Official Conroe information thread

[pointreyes]

Originally Posted by shaihulud

... and I understand the moniker change to AMD64, which is not bad, once EMT64 came around. At least AMD64 was not 64Today!

It's EM64T, not EMT64.
I'm able to remember it a certain way but I would sound like a AMD fanboy.
64Today! - Glad AMD did not decide to use that one.

BTW: Isn't SSE4 starting with the Conroe?

[Sandog]

Here's some nice Conroe benchmarks from a review.

http://techreport.com/etc/2006q1/conroe/index.x?pg=2

Strange seeing Intel on top isn't it? Especially in popular games.

Note: At the beginning of this review they said they had to modify the voltage regulator to get the voltage down on the D975XBX in order to run the Conroe. This is what rev.304 did to the board I believe. It added/changed resistors that adjust the CPU voltage regulator.

When Intel introduced the Conroe at IDF and did a system comparison to AMD's chip a lot of people assumed a setup in favor of Intel. Really if anything that was a conservative test. With all the review sites and enthusiasts out there waiting to scrutinize the Conroe worse than NASA scrutinizes the space shuttle, I doubt Intel would have fudged any settings. The Conroe that appears in Q3 will be looking as good if not better than what Intel showed at the IDF due to final revisions, a new chipset and chipset drivers, and new BIOS's down the road. AMD is quickly honing their already good chips and it will be a close finish but if you are an Intel fan this is very good news no matter what AMD does.

[Sandog]

Originally Posted by pointreyes

It's EM64T, not EMT64.
I'm able to remember it a certain way but I would sound like a AMD fanboy.
64Today! - Glad AMD did not decide to use that one.

OK I give, what to AMDers refer to it as? I can take it.

BTW: Isn't SSE4 starting with the Conroe?

I could speculate on that if you want me to. By the way, down here speculate just means you won't be coming home on time.

[shaihulud]

It's EM64T, not EMT64

Damn fingers....corrected

Yes, SSE4 starts with NGA.

OK I give, what to AMDers refer to it as? I can take it.

It is labeld as AMD64, now http://www.amd.com/us-en/Processors/..._13278,00.html Or did you think I was fishing for something with the naming statements? I was not really trying to. I just do not particularly like the AMD naming scheme. I think it can be much better.

[Sandog]

Originally Posted by shaihulud

Damn fingers....corrected

You just need a new revision keyboard with less erroneous data like me

http://www.theinquirer.net/?article=31007

"But what truly seals the deal on waiting is the price of those Conroe chips being priced very nicely. Almost too low that I and many others cant wait for them. The 65nm process allows Intel to mark down chips as well not entirely because of poor demand but it does make sense to try and liquidate inventory (90nm) when you have a new generation of 65nm parts coming soon. "

[pointreyes]

Originally Posted by Sandog

OK I give, what to AMDers refer to it as? I can take it.

I think it's only me that came up with this meaning for EM64T but always afraid people might think I'm being an AMD fanboy. I'm being more of a person that is just tired of marketing.

Elaborate Marketing [using] 64 Techniques.

I guess AMD64 could mean:
Advanced Marketing Delineation [for] 64 [-bit]

[Sandog]

Originally Posted by pointreyes

I think it's only me that came up with this meaning for EM64T but always afraid people might think I'm being an AMD fanboy. I'm being more of a person that is just tired of marketing.

Elaborate Marketing [using] 64 Techniques.

I guess AMD64 could mean:
Advanced Marketing Delineation [for] 64 [-bit]

Ahh, something along the line of Fix Or Repair Daily (Ford). I was trying to figure out what rhymed with EM64T (extended memory 64bit technology).

[shaihulud]

It seems that the economic battle has started. This is even though it has not “started.” Hmnnnn

http://www.amdzone.com/

[Intel_User]

I was always careful never to say half which I never did, I have heard talk before of what the difference was but of course it is not half. Mathematically, there is a 38% reduction in size between the two dies. Intel also went from 12" to 16" wafers I believe which is a 33% increase in diameter.

Theoretically it is half. Let me explain. 65nm is 65/90=0.7222..., but its only one side. So we have to square the 0.7222... result, which makes it... 0.5216. Its approximately 37% reduction in one side, and 37% on another side. Of course its up for Intel and others to follow exactly, so it may not be 0.52, but higher.

Here are Intel's SRAM 6T cell sizes(6T means 6 transistors, and cell is sort of like a "group" of those transistors)

130nm 1st: 2.09um
130nm 2nd: 2.00um
90nm: 1um
65nm: 0.58um

For the wafers, Intel went from 8" to 12". 8" is 8 inches, so that's approximately 200mm in other terms. 12" is 12 inches which is approximately 300mm. Looking at the numbers alone, that's only 50% increase, however you gotta remember to square them after. 200mm wafer is really 200x200 or 40000mm2, and 300mm is really 300x300 or 90000mm2. So the actual increase is 90000/40000=2.25x

16" is 400mm, which is being researched by Intel right now.

