapest thing when investing in a server. Server software is the killer.
 
soulfly1448 said:
I stated before in another thread that AMD prolly has something in the works to battle Conroe.

Someone start a thread about what AMD's going to do to battle Tukwila and or Nehalm :drool:
One or the other is due out right around or just after K8L, unless Intel doesn't see a need IE no worthy comp. from AMD :(
 
greenmaji said:
Someone start a thread about what AMD's going to do to battle Tukwila and or Nehalm :drool:
One or the other is due out right around or just after K8L, unless Intel doesn't see a need IE no worthy comp. from AMD :(

are you sure it won't be intel needing something to do battle with K8L?

no, and none of us do(well, maybe 1 or 2 of us do, but can't say, lol)

Cheator said:
K8L will be sometime in the Q2 of 2007. This has been stated many times in this section, meaning that quad core from AMD will be out in the first half, not the second. socket 1207 quad cores might take longer since the socket isn't out yet.

Socket F should be out in August, not a firm date, but most likely before the year is up
http://www.theinquirer.net/?article=32188


If RHT and '4x4' are even on the low-end of what we'd expect, we're all in for a treat, the only thing that should be argued is wether or not Intel will answer, let's hope so, gotta keep the competition hot, or we lose

i skipped 754/940/and 939 for amd, and haven't gone Intel of course, because they didn't offer a good price/performance either, these price drops along with new tech is what i'm watching, and will make my move based on that
 
hangten said:
Molester, enjoyed reading this thread. I take it your going to get an AM2 system?

waiting on RHT confirmation, and see if intel's 'core multiplexing' pans out, this new flood of tech leaves alot to be taken in, and i wanna make sure i have a good upgrade path

i would love to know if K8L will be on socket AM2

mainly though, i'm waiting on '4x4', which i'm sure is/will be AM2
 
Molester said:
waiting on RHT confirmation, and see if intel's 'core multiplexing' pans out, this new flood of tech leaves alot to be taken in, and i wanna make sure i have a good upgrade path

i would love to know if K8L will be on socket AM2

mainly though, i'm waiting on '4x4', which i'm sure is/will be AM2

Something that may come into play is that supposidly AMD has been granted a patent for this technology. Now, normally I'm sure intel would probably just implement it into its cpu's (providing the patent is true) and fight it out in court, but with all the bad press it's getting lately from the other AMD lawsuit, it might think twice. They also may find a way to circumvent the patent and still use the technology, it's happened before. 1 thing's for sure, the next year - 18 months will be very interesting.
 
xFlankerx said:
...
4x4 is a bad name anyway. It should be 2x2. Two cores x Two Sockets.


I believe they are refering to the fact that there will be 4 CPU cores and 4 GPU's (quad SLI) on the platform. Hence 4x4.
Someone correct me if I'm wrong on that.
 
Roman79 said:
I believe they are refering to the fact that there will be 4 CPU cores and 4 GPU's (quad SLI) on the platform. Hence 4x4.
Someone correct me if I'm wrong on that.

don't forget about quad-crossfire, and possibly 2 quad-core cpu's :D
 
Back