Deneb's naming scheme is exposed

[Cookie Monster]

Link

Upcoming 45nm Desktop Processors AMD also plans to bring this higher-performance and more energy efficient 45nm processor technology to the desktop PC market in Q1 2009 with the AMD platform codenamed ?Dragon.? This platform will be the second generation AMD performance desktop platform, featuring all next-generation components in comparison to the first generation AMD ?Spider? platform released in 2008. The AMD ?Dragon? platform is designed to harness the power of fusion by optimizing the performance of new 45nm AMD Phenom? II X4 quad-core processors with award-winning AMD 700 Series chipsets and award-winning ATI Radeon? HD 4000 series graphics.

Thank god for the sane person who approved of the naming scheme. The previously rumored 5 digit naming scheme was awful imo.

Deneb is officially set to launch at Q1 2009.

edit - Translated Link

Some points:
- DDR2 800 is supported
- IPC performance increase
- Shanghai is faster than barcelona with the same clocks by approx 20%.
- Supposedly the new immersion lithography is used in the 45nm shanghai parts (something intel is going to use starting from 32nm process)
- Yields are good
- Smart fetch is introduced, this feature supposedly reduces idle power consumption by about 21%.
- Istanbul is coming in 2009 and it brings further performance improvements/power savings instead of just having 2 more cores.

 

[error8]

Originally posted by: Cookie Monster

- Shanghai is faster than barcelona with the same clocks by approx 20%.

If this is going to be true and the overclocking is improved over the current Phenoms, then it looks like it's going to be a winner. Is still gonna be slower then the i7, but if it's priced right, AMD will offer a good alternative to Intel quads.

 

[Martimus]

That doesn't mean that the 5 digit naming scheme is out. I do hope that it is though.

 

[nyker96]

the translation is too hard to read, not sure what they are saying. probably easier to just read the original japanese.

 

[Nemesis 1]

Originally posted by: Cookie Monster
Link

Upcoming 45nm Desktop Processors AMD also plans to bring this higher-performance and more energy efficient 45nm processor technology to the desktop PC market in Q1 2009 with the AMD platform codenamed ?Dragon.? This platform will be the second generation AMD performance desktop platform, featuring all next-generation components in comparison to the first generation AMD ?Spider? platform released in 2008. The AMD ?Dragon? platform is designed to harness the power of fusion by optimizing the performance of new 45nm AMD Phenom? II X4 quad-core processors with award-winning AMD 700 Series chipsets and award-winning ATI Radeon? HD 4000 series graphics.

Thank god for the sane person who approved of the naming scheme. The previously rumored 5 digit naming scheme was awful imo.

Deneb is officially set to launch at Q1 2009.

edit - Translated Link

Some points:
- DDR2 800 is supported
- IPC performance increase
- Shanghai is faster than barcelona with the same clocks by approx 20%.
- Supposedly the new immersion lithography is used in the 45nm shanghai parts (something intel is going to use starting from 32nm process)
- Yields are good
- Smart fetch is introduced, this feature supposedly reduces idle power consumption by about 21%.
- Istanbul is coming in 2009 and it brings further performance improvements/power savings instead of just having 2 more cores.

Could you link to were Shanghi is 20% faster than Barcelona. Please . Your talking about Desk top DeneB . I would like to see were Shanghi is 20% faster than K10 on desltop Apps. Please . Thank you.

 

[TidusZ]

While your serving nemesis do you think you could get me a medium pizza and a lite beer? ty. Furthermore, if AMD can compete with the lower end I7's in price/performance and not be a fire hazard when overclocked then I'd say it'd be a winner for sure. I'm not too keen on spending 1000 on a high end i7, so its either wait for intel's new, new chips, or grab these denebs if they are as good as we are hoping.

 

[Cookie Monster]

Originally posted by: Nemesis 1
Could you link to were Shanghi is 20% faster than Barcelona. Please . Your talking about Desk top DeneB . I would like to see were Shanghi is 20% faster than K10 on desltop Apps. Please . Thank you.

Im talking about shanghai. It was stated in that translated article about the 20% performance figure. Whether its true or false is another story.

 

[Nemesis 1]

Originally posted by: TidusZ
While your serving nemesis do you think you could get me a medium pizza and a lite beer? ty. Furthermore, if AMD can compete with the lower end I7's in price/performance and not be a fire hazard when overclocked then I'd say it'd be a winner for sure. I'm not too keen on spending 1000 on a high end i7, so its either wait for intel's new, new chips, or grab these denebs if they are as good as we are hoping.

So whats wrong with the $284 IC7 . Deneb is going to be cheaper at the high end than Intels Lowend @$284. The M/B should be alot cheaper for Deneb . The Memory should also be cheaper . Can't argue against Deneb price thats a fact.

