AMD released its first 65 nm processors in December, covering the mainstream and the lower high-end markets with the new Brisbane dual core. Performance hasn't changed much with the die shrink, but power consumption has dropped nicely. But how overclockable is the new processor? And when will it be available? We find out.
Nothing more than AMD projected, in fact a little less. Reducing the L2 cache negativelty impacted overall performance while freeing up more die space. Good for AMD in terms of manufacturing capacity, hopefully good for the consumer in terms of potential price reductions, bad for the consumer in terms of performance.
If I were buiding a bank of servers, the potential energy savings might excite me, but Im not. Im a 1 system at a time DIYer and I want perfomance. I dont want "platformance" or "mega tasking" or other advertising gimmicks. I would much rather have an overall 11% gain in application processing power than an 11% reduction in power usage.
I'll pass on brisbane and wait to see what K8L has to offer.
Has AMD stated publicly that K8L will put them back in the lead? The only places I have read this is Blogs and such. If AMD has set any expectations for K8L themselves, could someone give me a link? Else i'm afraid that it might just be the fans and blogs generating the hype (much like the 65nm was).
Has AMD stated publicly that K8L will put them back in the lead? The only places I have read this is Blogs and such. If AMD has set any expectations for K8L themselves, could someone give me a link? Else i'm afraid that it might just be the fans and blogs generating the hype (much like the 65nm was).
I don't recall AMD or anyone for that matter stating 65nm alone would be the thing to get AMD back in competition. I do however recall K8L on 65nm being claimed by many posters as the thing to get AMD back.
isn't aimed at you anymore gr8Mikey
Personally i don't see what the big fuss over the die shrink is. AMD did the same thing with 90nm on K8, no one complained much then. Bit your tongues and wait for a new product not a die shrink.
Personally i don't see what the big fuss over the die shrink is. AMD did the same thing with 90nm on K8, no one complained much then. Bit your tongues and wait for a new product not a die shrink.
130nm -> 90nm did NOT result in increased cache latency and decreased performance - in fact performance improved slightly due to enhancements to the IMC.
That is the difference.
A dumb shrink simply should NOT result in decreased performance.
Personally i don't see what the big fuss over the die shrink is. AMD did the same thing with 90nm on K8, no one complained much then. Bit your tongues and wait for a new product not a die shrink.
130nm -> 90nm did NOT result in increased cache latency and decreased performance - in fact performance improved slightly due to enhancements to the IMC.
That is the difference.
A dumb shrink simply should NOT result in decreased performance.
Personally i don't see what the big fuss over the die shrink is. AMD did the same thing with 90nm on K8, no one complained much then. Bit your tongues and wait for a new product not a die shrink.
130nm -> 90nm did NOT result in increased cache latency and decreased performance - in fact performance improved slightly due to enhancements to the IMC.
That is the difference.
A dumb shrink simply should NOT result in decreased performance.
Hmm, but really, what's the fuss about it?
The performance decrease is minimal... and these parts were never aimed at enthusiasts to start with.
Nobody expected anything out of it, excepting lower power consumption (and cheaper to manufacture and increased potential production capacity), and i think this was accomplished.
When Intel went 90nm -> 65nm they managed to double the cores and at the same time consume less power, adding an extra 1/2 hour average battery life - DOTHAN -> YONAH
This is why Intel needs new motherboards for every core - the VRM changes for lower vcores (and with that, higher clock speeds, lower heat etc) while AMD wants to stick with "backward compatibility"- there as bad as each other.
ROFLMAO @ availability of AMD 65nm products.
AMD not only took there dam time bringing 65nm, they didn't bring any improvements with it, on the other hand 130nm -> 90nm for AMD was the same and look how far it came, it seems it scaled better then Netburst, well both Intel and AMD's new cores seem to do better then Netburst in performance, efficency and ram (and stupid) clock speeds.
For now, AMD cant offer me an attractive, well priced product, Intel has me and thats what i bought, simple as that!
K8L hopefully will arrive on time and bring back the competition!
Personally i don't see what the big fuss over the die shrink is. AMD did the same thing with 90nm on K8, no one complained much then. Bit your tongues and wait for a new product not a die shrink.
130nm -> 90nm did NOT result in increased cache latency and decreased performance - in fact performance improved slightly due to enhancements to the IMC.
That is the difference.
A dumb shrink simply should NOT result in decreased performance.
Hmm, but really, what's the fuss about it?
The performance decrease is minimal... and these parts were never aimed at enthusiasts to start with.
Nobody expected anything out of it, excepting lower power consumption (and cheaper to manufacture and increased potential production capacity), and i think this was accomplished.
