The current GPU and CPU disadvantages


The current GPU is very powerful and the float point process ability even better than current mainstream CPU products. Why not use these resources to do other process when they are not used for games? If so, will the system hardware performances be strengthened largely?


In fact, AMD has realized it. But as the design thought and structure are different so they can not joint simply. The let us see their advantages and disadvantages.


http://www.hardspell.com/pic/2007/6/14/4750395a-086e-4f57-9ddf-531ac3b88a044501200.jpg

http://http://www.hardspell.com/pic/2007/6/14/ecf69baf-1d25-4c79-8876-ab00e79503f74501200.jpg

http://http://www.hardspell.com/pic/2007/6/14/5f87c4ec-21ba-426a-8027-cb9f1e0bf0734501200.jpg

from:http://www.hardspell.com/english/doc/showcont.asp?news_id=824

really faster faster???

Hi,

This technology from AMD is also known as "Fusion".

Intel will launch this technology with their "Nehalem" microarchitecture, around 2008. More about this here:

http://www.hardwaresecrets.com/article/434

In theory this technology will increase on-board video performance a lot, however we will have to wait and see.

Cheers,
Gabriel.

Ever since the first time theory behind this arctitecture was introduced by AMD i never really cared much about having the GPU integraded with the CPU. Its going to reduce space in which CPU cores could have been placed, the GPU wont have its own dedicated memory storage space and will be leaching from the system memory from what i'm guessing, if so... if not seperated with dedication to the MCH, will also be sharing the same Memory Controller as the CPU line... thats going to bog the bandwidth by 50-50 when requests are on full load to the main memory banks...

*Gabriel please clear this up for me... if this is so i'd like to know*

with the DDR3 modules comming out soon, and the 3.2 ghz FSB for AMD's phenom core's, I doubt bandwidth will be a huge problem, as combines GPU&CPU is a little way off yet, so maybee we can be expecting DDR4 modules by the time its out?
but I think what will happen is that the GPU will get its own set of DIMM socket's

with the DDR3 modules comming out soon, and the 3.2 ghz FSB for AMD's phenom core's, I doubt bandwidth will be a huge problem, as combines GPU&CPU is a little way off yet, so maybee we can be expecting DDR4 modules by the time its out?
but I think what will happen is that the GPU will get its own set of DIMM socket's

Alright i hope so bro, if it does have its own onboard mem slots 4 GPU they may be physically upgradable now as well which could be kool.

Ever since the first time theory behind this arctitecture was introduced by AMD i never really cared much about having the GPU integraded with the CPU.

From my understanding -- I may be wrong -- this idea is more targeted to low-end PCs. I don't think they will be able to embed a high-end GPU into a CPU. The reason why is that if you think about GPUs, their technology changes every 9 months, while for CPUs they only change every 18-24 months. There is a huge gap between the two.

But we will have to wait and see. In my opinion AMD and Intel should have two separated CPU lines in the future, one with embedded GPU, another without.

Cheers,
Gabriel.

From my understanding -- I may be wrong -- this idea is more targeted to low-end PCs. I don't think they will be able to embed a high-end GPU into a CPU. The reason why is that if you think about GPUs, their technology changes every 9 months, while for CPUs they only change every 18-24 months. There is a huge gap between the two.

But we will have to wait and see. In my opinion AMD and Intel should have two separated CPU lines in the future, one with embedded GPU, another without.

Cheers,
Gabriel.

Very true, their would be advancement rate issues at hand. Personally AMD can mess with GPU integration as Intel should be working on 45nm's which for them finally integrate the MCC into the CPU as AMD has had for years now. Not to mention the supposed 80-core CPU Intel had been talkin about which should be on its way in another 4 years or so now, doubt they will produce that one on time with talk of now trying to copy AMD's GPU integration theory. *Don't get me wrong i love Intel, i just have high expectations from them and believe they should put their mind on their own goals*

AMD has always been a step infront of Intel. But for the GPU CPU intergration I think could work. That's the idea of the $100 pc. Intergrating everything. Could be great for entry level or business pc's. Only problem you'd have is heat. But I think IBM has come up with a solution for such kind of thing. Will go search and read up about it.

