Intel Core 2 Quadro review by Tom's Hardware

[Magnus]

I don't know how AMD will counter this new development. They will have to wait with their 4x4 chips by second half of 2007. I think AMD will have to hurry it up or they will be sorely behind.

[Fozzik]

The first quad-core from Intel are not true quad-core...they are just two dual-core processors stuck on the same package and communicating through the front side bus, just like their first dual-core CPUs were. Unless something has changed, the first quad-core from Intel are really not that big of a deal. They like to be first just so they can say they were...but it will be a while before either company has a true quad-core CPU on the market (I believe AMD will be first), and it will be even longer before software catches up and really makes a quad-core worth-while for a typical desktop PC.

I guess if you are the type who does a lot of work with applications which will take advantage of multiple threads, it could be a performance boost... for most games, they still aren't really making use of dual-core all that much.

[Magnus]

Yes, I'm aware of this fact. What I was saying, it will hurt AMD financially.

[AgentJ]

Interesting article, read it myself. Truthfully tho, im not the least bit interested in quad-cores yet, of any type. :confused:

Since as said prev, most games dont even use dual-core tech yet*.

[Fozzik]

I actually doubt it will hurt AMD at all. The market will be very small for quad-core at first, and I think the deal between AMD and Dell (and the AMD/ATI deal) far outweigh any trouble they might see from Intel's Kentsfield.

Just my opinion, though. It's hard to tell what the market will end up interested in. I know Dell is interested in AMD's 4x4 platform for their XPS line, and it will be fun to watch what starts coming out of the AMD/ATI combined engineering teams. AMD is without a doubt behind Intel architecturally right now on the desktop. I think they'll be ok, though. Intel was behind pretty significantly for almost 3 years before they shipped Conroe...and most consumers probably never even knew it. :)

[Gargen]

What I want to know is, what will they call it when 8 core CPUs come out? I'm hoping to have a Octo Core CPU myself some day! :D

[akherat]

AMD is without a doubt behind Intel architecturally right now on the desktop.

That's a very debatable statement. From a design architecture standpoint, AMD is still (IMHO) superior. Core 2 is definitely more competitive, but in the design realm, Intel still has some catching up to do.

[darkarius]

That's a very debatable statement. From a design architecture standpoint, AMD is still (IMHO) superior. Core 2 is definitely more competitive, but in the design realm, Intel still has some catching up to do.

I an a AMD fan too but you have to admit that the Conroe is simply a better chip and par for par kicks the mess out of AMD's current gen chip. However while Intel currently has the edge on tech I think that AMD has a vast advantage on how the market flows and how to cater to it's audience.

Also I am sure AMD has something up it's sleeve to cure the conroe infection that plagues us today. This seesaw match between AMD and Intel will be going on for some time unless one discovers some secret tech (and can keep the tech secret) that the other can not fathome or come up with an alternative to it, Kinda like the Atom Bomb :pirate: .

[Gargen]

That's a very debatable statement. From a design architecture standpoint, AMD is still (IMHO) superior. Core 2 is definitely more competitive, but in the design realm, Intel still has some catching up to do.

Other than the on die memory controller, what advantages does ADM CPU architecture have over Intel CPU architecture? I can think of several Intel advantages that easily out do the memory controller when put together.

I'm not trying to be argumentative, I'm actually just curious. The actual CPU architecture stuff is something that I'm still learning. Soo... debate me. :D

[akherat]

Other than the on die memory controller, what advantages does ADM CPU architecture have over Intel CPU architecture? I can think of several Intel advantages that easily out do the memory controller when put together.

I'm not trying to be argumentative, I'm actually just curious. The actual CPU architecture stuff is something that I'm still learning. Soo... debate me. :D

Keeping in mind we are talking strictly about design here, not measured performance (you can still drive a nail with a sledgehammer, after all).

The single biggest design difference and where I think AMD is going to have an advantage going forward is HyperTransport linking. I don't know that I can adequately describe what it is and how it works without using a bunch of graphics, but suffice to say it allows those things connected directly to the CPU to talk directly to the CPU without having to pass over the FSB, and it does it with more available bandwidth.