 

[TidusZ]

Originally posted by: Nemesis 1

Originally posted by: TidusZ
While your serving nemesis do you think you could get me a medium pizza and a lite beer? ty. Furthermore, if AMD can compete with the lower end I7's in price/performance and not be a fire hazard when overclocked then I'd say it'd be a winner for sure. I'm not too keen on spending 1000 on a high end i7, so its either wait for intel's new, new chips, or grab these denebs if they are as good as we are hoping.

So whats wrong with the $284 IC7 . Deneb is going to be cheaper at the high end than Intels Lowend @$284. The M/B should be alot cheaper for Deneb . The Memory should also be cheaper . Can't argue against Deneb price thats a fact.

You pretty much nailed it, the thing that's wrong is the exhorbitant prices of the mobo and ddr3, and the unacceptable heat that it generates when overclocked (80+ celcius at 3.8 on good air cooling, give or take from several sites -> check out hardocp).

The price of the chip itself isn't bad, but for the performance you get its not justified by the high overall system price and heat problems. I'm hoping a revision will fix it all and by then prices will be down as well. Either that or Deneb will do the trick.

 

[Idontcare]

Originally posted by: Nemesis 1
So whats wrong with the $284 IC7 . Deneb is going to be cheaper at the high end than Intels Lowend @$284. The M/B should be alot cheaper for Deneb . The Memory should also be cheaper . Can't argue against Deneb price thats a fact.

Yep. Deneb could very well turn out to be the real sleeper CPU for the folks who aren't overly interested in running 8-threaded apps (or 8 instances of a single-threaded app).

Of course for those folks the price point to compare performance/dollar with is yorkfield and the P35/X38 DDR2 boards.

 

[Nemesis 1]

Originally posted by: Cookie Monster

Originally posted by: Nemesis 1
Could you link to were Shanghi is 20% faster than Barcelona. Please . Your talking about Desk top DeneB . I would like to see were Shanghi is 20% faster than K10 on desltop Apps. Please . Thank you.

Im talking about shanghai. It was stated in that translated article about the 20% performance figure. Whether its true or false is another story.

Oh! OK heres the threads name.
Deneb's naming scheme is exposed . I wasn't tring to be a funny guy. You are aware that server k10.5 isn't any where near 20% faster in desktop apps. The ones shown anyway. So Its really not right to spout those numbers when with what we have seen . those numbers you give don't apply to the desktop as we have seen. Not saying deneb platiform won't be faster, But so far nowhere near that kind of gain.

Its much like IC7 for the desktop. Many argue that IC7 offers little over Penryn . But watch what happens on servers.

Wouldn't it be ironic if it turns out that IC7. Is a better performance increase over penryn on the desktop . Than Deneb is over K10 on the desktop.

With no one moving away from penryn to IC7 because of performance differance. If Deneb is even less improved over K1O thats be interesting . We have also seen hugh gains with IC7 over penryn . in multithreaded apps. So pure averages will show IC7 made the biggest performance gains. I am sure when there tested side by side . The testers will use what ever apps. fit their agenda.

 

[Nemesis 1]

Originally posted by: TidusZ

Originally posted by: Nemesis 1

Originally posted by: TidusZ
While your serving nemesis do you think you could get me a medium pizza and a lite beer? ty. Furthermore, if AMD can compete with the lower end I7's in price/performance and not be a fire hazard when overclocked then I'd say it'd be a winner for sure. I'm not too keen on spending 1000 on a high end i7, so its either wait for intel's new, new chips, or grab these denebs if they are as good as we are hoping.

So whats wrong with the $284 IC7 . Deneb is going to be cheaper at the high end than Intels Lowend @$284. The M/B should be alot cheaper for Deneb . The Memory should also be cheaper . Can't argue against Deneb price thats a fact.

You pretty much nailed it, the thing that's wrong is the exhorbitant prices of the mobo and ddr3, and the unacceptable heat that it generates when overclocked (80+ celcius at 3.8 on good air cooling, give or take from several sites -> check out hardocp).

The price of the chip itself isn't bad, but for the performance you get its not justified by the high overall system price and heat problems. I'm hoping a revision will fix it all and by then prices will be down as well. Either that or Deneb will do the trick.

Well lets wait for deneb and run both cpus at same O/C than we can talk about that .

If your O/C an IC7 using 8 threads yep with stock cooling it heats up . It will likely use more power to . But thats when you have to look at work completed in time. Than figure power usage . Simple enough concept . Don't ya agree.

There is NO heat problem . This is the same shit as it won't overclock . Just lets wait get these 2 side by side. Bring a urn tho as IC7 will kill deneb .

Than there is this . The top end penryn is going to be cheaper than the cheapest IC7 . Dened probably loses to penryn on the desk top also . Penryn M/B are cheap . DDR2 is cheap so intel is offering alot . Hopefully AMD can withstand it . But trueth is . When Merom came out Intel only had P4s to fight with at the lowend. Now they have Penryn. See the differance . Its hugh .