AMD has always been a step infront of Intel. But for the GPU CPU intergration I think could work. That's the idea of the $100 pc. Intergrating everything. Could be great for entry level or business pc's. Only problem you'd have is heat. But I think IBM has come up with a solution for such kind of thing. Will go search and read up about it.

the complete opposite, Intel has usually been one step ahead of AMD

the complete opposite, Intel has usually been one step ahead of AMD


Uhm, I don't hink so. Go read up a little on the history. Here's just a little proof...

http://www.hardocp.com/images/news/1163691441BxRprUkDWZ_1_15_l.gif

Uhm, I don't hink so. Go read up a little on the history. Here's just a little proof...

http://www.hardocp.com/images/news/1163691441BxRprUkDWZ_1_15_l.gif

just because they invented these technology's doesn't mean that they were better the only major point at which AMD was on top was from the athlon xp's --> dual core X2's the minute the core 2 duo's came out intel was decimating AMD, and before that the p1,2,&3 was almost alway's better intels 286,386& 486 was better than their AMD counterpart.

Hi guys,

This kind of discussion of who is better won't lead us anywhere.

However it is just a matter of studying history. Intel is always in front of AMD and probably always be, since Intel is a far bigger company and has far more money than AMD to rush new releases.

The slide posted here only shows that AMD was the first to use 64-bit extensions and integrated memory controller and yes, this was the only time AMD was technologically ahead. But that is just one time inside a 30+ years of microprocessor history.

History lesson.

Until the 386 processor AMD bought the authorization from Intel to manufacture the CPUs. Meaning that the 386 from AMD was exactly the same CPU as the 386 from Intel and vice-versa (Intel gave AMD the CPU manufacturing masks). In order to give an edge to the consumer, AMD offered CPUs with 40 MHz clock rates, a step Intel never offered (Intel until now works with 33 MHz multiples). E.g. the maximum 386 Intel offered was 386DX-33, AMD on the other hand offered 386DX-40. However this does not mean being ahead, as these CPUs were internally the same, as both were designed by Intel.

The things started to change with the 486. AMD was fed up of paying royalties to Intel and decided to design its own CPUs. So they reversed-engineered the 486 and launched their own 486, which was internally different from Intel's 486. At one hand that wasn't hard as 486 only presented 6 new instructions over the 386, on the other hand 486 integrated the 387 math co-processor inside the CPU -- this was the first time the CPU had an FPU inside. 486 from AMD behaved like the 486 from Intel except that on 486DX4 from Intel it had 16 KB L1 memory cache and 486DX4 from AMD had 8 KB L1 memory cache, thus being slower -- but cheaper. Also, AMD continued to provide clocks also in 40 MHz steps (486DX-40, 486DX2-80 and 486DX4-120), clocks never provided by Intel.

Intel of course was fed up with this and also with Cyrix, that at the time launched a louse CPU called 486DLC which was a relabeled 386DX and decided to use "Pentium" and not 586 for their next CPU line. Using 586 people would copy the name once again, but using a trademark nobody could copy.

Pentium was released on March 22, 1993. AMD's counterpart, K5 (a.k.a. 5K86) was released in March, 1996 -- i.e. 3 years later.

At this time AMD was struggling big time to catch up with Intel. Pentium-133 was released on June, 1995, while its competitor from AMD, K5-PR133 (which worked at 100 MHz, BTW), was released only on October, 1996. That is 18 months later. Not only that. At the same time AMD was launching K5-PR133 Intel was launching the whole Pentium MMX line, including the 200 MHz version.

AMD's answer to Pentium MMX, K6, was only launched April 2nd, 1997, at least here the gap shortened to 6 months. By the way, K6 CPU wasn't designed by AMD and it isn't based on K5, and that explains the success of K6 over K5. K6 was designed by Nextgen and was called NX686. AMD had just bought this small company and relabeled NX686 as K6.

