PDA

View Full Version : First hardware action of the new year


Fozzik
12-31-2005, 09:21 AM
Well, the holidays are flying past as usual. Here we are on the last day of 2005, and 2006 looks to be another interesting one when it comes to computer hardware. We're already hearing about the first new releases of the year, and they aren't far away.

First up, Intel has officially announced their first desktop processor built using the 65nm process. It turns out Intel will also be dropping the "Intel Inside" slogan they have been using for so long, in favor of the new and much more exciting *snore* "Leap Ahead". We will be seeing the old logos, slogan, and processor naming conventions for a bit longer...and sadly the new doesn't seem to be the least bit less confusing than the current. It'll be different, though.

So, Intel's first release of 2006 is the Pentium Extreme Edition 955. Using 65nm traces allows for a bump in clock speed up to 3.4GHz, and for a larger level 2 cache - 2MB per core. This is still the exact same Netburst architecture as the 600-series Pentium 4's...but with two cores on the substrate. The shrink from 90nm traces down to 65nm does reduce power consumption a bit on a per-core basis, but not enough to bring it below the Athlons from AMD.

This chip, due in no small part to the fact that the Extreme Edition cores are the only dual-cores from Intel that have HyperThreading turned on, performs pretty well. It is able to best an Athlon X2 4800+ in some benchmarks, and is much more competitive with its rival in every area. There is of course the exception...games. This new flagship from Intel does pick up some ground in game performance verses AMD's fastest, but not nearly enough, and remember that AMD has a new dual-core FX-60 on the way very soon.

The Pentium EE 955 is a step in the right direction, since it seems to show that they are able to make the transition to 65nm without too much trouble. It remains to be seen how large-scale production will go with this new process, the Extreme Editions are very low volume parts. This chip will most likely be over $1000, so they probably won't be shipping a huge number of them. With faster gaming performance still squarely in the hands of AMD, and the introduction of Intel's new architecture only about 6 months away...this chip isn't one I would recommend.

On the graphics front, ATI and Nvidia are speeding towards their next releases: the G71-series from Nvidia, and the R580 (X1900) from ATI. It appears that each company has a different picture of where game performance is needed most. ATI is going very heavy on the shader engines, 48 of them in the new chip while still only having 16 pixel pipelines. Nvidia on the other hand will likely have 32 pixel pipelines in their new chip, with dramatically fewer shader engines.

This means ATI is banking on games wanting to do a LOT more work on each pixel, and Nvidia is taking the brute force approach...with a much larger number of pixels getting spit out each clock cycle. It will be interesting to see how things shake out, but as usual it will depend on the individual game and settings. With Nvidia now making the move to 90nm (ATI is already there with the X1800), we should see much higher clock speeds out of both companies, and a good hop in performance over the current top end cards.

It's looking like both new chips will be available within the next 2 months, so we don't have long to wait to see how things shake out. After ATI's many problems in 2005, let's hope they can execute well out of the gate this year. On paper it looks like a very interesting and competitive race going forward, and that's the best thing for everybody.

Belisarius
12-31-2005, 09:58 AM
Thanks for the heads up Fozz. Look forward to the battle betwix ATI & nVidia. Here's hoping ATI can get its stuff in one sock.

Havelock
12-31-2005, 12:22 PM
I wonder if we'll see a new Crossfire mastercard shortly after the introduction of the X1900.

Sergo
12-31-2005, 12:55 PM
Always enjoy your reports, Fozzik. I'll be a more and more frequent reader here as I inch towards my Vanguard computer. I'm considering building one this time, but only if I could save more than 50 bucks. Gotta bunch of research to do on newegg vs. abspc to see how the prices add up.

Fozzik
12-31-2005, 12:58 PM
I wonder if we'll see a new Crossfire mastercard shortly after the introduction of the X1900.

Well, from what I've read, ATI will be completely replacing the X1800 series... there will be a new XL, XT, and Crossfire edition...as well as a new All-In-Wonder card. Each will have the new chip running at various clock speeds, probably with various amounts of hardware turned on.

I really hope this isn't true, because of just how silly it looks, but I've seen that ATI may be yet again changing the naming letters on their cards. The new X1900's may be named like this -

X1800XL replaced by X1900XT
X1800XT replaced by X1900XTX

You just never really can have enough X's in the name of your product, as far as marketing departments are concerned. If this new naming convention is true, it's going to get very confusing for a while, with people talking about the "XT" and meaning the "XTX", or talking about the "XT" and meaning the "XL", or the other way around.

This is of course carefully orchestrated by ATI, because as everyone knows, the more confusing and nonesensical the names of your products are, the more likely you are to trick some helpless computer n00b into buying something they didn't want. Why a marketing department would want to trick people into buying things they don't want, I don't know...since it seems that someone will probably end up accidentally buying a cheaper card just as often as they accidentally buy a more expensive one.

