AMD & Intel Price War -- Awwww, Yeah!

[x-dANGEr]

It looked too cheap that there must have been something wrong.. (JK) :P

[Lemur]

As some would say, it has begun.

AMD slashes prices on dual cores:

AMD's flagship processor, the Athlon 64 FX-62 for Socket AM2, has dropped to $827 from $1,031 (all prices are per CPU in quantities of 1,000). The rest of the company's dual-core CPU lineup has seen price reductions as well:

• Athlon 64 X2 5000+ from $696 to $301
• Athlon 64 X2 4600+ from $558 to $242
• Athlon 64 X2 4200+ from $365 to $187
• Athlon 64 X2 3800+ from $303 to $152

[AntiochusIII]

Wow, just look at that slashing and gnashing of prices...

GOD BLESS EVERYTHING!

MY TIME HAS COME!

(For the new computer, of course)

And today just happens to be the day when my haggling had finally gotten myself a solid permission to start researching/building a new comp, which had been an issue since God-knows-when, as some orgers might noticed by me asking for advice over here...

Hm, somebody likes me.

[Geezer57]

I attended a meeting today at a local computer distributor (as a former reseller, still on their contact list), where a presentation was made by a regional AMD representative. He indicated they're migrating their desktop processors from .90 micron process to .65 micron by the end of the year (earlier for mobile processors), and that this migration will also include an architechture "refresh". He was quite confident that Intel's recent gaining of the performance lead (with their Conroe series) won't last more than a quarter (3 months or so), and that AMD will be back on top again performance-wise soon.

By mid-2007, they should have introduced Socket AM3, which is backwards-compatible with Socket AM2 processors, but offers much higher Hypertransport speed (5X) and support for both DDR2 and DDR3 memory. Interesting stuff...

[Big_John]

I attended a meeting today at a local computer distributor (as a former reseller, still on their contact list), where a presentation was made by a regional AMD representative. He indicated they're migrating their desktop processors from .90 micron process to .65 micron by the end of the year (earlier for mobile processors), and that this migration will also include an architechture "refresh". He was quite confident that Intel's recent gaining of the performance lead (with their Conroe series) won't last more than a quarter (3 months or so), and that AMD will be back on top again performance-wise soon.

By mid-2007, they should have introduced Socket AM3, which is backwards-compatible with Socket AM2 processors, but offers much higher Hypertransport speed (5X) and support for both DDR2 and DDR3 memory. Interesting stuff...

that could be true, but that could also be AMD propaganda to keep the distributors and retailers on board. fabing a smaller process is great, but i doubt that will boost the K8 much. imo, AMD's going to need a new architecture to jump ahead of conroe in the benchmarks, not just a "refresh". but you never know...

regardless, right now is a great time to buy a new processor no matter which way you go!

[orangat]

There is no way AMD is going to catch up in a quarter with some tweaking, not to mention the Conroes could still improve by integrating the memory controller.

And the K8L is not due till 2008. It'll be a while before AMD regains the lead.

[Geezer57]

There may have been an element of "damage control" in the statements by AMD's rep mentioned in post #64 - he was, after all, a sales representative rather than an engineering/technical one. Granted, the projections brought forth in that meeting were probably optimistic, but AMD (being smaller) should at least theoretically be able to move faster than Intel, so I wouldn't count them completely out.
Intel's new Conroe cores can only be good for the enthusiast/consumer, as AMD (as underdog) is doing what they've done several times before: become the low-cost/midrange value provider, until they can catch up. Remember the K6 vs. Pentium II situation, etc.? Intel has a tendency to rest on their laurels when not pushed (and charge a premium while doing so), so we as consumers need AMD to not be too far behind. The closer and fiercer their fight, the happier we should all be at this end!

[Big_John]

There may have been an element of "damage control" in the statements by AMD's rep mentioned in post #64 - he was, after all, a sales representative rather than an engineering/technical one. Granted, the projections brought forth in that meeting were probably optimistic, but AMD (being smaller) should at least theoretically be able to move faster than Intel, so I wouldn't count them completely out.
Intel's new Conroe cores can only be good for the enthusiast/consumer, as AMD (as underdog) is doing what they've done several times before: become the low-cost/midrange value provider, until they can catch up. Remember the K6 vs. Pentium II situation, etc.? Intel has a tendency to rest on their laurels when not pushed (and charge a premium while doing so), so we as consumers need AMD to not be too far behind. The closer and fiercer their fight, the happier we should all be at this end!

definitely. the conroe wouldn't exist if it weren't for AMD. competition is a good thing, and that's reason enough to buy AMD (the current 'underdog') for some.

anyway, here's a funny little story i ran across.
http://www.theinquirer.net/default.aspx?article=33329

[orangat]

Architecturally the A64 is pretty much tapped out and no quick fix is going to be able to surpass the Conroe. Its going to be a long wait till 2008.

[x-dANGEr]

I guess I'll postpone my new PC to 2008 then.. It only has been waiting 3 years.. 0-0

[hoom]

I think there is less that needs to be changed in A64 to catch up to Conroe than some think.
Its already the right basic architecture (short pipes doing more per clock).
A64 is only marginally weaker per clock in theory, actually does more per clock in a bunch of places but Conroe is just more efficient at keeping its functional units actually doing stuff, & handling cache.

I think it would not take too much tweaking to improve those aspects of A64 & at least get closer to conroes efficiency.

[orangat]

I think there is less that needs to be changed in A64 to catch up to Conroe than some think.
Its already the right basic architecture (short pipes doing more per clock).
A64 is only marginally weaker per clock in theory, actually does more per clock in a bunch of places but Conroe is just more efficient at keeping its functional units actually doing stuff, & handling cache.

I think it would not take too much tweaking to improve those aspects of A64 & at least get closer to conroes efficiency.

What do you mean by the A64 doing more per clock in a bunch of places?

[hoom]

Theoretically in terms of numbers of functional units & how many cycles it takes to do a certain instruction.
Conroe is not far behind (P4 was way behind) & gets the lead in a bunch of other bits.

[orangat]

Theoretically in terms of numbers of functional units & how many cycles it takes to do a certain instruction.
Conroe is not far behind (P4 was way behind) & gets the lead in a bunch of other bits.

What functional units/instructions?

[hoom]

See here & next page.
Though in re-reading it the balance seems more in favour of conroe than I thought on first reading.

[orangat]

See here & next page.
Though in re-reading it the balance seems more in favour of conroe than I thought on first reading.

I still fail to see how you can consider the A64 'to be doing more per clock'.

From the article, the only advantage of the A64 is its throughput for simple shift/rotates which is because it has more independant agu's not because of better latency. The Conroe is simply just as good or better overall in the back end especially in the fpu pipeline where it definitely better.

[hoom]

OK so 'a bunch' was overstating it but my point is that K8L is already stated to be addressing most of the shortfalls (branch prediction, wider SSE, out of order instruction etc).

If AMD can bring those improvements into dual core before the official K8L (quad core), I think the gap will be much reduced.

Quad core wise, AMD is looking healthier unless Intel can pull out some sort of double frontside bus since each AMD core has its own direct access to memory & conroe is already pretty constrained by frontside bus.