The comparison between Pentium II, Pentium III and Pentium 4 against AMD CPUs is a little bit harder, because starting here the models do not have an exact correspondence. For example, in physical aspect the competitor for Pentium II would be the first Athlon CPU, because both of them were packed inside a cartridge. However the comparison between the two is unfair because Pentium II was available from 233 MHz to 400 MHz while Athlon was available starting at 550 MHz. So we must compare competitors of this time based on their clock speeds.

Pentium II was released May 7th, 1997. K6-2 was released May, 1998. One year later.

Pentium III was released February 26th, 1999. Athlon (cartridge) was released August, 1999. 6 months later.

Pentium III using socket 370 was released October 25th, 1999. Athlon using socket 462 was released June, 2000. 8 months later.

Pentium 4 using socket 423 was released January 3rd, 2000. At this time AMD competitor was still Athlon/Athlon XP.

Pentium 4 using socket 478 was released September, 2000. At this time AMD competitor was still Athlon/Athlon XP.

Athlon 64 was released only September 23rd, 2003. 3 years after socket 478 Pentium 4.

For dual-core CPUs both companies released their first models on the same month, with Pentium D launch on May 26th, 2005 and Athlon 64 X2 launch on May 31st, 2005.

Well, I think I made my point here. But don't get me wrong. I am not saying that Intel is better than AMD. I am only answering to the statement that AMD launches things before Intel, which is untrue.

Cheers,
Gabriel.

First of all, I never said AMD launches something before Intel. I said that AMD has always been a step ahead of Intel.
AMD did buy the authorization from Intel, but it doesn't mean Intel produced their chips. AMD and Intel was sharing their technologies with each other, until the 386 (not 486) processor where Intel stopped. That was in the late 80's. AMD launched the Am386 processor in 1991.

" By the end of the 1980s, one company controlled the major PC semiconductor architecture. Customers had only one source — and little choice — until AMD restored competition to the marketplace in 1991."

AMD only launched it's Am486 processor in 1993 (long after the battle between the companies were over), and Intel launched it's i486 processor in 1992.

If Cyrix is mentioned, then Harris and NEC must also be mentioned. At that stage the NEC chip was the fastest chip available. But the fight at that stage never was between Intel and those three companies. It was between AMD and Intel for eight years. (History 101: AMD and Intel got the 'block' processor idea from Harris.)

The Intel Pentium was indeed launched in 1993, but it's counterpart isn't the 5K86, but the Am486, wich was also launched later the same year by AMD. This was also the year both Intel and AMD started using 'micron' technology.

In 1994 Intel found a flaw in their processors, affected millions of people around the globe (Wich they say is "minimal"...).

Under my understanding, the Pentium-133 is part of the MMX technology Intel released, because I had one. The Pentium MMX technology was released in 1997, aswell as the Petium II. So it wasn't in 1995 or 1996.

In 1996 AMD "PARTNERED" with NexGen, creating the K6 processor family...

"AMD partnered with NexGen to create the AMD-K6® microprocessor family, designed to restore competition to the supply chain and offer an outstanding solution for desktop computing."

In the physical aspect, you can't compare the AMD Athlon to the pentium II, because the Pentium was released two years before. Release dates of the Pentium III and Athlon was both in 1999. So rather compare the Athlon (even if in a cartridge) to the Pentium III or Pentium 4.

Intel may have been feeling minipulated by the fact that Athlon had been faster than Pentium, and decided to launch the Pentium 4 in 2000, wich didn't bother AMD, because AMD had launched their AMD-760 chipset with double data rate(DDR) memory technology for their processors, wich speeded up their processors more than before.

"October: AMD announces AMD-760™ chipset, featuring Double Data Rate (DDR) memory"

The AMD AthlonXP was only released in 2001 (not 2000), but before the AthlonXP, they launched AthlonMP wich were their first multiprocessing platform. The AthlonXP was built on QuantiSpeed architecture wich was designed for pure application performance, and also featured Enhanced 3DNow!