I don't know why but many here view things like thats what I am buying . That cool thats what were about. BUT. Its not about us . Its about HP and Dell and the like . AMD is going to have a very hard time selling to these guys if Intel undercuts AMDs pricies . Which we all know intel is going to do . You guys believe AMD is better off today than 2 years ago . YOUR wrong. Intel by LAW can undercut AMDs new tech with intels old tech . This isn't looking good for AMD at all.

 

[Cookie Monster]

Originally posted by: Nemesis 1
Oh! OK heres the threads name.
Deneb's naming scheme is exposed . I wasn't tring to be a funny guy. You are aware that server k10.5 isn't any where near 20% faster in desktop apps. The ones shown anyway. So Its really not right to spout those numbers when with what we have seen . those numbers you give don't apply to the desktop as we have seen. Not saying deneb platiform won't be faster, But so far nowhere near that kind of gain.

Its much like IC7 for the desktop. Many argue that IC7 offers little over Penryn . But watch what happens on servers.

Wouldn't it be ironic if it turns out that IC7. Is a better performance increase over penryn on the desktop . Than Deneb is over K10 on the desktop.

With no one moving away from penryn to IC7 because of performance differance. If Deneb is even less improved over K1O thats be interesting . We have also seen hugh gains with IC7 over penryn . in multithreaded apps. So pure averages will show IC7 made the biggest performance gains. I am sure when there tested side by side . The testers will use what ever apps. fit their agenda.

The thread title? yes deneb's naming is exposed, but i do mention about "leaked slides" and these are about shanghai. I could have made a different thread to keep the discussions seperate however.

When i mention shanghai and barcelona, im not referring to any desktop apps at all. These are server chips and i.e the 20% pefomance figure refers to the server related benchmarks. Like i said, this is what the article stated. Obviously according to AT, shanghai is roughly 10~15% faster than barcelona (with a few exceptions).

Everyone is eagerly waiting for denebs performance, but imho i expect maybe ~10% performance improvement.

 

[Dadofamunky]

Looking forward to seeing Deneb and really hope it does well. I have a lot of goodwill towards AMD for all the obvious reasons and want to see them stay in the game. Especially with the way things are going in the broader economy right now, and Intel anouncing negative guidance.

 

[Idontcare]

Originally posted by: Cookie Monster
Everyone is eagerly waiting for denebs performance, but imho i expect maybe ~10% performance improvement.

I forget where I read it, but someone who did a review of Shanghai (I'll be embarrassed as hell if it was AT) mentioned that the L1 cache had been reworked in 45nm to eliminate what was a significant but unmentioned performance issue in the 65nm implementation and that this L1$ improvement combined with the larger L3$ would make for a >10% IPC bump over Barcelona/Agena on its own even if AMD did nothing else to the core logic of Deneb/Shanghai.

No idea if it was a reputable observation, it clearly did not impress me too much as being credible as I didn't bother to note the source or remember the link. But there is hope, maybe?

 

[nyker96]

good for you for exposing such a heinous conspiracy by AMD!

 

[Cookie Monster]

Originally posted by: Idontcare
I forget where I read it, but someone who did a review of Shanghai (I'll be embarrassed as hell if it was AT) mentioned that the L1 cache had been reworked in 45nm to eliminate what was a significant but unmentioned performance issue in the 65nm implementation and that this L1$ improvement combined with the larger L3$ would make for a >10% IPC bump over Barcelona/Agena on its own even if AMD did nothing else to the core logic of Deneb/Shanghai.

No idea if it was a reputable observation, it clearly did not impress me too much as being credible as I didn't bother to note the source or remember the link. But there is hope, maybe?

For the past 2 years, it was "the empire strikes back". But could next year be "when the jedi returns"? i hope so too.

AMD still has some time to refine their 45nm process and i.e release higher clocked denebs (>2.7GHz) later in Q109. Its interesting to note the north bridge clocks of the new shanghai chips.

The Opteron 2384/2382 has a NB speed of 2.2GHz, while the rest has 2.0GHz. That is alot higher compared to the NB speeds barcelona launched with. (For reference - 1.8GHz for opteron 2350 and <1.6Ghz for the rest). 400MHz increase.

So hopefully, seeing 2.4GHz or higher NB speeds on denebs could well be a possibility i.e adding in the extra % performance against its predecessor.

@nyker96

my memory is sketchy but is today the "lets be sarcastic" day

 

[Idontcare]

Originally posted by: Cookie Monster
The Opteron 2384/2382 has a NB speed of 2.2GHz, while the rest has 2.0GHz. That is alot higher compared to the NB speeds barcelona launched with. (For reference - 1.8GHz for opteron 2350 and <1.6Ghz for the rest). 400MHz increase.

I had no idea the L3$ on Barcelona was clocked so much lower than it was on Agena. Or do I just have skewed sense of what Agena's L3$ clocks were because everyone here overclocks their Phenom NB to 2.4GHz?