Outside of that, it's mostly efficiency of design - Intel may very well be faster at the finish line with Conroe, but the level of effort to get there is where AMD is better by design. If you're looking for something more statistical in nature, look at how well Intel fares at handling highly threaded applications like databases or web servers versus an AMD. Thinks like number of and distance between gates in the CPU, how power is utilized (not a reference to power draw, but what is done with the power that is drawn), and communication with other components (AMD supports NUMA with 2 processors, Intel requires 4 - assuming the OS is NUMA aware). An AMD CPU does a better job of getting the most out of what it has available than an Intel CPU.

I realize that I'm not really clarifying my position too much, but I think if you're really interested, check out HyperTransport linking and sort of poke around those places that describe it, and you'll find what I'm (trying) to talk about.

[Fozzik]

AMD's onboard memory controller is nice...but at the moment it isn't providing them with any advantage. Once they improve and update the rest of their design to be more in line (or surpassing) the Core 2 Duo... the onboard memory controller should again put them on top.

AMD's platform, especially in the server space, is by far better. That doesn't really count as part of the CPU though. I know in an overall sense, it does...because you can't really separate the platform from the CPU. In a strict sense though...even with the worse platform, even with the higher latency of an off-chip memory controller...Intel has a clear performance lead. That means their CPU architecture is pretty darn good.

AMD will answer, I'm sure. Competition is a great thing for innovation. :)

[akherat]

There really isn't any debate that Intel has the performance lead on the desktop side of things for the time being, irrespective of design differences. Server performance tends to be measured a little differently, and so it doesn't always translate straight across. My spending my professional life in the server world sort of bleeds into all my IT discussions, and I'd agree with Fozzik on the server side of things 100% in almost any application (AMD currently has a superior platform).

I guess what I was really trying to get across, is even though in the computer world faster generally means better, how you get to that speed can be just as important.

As with a car, you can't be the worlds fastest and also be poorly designed. Intel makes a good product. But if you pull out the tale of the tape, IMHO, even though it may not be as fast at the finish line, AMD has a better designed, better engineered product.

[Gargen]

You'll get no argument from me that the AMD platform rocks the socks off the Intel platform. I generally consider things like hypertransport and 4x4 to be more of a platform thing (ie not part of the CPU "archiecture") and aren't that big of a deal on the desktop (keep in mind the quote you used in your first post).

Po-tay-to po-tah-to, I think we agree... well except for thisOutside of that, it's mostly efficiency of design - Intel may very well be faster at the finish line with Conroe, but the level of effort to get there is where AMD is better by design.but I don't think either of us is going to convince the other or really knows exactly what we're talking about at the level needed anyway.

[akherat]

You'll get no argument from me that the AMD platform rocks the socks off the Intel platform. I generally consider things like hypertransport and 4x4 to be more of a platform thing (ie not part of the CPU "archiecture") and aren't that big of a deal on the desktop (keep in mind the quote you used in your first post).

Po-tay-to po-tah-to, I think we agree... well except for thisbut I don't think either of us is going to convince the other or really knows exactly what we're talking about at the level needed anyway.

I think I see more of what you're talking about in the differences between CPU and platform. So, with that understood, of the more significant differences is in gating - now there may be an argument on the semantics of what I'm referring to here, but I'll try to clarify -

gates are the points within the transistor fields on the CPU where power is applied. The further you can go down a path without having to 'refresh' the power, the better you can do with transistor density and stuff like that. AMD gating designs allow for significantly longer chains (admittedly, I haven't seen the specifics of the conroe design) - this is one of the reasons that historically, AMD's overclocked much better than Intels. Fewer gates means you can better account for thermal dynamics. Conroes look to overclock pretty well, so Intel may have done some work here.

There are similar points that could be made - as I said before, Intel has a good design; AMD's design I feel is superior because they can do just as much work without as much effort.

**edit**
Incidentally, HyperTransport is a huge deal on the desktop, because it offloads the front side bus - additional cores dramatically increases the amount of FSB traffic, something I believe Intel users are going to discover as more applications begin taking advantage of additional available execution threads.