Note: At the beginning of this review they said they had to modify the voltage regulator to get the voltage down on the D975XBX in order to run the Conroe. This is what rev.304 did to the board I believe. It added/changed resistors that adjust the CPU voltage regulator.

Despite that some sites report that Conroe system had sub-par I/O drivers. Don't know why. They mentioned the Conroe system loading apps slow. I can't find the site but I remember it clearly as I found it surprising(on a 975XBX??). Maybe Conroe still has more features to be shown.

Intel provides design guides with their chipsets that explain how to lay everything out. What a gimme. If every manufacturer follows the circuit layout that Intel provides they would have a board similar to an Intel board and the least problems. I know ASUS does a good job. It's called support.

problem with them.

Asus does a very good job on their boards, true. Back then on the Pentium III/Pentium 4(early) days, I had an Asus board. I had a problem so I went to a support site. I had to register, but I couldn't because the page wouldn't load at all. I tried it day after days before giving up. That sort of ruined my views on Asus. So I trust Intel now.

Yea, Intel really gives lots of info on their products. When I want to talk to people about TDP on Nvidia/SiS/Via chipsets, I can't, since Nvidia/SiS/Via has crappy documents, that show marketing information. Intel had better documents 15 years ago than those companies do now.

I really think the best time to upgrade will be after Vista comes out. Everything is going to change with that operating system, software and hardware. Intels 965 chipsets were designed for Conroe and Vista but until the OS is here and you can run it fully you will never know. Especially when MS is changing 60% of the code. Intel probably started the 965 chipset mid 2005. About the only safe hardware to buy 'till then is a hard drive lol.

Vista's only problem for support on hardware is the video card. The OS does provide extra features like NAND flash hard drives, but they are really extra. To have full Vista support you need graphics hardware which supports Aero, which is supposedly DX10, WGF 2.0. If that's true then NONE of the current graphics cards are Vista capable. The support for 965 chipsets regarding Vista is because the integrated graphics is rumored to have FULL support for Aero graphics.

[Intel_User]

Let's add some die size numbers.

Conroe, the 4MB version is apparently around 140mm2 die. From this pic: http://www.anandtech.com/tradeshows/...spx?i=2203&p=2

It looks like 2MB cache in Yonah takes ~35mm2. Conroe is said to be using similar/same L2 cache, so we can expect same size per MB. Yonah is 90.3mm2 die. So 2MB Conroe, would be 105mm2 die. Yonah's die size for two cores are 90-35=55mm2. Conroe's core size would be 140-70=70mm2.

So, the core for Core(lol) has grown 20-30% over Yonah, which is remarkable considering the advancements Core has made over Yonah.

Yonah, is very efficient too. Yonah has 151.3 million transistors while Dothan has 140 million transistors. All the enhancements that went to Yonah, like dual core, enhanced media performance, logics required for shared cache, 3x decoder bandwidth only cost 11.3 million extra transistors. I am sure there are many that doubt the core for Dothan is that small. It has to do with the fact they optimized the design for die space, power consumption.

[h20-ski]

Although one thing to about is that the caches take up the most die space. One interesting point brought up by The Inquirer is that games are now for the most part cpu limited, and with the games going SMP dual core. And their thoughts is that a SMP computer would best suit these games.

[Intel_User]

One interesting point brought up by The Inquirer is that games are now for the most part cpu limited, and with the games going SMP dual core.

The thing about Inquirer, is except some very few authors, they have very wild imaginations. The most recent example was when they were predicting VLIW based architecture for Conroe. Which is ridiculous, considering all the other rumors. Why would Intel go for another VLIW, when they might as well go for their Itanium?? As far as architecture goes, EPIC better performance wise than VLIW.

If you look at benchmarks like FEAR, and CoD, they are very graphics limited. It used to be back in the Pentium II/III days, CPU mattered more. But now with almost everything on the graphics, with T&L and all the stuff, CPU don't really matter anymore.

[Intel_User]

Although one thing to about is that the caches take up the most die space.

Not 100% true. Caches are most efficient in die size per transistor, and also very power efficient per die size. If you look at Prescott, 1MB L2 cache takes up only 16mm2 out of the total 112mm2 die. If you look at Yonah, or Conroe die pics, caches take up less space than rest of the core.

But even if they are efficient, having so much will counter that. Remember, 2MB L2cache on Dothan takes over 70% of the transistors but take little over half the die size.

[h20-ski]

I am looking for the benchmarks of FEAR and CoD 2, and while performance gained by SMP right now is for the most part non existant, a game built from the ground up to be SMP enabled, should perform better, than a patched game. The thing is that unless you are running a massive resolution, even with all the eye candy turned on, for the most part the CPU is the limit of your frame rates. If you look at Valves' stastics the majority of people are running 1280 X 1024, not the resolutions that are GPU limited.

[Sandog]

Good facts guys, were going to all be well prepared to accept Conroe when it comes out from reading this thread. This is what makes it so worthwhile to come to ABXZone.