But who are they aimed at? The mass consumer? Well if thats the target, AMD missed the mark with its pricing and availability. Intels C2 is still cheaper and still has a significant performance advantage for a given clockspeed. AMDs current Uarch will never catch up, either at 90nm or 65nm. The only thing AMD can do until K8L comes out (assuming its a competative Uarch) is try to offer better perfromance for the cash outlay. They're not doing that either, at least not right now.
What has AMD aimed at the enthusiast? QFX Need we say more?
Hmm, but really, what's the fuss about it?
The performance decrease is minimal... and these parts were never aimed at enthusiasts to start with.
Nobody expected anything out of it, excepting lower power consumption (and cheaper to manufacture and increased potential production capacity), and i think this was accomplished.
I guess 'minimal' is a term that is subjective and can be loosely defined.
So I guess you see a 90nm 4600+ outperforming a 65nm 4800+ as a 'minimal' performance decrease then? OK, no problem, I'll just agree to disagree, I won't argue over semantics.
I think the actual performance decrease from the slower L2 cache is quite small, relatively speaking. The slower memory speeds as a result of the half multipliers on Brisbane is the real killer for performance. I would say the majority of the performance loss from Brisbane is a result of the lower memory speeds and not the slower L2 cache.
In a worst case scenario, we can have Brisbane X2 running at DDR2-700 speeds as opposed to DDR2-800 speeds on Windsor X2.
Hmm, but really, what's the fuss about it?
The performance decrease is minimal... and these parts were never aimed at enthusiasts to start with.
Nobody expected anything out of it, excepting lower power consumption (and cheaper to manufacture and increased potential production capacity), and i think this was accomplished.
I guess 'minimal' is a term that is subjective and can be loosely defined.
So I guess you see a 90nm 4600+ outperforming a 65nm 4800+ as a 'minimal' performance decrease then? OK, no problem, I'll just agree to disagree, I won't argue over semantics.
I think the actual performance decrease from the slower L2 cache is quite small, relatively speaking. The slower memory speeds as a result of the half multipliers on Brisbane is the real killer for performance. I would say the majority of the performance loss from Brisbane is a result of the lower memory speeds and not the slower L2 cache.
In a worst case scenario, we can have Brisbane X2 running at DDR2-700 speeds as opposed to DDR2-800 speeds on Windsor X2.
The L2 cache latency is higher resulting in lower performance, along with slower memory speeds.
In reality, the home user will buy one and install it in a typical system, first they cant get one cause there 1) late and 2) not even available (PAPER LAUNCH!!!!!), and why bother aiming at the 65nm ones over the 90nm - same performance (within 10%), overclockability, and what, 11% less power consumption, oops a video card will instantly chew that, the cpu isnt the only thing to use it, another drive or two and more ram will take that 11% to nothing, theres no gain from the new for end users, its a loss if anything (performance).
LOL if anything that 11% less power consumption is because of the performance loss
On a better note, atleast it isnt like a prescott - hotter and slower
I guess 'minimal' is a term that is subjective and can be loosely defined.
So I guess you see a 90nm 4600+ outperforming a 65nm 4800+ as a 'minimal' performance decrease then? OK, no problem, I'll just agree to disagree, I won't argue over semantics.
I think the actual performance decrease from the slower L2 cache is quite small, relatively speaking. The slower memory speeds as a result of the half multipliers on Brisbane is the real killer for performance. I would say the majority of the performance loss from Brisbane is a result of the lower memory speeds and not the slower L2 cache.
In a worst case scenario, we can have Brisbane X2 running at DDR2-700 speeds as opposed to DDR2-800 speeds on Windsor X2.
It is indeed subjective.
But look here. Itunes: -2.2%
Cinebench: -1.4%
Quake4: -4.3%
Oblivion: -0.4%
HL2: -4.1%
In all the other tests is a draw...
This is what i call "minimal".
Now explain me what you call "killer for performance"
Most of these are being shipped to the "typical home user" in the form of Dells and HPs. I really don't think AMD gives a rat's ass about making them widely available to us via normal retail channels, and people who buy from large OEMs will never notice the small difference in performance. At the same time, AMD has increased the amount they can produce (and is continuing to transition) and charging the same for them. All in all, it's very good for AMD, but it doesn't bode too well for enthusiasts (but AMD doesn't have much of the enthusiast market right now, so it doesn't matter).
Personally i don't see what the big fuss over the die shrink is. AMD did the same thing with 90nm on K8, no one complained much then. Bit your tongues and wait for a new product not a die shrink.
130nm -> 90nm did NOT result in increased cache latency and decreased performance - in fact performance improved slightly due to enhancements to the IMC.
That is the difference.
A dumb shrink simply should NOT result in decreased performance.
Yea look at Cedar Mill!
