HTPC AMD CPU - 90nm X2 5200+ EE or 65nm x2 5000+?

[NDLunchbox]

Hi,
I'm rebuilding my HTPC and want to move to an AM2 proc. I was sold on the 65nm Brisbane 5000+ because it was a 65W part and cooler parts = less fan noise.

But then I did some research and found out that:
A) Brisbanes are actually a little slower than their 90nm counterparts.
and
B) AMD has Windsor 90nm "Energy Efficient" parts that are rated at the same wattage and clock with twice the L2 cache (Like the X2 5200+).

So my questions are, since they cost about the same and have the same wattage - should I resist the "latest and greatest" (i.e. 65nm CPUs) and go with the Energy Efficient 90nm part with more cache? Will the larger cache translate into improved performance in media apps (ripping DVDs, transcoding, etc.)?

No matter what I'm going to wait for the April 9 price cuts, but I wanted to get some input on weather the 65nm parts had any advantage over the 90nm EEs.

Thanks,
-LB

 

[ajfink]

No real reason to go for the slower piece in this case.

 

[NDLunchbox]

PS - I have no intention of overclocking (I have read the Brisbane can be ramped up quite a bit).

 

[NDLunchbox]

Interesting... if nothing more compelling is announced by mid April, I suppose I will go with the 5200+ EE.

One concern though... I think I read that AMD is having yield issues with the larger cache chips and will not offer 2MB L2s in non FX parts. Does this mean that 5200+ EE SKU will be hard to come by soon? It took me a while to find it as it is...

 

[ajfink]

Why are you people voting for the 65nm part? If you have a reason, explain it, because it doesn't make sense to me - for all the reasons Jack as explained.

I'd actually say 65nm is at least a decent overclocker, though, seeing as even the 1.9Ghz parts are routinely known to hit ~3Ghz on good air or better. 90nm has higher clocking overall, though, but I don't think as easily to 3Ghz as 65nm is doing.

 

[ajfink]

It's a decent overclocker, I should have been more clear --- percentage wise, it does a good job.... the ceiling clocks though are a bit discouraging... that will go up over time, but we may or may not see it on Brisbane. Barcelona is a big question mark, it may or may not, depends on how much complexity went into the logic.

However, if Barcelona does not OC as well as Brisbane, this is NOT an indication that they took a step back.... it is a convoluted and complexed interaction between process technology and architectural features at the circuit level. So we may not end up knowing how much over time 65 nm improved directly, as it may be masked by the architectural features.

But it will improve.... this is staple AMD.

Jack

Oh I won't argue that the ceiling is a bit lower than 90nm - I'm just saying that nearly all the chips reach that limit with comparative ease. X2 3600+ all over the place hitting 2.8Ghz on stock volts and such...you'd never get a 1.9Ghz 90nm part to do that.

It will be interesting to see how Barcelona overclocks. Bit more mature process...different architecture...who knows. There are apparently 2.9Ghz dual-core parts planned from the beginning, so I would imagine getting into the lower 3Ghz-range should be rather doable.

We'll see.

 

[udontnome]

Jeez JJ, that dates me... hadn't seen that formula written out since '81, lol. what's with the banner at the top, which would u choose... x2... I wouldn't choose an x2 for my sock drawer.

 

[gOJDO]

90nm 5200+

 

[ajfink]

If I had to guess --- I expect Barcelona to OC ceiling lower than Brisbane would .... simply because Barcelona is a much more complex beast.

A few reasons .... you see me argue simple transistor parametrics, which is true, my arguments are correct. What you will also see me discuss (on occasion anyway) is that the data for a single transistor alone is not the determinant for the top clock of a circuit.

Each transistor is more or less a capacitor... that is the way the transistor works... it is a capacitor suspended over a triode. Anyway, the total capacitance is the determinant of the speed at which the circuit can clock.... the more transistors within the functional block, the more total capacitance of the circuit.

For example, if AMD or Intel constructed a 16 bit wide bus, and to make that bus work required 100 transistors (I am pulling out numbers just to illustrate), then if they widen that to 32 bits and to do so took 180 transistors.... the 32 bit wide bus would clock slower over all... but because it is 16 extra bits wide it will have higher bandwidth... just as an example. The reason, 100 transistors add up to X capacitance, 180 transistors add up to y.... the total capacitance of Y is greater than X, hence slower by the propogated delay... simple equation, Td=C*V/I, where C is the capacity, and Td is the switching speed.... more C means larget Td, hence longer to switch the circuit to the end result.

Barcelona will spend enormous transistor budget to get the IPC logic way up.... thus, it will clock down for the most part.

AMD's roadmap has a 2.9 GHz dual core Agena on the horizon, but it will be a way out there as they tweak up the process....

This is the reason I say we may not be able to quantify the 65 nm improvement over time because they may not carry Brisbane through to the end of the process life cycle....

I was very careful to make clear, if Barcelona does not reach Brisbane clocks or even exceed Brisbane clocks.... this is by no means an indication that they took a step backward. i.e. You will not see me poke fun at Barcelona clocks being lower than Brisbane because I know the reason at it is a good reason.... hence Barcelona launching at only 2.3 for quad makes sense to me.

Jack

Indeed - but I think the 2.3Ghz quad has more to do with energy efficiency than anything else,

Either way, AMD NEEDS Barcelona to get clockspeeds upwards of 3Ghz quickly to stave off the higher clockspeed of Penryn. That's the way it seems to me, anyway...

 

[NDLunchbox]

Arrrh... pirates have hijacked me thread!

 

[ajfink]

Arrrh... pirates have hijacked me thread!

:lol:

Sorry...but we DID answer your question as best we could!

 

[NDLunchbox]

I know, I was just kidding... didn't much care for the sock-draw comment though - not a very constructive reply.

 

[ajfink]

I know, I was just kidding... didn't much care for the sock-draw comment though - not a very constructive reply.

Just got to get past the people who for some reason don't contribute constructively and would rather say "AMD sux" or "screw Intel" instead.

Don't worry about it,

 

[NDLunchbox]

Thanks - I'm hardware agnostic, I see the good in all gadgets!

 

[TSIMonster]

I vote for the Brisbane simply because I have one and love it thus far. It is the coolest running CPU that I have ever had. At 2.6Ghz it idles at around +-30 degrees, unsure of load temp.

This is also my first dual core desktop... (I somewhat lean AMD(burnt by netburst, literally)) But I do love the Pentium Ms and I also have a core duo notebook.

 

[ajfink]

Burnt your NetBurst? Story time, methinks,

 

[dasickninja]

I subject a Prescott to a kind of heat it couldn't survive. Thermite. Nuff said.

 

[NDLunchbox]

I have a room full of Netburst Xeon DL360's... they do run hot! We had an A/C unit go down and it hit 100+ degrees F in there. I've got a stack of Clovertowns running VMWare on order... 4 to replace 30. Can't wait, hope they get here before summer!

 

[ajfink]

I have a room full of Netburst Xeon DL360's... they do run hot! We had an A/C unit go down and it hit 100+ degrees F in there. I've got a stack of Clovertowns running VMWare on order... 4 to replace 30. Can't wait, hope they get here before summer!

I can't even imagine how much cooler it will be.

 

[TSIMonster]

I know, we had a server room with about 30 of the same servers..... we had to build a new SUPER cool IT room to hold them. It reached 100+ nearly everyday last summer. Unfortunately... my boss only looks at Ghz, so he'll probably keep getting the same ole same ole...