AMD's 45nm Opteron's to Launch this Month?

[Rob Williams]

From our front-page news:

We found out just a month ago that AMD would be launching their 45nm desktop Quad-Cores in January (it could be even earlier), and today we learn that on the Opteron side of things, we'll be seeing product a lot sooner - as early as this month.

The first processors to hit the market will be those with an "average power consumption" - of around 75w - and will range between 2.3GHz - 2.7GHz, which is supposedly comparable to 1.7GHz - 2.5GHz of the 65nm products. Out of the nine processors to be launched sometime this month, five will be of the 2-way variety while the other four will be for 8-way configurations.

In January, AMD also follow-up with their Shanghai launch, with initial processors having a power-consumption of around 55w. Clock speeds don't seem to be known right now, but it does appear that higher-performance models will also become available at the same time, or shortly after. Those will feature a TDP of 105W - still much less than current offerings. Hopefully they will be able to breath new life into AMD's offerings.

AMD quite apparently has pulled in the launch of its Shanghai processor at least one full month, which probably has been motivated by Intel’s strong showing in the 2-way and 8-way segments. If Shanghai in fact is as good as AMD claims it is, then Shanghai really is what Barcelona should have been.

Source: TG Daily​

 

[Rory Buszka]

Here's a thread I found at TechPowerUp supposedly offering 'proof' that the 45nm 'Deneb' Phenoms will clock to 4 GHz (though admittedly, that is a pretty insane air-cooling setup.) If it's true, I'll be keeping my wallet on standby.

http://forums.techpowerup.com/showthread.php?t=73521

 

[Rob Williams]

If that's a real screenshot, then AMD is no doubt on the right track, and Intel might finally have reason to raise an eyebrow. There's a lot I'd question, though. First, the smallest known Deneb model is supposed to be the Phenom X4 20350, clocked at 2.8GHz. Why does one screenshot show a 2.2GHz stock speed? It doesn't make sense to underclock the CPU for the sake of a screenshot.

Past that, two different versions of CPU-Z are used... why? Version 1.46 was released towards the end of June, long before the application would have been able to add proper support for the new processors. It could be that nothing would have had to change (given it's just a process shift), but even though, it takes away a lot of the credibility.

Finally, how is it possible that such overclocking potential could arise only from a process change? Current Phenom CPUs have a hard time going much beyond 3.0GHz, yet Deneb can achieve close to 1GHz higher? When Intel made the move to 45nm, we didn't see stark differences like this. Even launch 65nm Quad-Cores from Intel could hit 4.0GHz with a large air cooler... we don't see 4.8GHz or higher with the 45nm counterparts.

As much as I'd love to see AMD finally give Intel a reason for alarm, I find these screenshots hard to believe. If it was an architecture overhaul, it would be understandable, but to see such increases from the "simple" move to 45nm... it seems a bit off.

That skepticism aside, if AMD does manage to push out processors like that, then it's going to be a good day for enthusiasts. But still, as we've seen in the past, clock-for-clock, AMD's processors are about 10% slower than Intel's offerings, so it's going to be pricing and pricing alone that decides the victor. After all, Intel's products use less power, run cooler and can likely overclock to a much higher stable frequency than even the Deneb's will be able to achieve.

I won't discredit AMD, but to see a part come out of nowhere that would really cause the consumer to consider them for the next build would be something else. It will happen eventually... I'm just not sure it's going to be with Deneb.

 

[Kougar]

I'm sure 4GHz is possible to acheive... 3.4GHz is doable enough now with just a basic Phenom and the right AMD southbridge after all.

But overclocking is != launching at 4GHz clockspeeds, like was rumored. Deneb @ 3.0GHz is not very much if all it can offer is a ~20% IPC improvement to go with that clockspeed. That would be great competition for Core 2... but is not enough to compete toe-to-toe against Bloomfield. So once again they are back to playing the value game. Although just that alone would be a good start for AMD, with their fab and financial situation now taken care of for the short term, they can get back to designing good chips.

Rob, I'd never have noticed the CPU screenshots differend in version/color! The TechPowerUp thread is not the source however (looks like he "forgot" to post a link pointing back to it), going to the source of the article you can find more screenshots of Deneb at 3.6GHz: Link

1.5v for 45nm for 3.6GHz is not very good though! My 65nm Q6600 needs less voltage than that.

 

[Rob Williams]

I agree, 4GHz might be possible, but I can't see it being possible without insane voltages, at which point, I don't care anymore. I'm all about stable overclocks, and if Deneb can only hit 3.6GHz with high voltage (some Intel Quad's only need 1.35v to hit that, like our QX9770), then I don't think there is much to be excited about.

As for that thread with the source images... it would mean more to me to see that with a long stress session. It's one thing to achieve a high overclock, it's another to achieve a high overclock and it actually remain stable when taxing it out ;-)

 

[Kougar]

Let me know if you find anyone using IntelBurn on one of them... Supposedly the program now works for AMD CPUs as well... there is honestly nothing better.

What's ironic is the author didn't start out attempting to design one of the best programs for stress testing either....

 

[Rob Williams]

Ahh man. I forgot about that application quicker than I found out about it. Doh. I'll have to download that one and begin stressing our own processors with it. I hope that it supports Nehalem as well.

 

[Kougar]

It should work fine on Nehalem. Although you might become memory limited if it spawns 8 threads... hmm. The program has only gotten better and he's ironed out the bugs with it with 1.7v