"In fact, 2001 was the year that the PC industry became aware that application performance and the overall user experience was a far more reliable processor performance metric than megahertz ever could be."

In 2002 both AMD and Intel starts using 0.13 Micron technology.

In 2003 AMD launched the Athlon 64, wich was built on AMD64 technology, and was a 64-bit processor. Something Intel couldn't provide untill the Core 2 Duo, three years after AMD.

In 2004 AMD had launched the world's first dual core processor, as well as the Sempron family. Where Intel had only launched it in 2006, the Core 2 Duo. Two years after AMD.

And then in 2006 AMD and ATI joined forces.

Hi,

First of all, I never said AMD launches something before Intel. I said that AMD has always been a step ahead of Intel.

Which, by the summary I wrote, can be easily seen that this affirmation is simply untrue. Like I said, AMD has got it right by embedding the memory controller inside the CPU and the 64-bit extensions, but like I said, this was just two points inside 30+ years of microprocessor history.

I just want to correct some factual errors:

The Intel Pentium was indeed launched in 1993, but it's counterpart isn't the 5K86, but the Am486, wich was also launched later the same year by AMD.

Nope. You are comparing apples to bananas. The competitor to Pentium (a 5th generation CPU) couldn't be 486 (a 4th generation CPU). It was in fact K5. Both Pentium and K5 used the same CPU socket.

Under my understanding, the Pentium-133 is part of the MMX technology Intel released, because I had one. The Pentium MMX technology was released in 1997, And then in 2006 AMD and ATI joined forcesaswell as the Petium II. So it wasn't in 1995 or 1996.

Nope. Pentium-133 didn't have MMX technology and was released June 12th, 1995:

http://download.intel.com/museum/research/arc_collect/timeline/TimelineDateSort7_05.pdf

In 1996 AMD "PARTNERED" with NexGen, creating the K6 processor family...

Nope. AMD bought NexGen in 1996:
http://en.wikipedia.org/wiki/NexGen

"AMD aquired NexGen, and turned the Nx686 design into what is now known as the AMD K6 processor."
http://www.sandpile.org/impl/nx6.htm

And then in 2006 AMD and ATI joined forces

It wasn't a merger but a buyout. AMD bought ATI.

As for your other comments, the ones that are not marked with concrete data is pure speculation and/or propaganda. E.g. "In fact, 2001 was the year that the PC industry became aware that application performance and the overall user experience was a far more reliable processor performance metric than megahertz ever could be." this statement isn't a factual data but pure propaganda, as this cannot be proved.

I'd also like to add that yes AMD launched DDR support before Intel, because at the Intel was betting on Rambus technology that never took off.

Cheers,
Gabriel.

There was three points, the memory controller, the 64-bit extensions and the dual core.

Again, even if comparing apples to banannas, the Pentium and Am486 was launched the same year. The may have not been the same generation CPU's, or used the same socket, but used the same technologies.

I may have been wrong about the Pentium-133 being MMX technology, but then again you can compare it to the Am5x86 running at 133MHz wich was an enhanced Am486, and again, using the same technology.

AMD may have bought NexGen, but NexGen didn't design the entire K6. The Nx686 was only the basis, wich means AMD added a few stuff.

I compared the history from AMD's site to Intel's site, and both was correct. I took information straight from the sites and quoted it. And AMD have been using technology to run at a low MHz and improve application performance, wich Intel only started doing with the Core 2 Duo's. Compare an AMD Athlon 64 2800+ running at 1800MHz against Pentium 4HT running at 3.0GHz. The Pentium 4's MHz goes higher, just to prove my answer. They only started running low with the Core 2 Duo's.
And it can be proved by benchmarkes.

And lastly, why are we argueing about what and who is better? It's for the end user to decide wich is better. I can just say that I'm an AMD fan, and will always be loyal to them.

billy_bob