I guess the most likely reason is that hardware manufacturers are trying to plan ahead, for those times when they have some real stinker products in their lineup. If the naming convention is moronic enough, they are very likely to sell more of the crappy products, because people without enough knowledge will mistake them for the good ones. Woe to us all.

The fist time a product comes out for the computer that is called "XXXXXXXX", I will become violently ill. Of course the competing company will be able to easily defeat that grand and powerful name by calling their new product "X^infinity"...which, when it occurs, will probably just cause the world to fold in on itself immediately, wiping out all existance as we know it. At least we can rest easy in the fact that science shows us it will be the marketing people who die last, and will get squeezed really uncomfortably for many years before it all goes black.

Now I'm rambling.

Sergo
12-31-2005, 01:28 PM
Perhaps it's some sort of code in Roman numerals. A secret message from the Illuminati or something.

Havelock
12-31-2005, 02:50 PM
Fozzik, it's the X ^ X that will truly be the uber card. Infinity is for sissies compared to the might X. :p

Fozzik
12-31-2005, 05:01 PM
X_x <----dead

Merwenn
12-31-2005, 07:25 PM
With the release date of Vanguard relatively far from this point, there's no reason not to wait until it has gone gold before building or upgrading a computer. Technology changes so much within even 6-8 months.

Sergo
01-01-2006, 10:47 AM
Relatively far? My good man, it's blessedly near for those of us who have been following it since 2003. You're quite right that it's too early to buy, but I better start learning now. I bought my last machine 2 years ago, and it already seems like everything has changed.

For instance, what are these numbers that go with memory? Like 5-3-3-2? In bridge, that's a hand with 5 spades, 3 hearts...

Traldan
01-01-2006, 11:15 AM
I'll be hopefully getting one of the X1900XXXXX's or whatever they are. :D

(Yeah, I'm a shameless ATI fanboy)

I'll let ya know about performance. We can compare X^X and X^Infintiy series cards. :)

Breogan
01-01-2006, 11:39 AM
For instance, what are these numbers that go with memory? Like 5-3-3-2? In bridge, that's a hand with 5 spades, 3 hearts...
These are the memory timings and they refer to the latency the memory suffers performing certain operations at a given speed. The lower these numbers are, the better (which means much more expensive).

Example: I've got a 1Gb stick which has 3-3-3-8 timmings at DDR400 by default, but I'm running it at 2.5-3-3-7 (you can specify the parameters manually in the BIOS and depending on the module, it might work or not).

The difference between a regular module and a low-latency one could be as high as 100$ depending on brand and timings, but the improvement a low latency memory offers is quite debatable, to say the least, and I believe the money is better spent buying a better video card or processor.

Fozzik
01-01-2006, 11:46 AM
Relatively far? My good man, it's blessedly near for those of us who have been following it since 2003. You're quite right that it's too early to buy, but I better start learning now. I bought my last machine 2 years ago, and it already seems like everything has changed.

For instance, what are these numbers that go with memory? Like 5-3-3-2? In bridge, that's a hand with 5 spades, 3 hearts...

Breogan hit the highlights, but if you are interested in a bunch of nerdy detail, here a re-post of something I wrote up on the forums quite a while ago -

the way memory is accessed is that the system requests the data from a certain address on the DRAM. The DRAM gets the request and the address, then goes looking for the data. The memory is laid out in sections called banks. Each bank is like a spreadsheet with rows and columns. The correct row is activated, and then the correct column is powered up, read, and the data is sent out to the system.

So, to describe the timing numbers, all of them are expressed in terms of memory clock cycles.

The four numbers you usually see stand for:

CAS - Column access strobe...talked about this one, it's the time to switch from one column to another. This is one of the most important, because the memory will be switching between columns almost every time a request comes in.

tRCD - Which stands for time of RAS to CAS delay. This is the amount of time between when a row is activated and when a column can be activated.

tRP - time for RAS precharge. This is the time it takes to flush an active row, before you can activate a new row.

tRAS - This is the minimum amount of time that a row must remain active before you can flush it and get a different one.

Then there was command rate, probably the most important along with CAS latency, because command rate is the number of clock cycles the memory takes before it starts getting to the business of finding your data and sending it out.

So, the numbers are generally written in that order -
CAS-tRCD-tRP-tRAS CMD
or
2-3-2-6 1T

You have to be careful about changing timings in the BIOS, and generally I like it when I can just run them at the settings they are designed for and leave them alone. There are dependencies...for instance you don't want your rows deactivating themselves before the column can be brought up and read (which is why tRAS is generally around double the tRCD number, which I think answers your question about why the last number is highest).

In terms of the whole process, here's how the numbers fit in.

1) Request is recieved (CMD)
2) Row is checked, if correct row is active, then column is brought up and read. (CAS)
3) If correct row is not active, wait (tRAS) then flush old row (tRP) and activate the correct row (tRCD) then go to step 2.