Sign in with
Sign up | Sign in
Your question

Deneb in December

Tags:
  • CPUs
  • Phenom
  • Product
Last response: in CPUs
Share
a b à CPUs
September 23, 2008 2:19:53 PM

http://www.theinquirer.net/gb/inquirer/news/2008/09/23/...

A SCREENIE PLONKED on the Expreview website is showing a new ... bla bla bla.

http://en.expreview.com/2008/09/22/amd-45nm-cpu-naming-...

According to sources, AMD will begin to optimize their naming scheme for their 45nm CPU: they will keep Phenom X4/X3 brand, but numbers after it will have five digits.

From Expreview:

Sources mentioned that the name of two 45nm CPU this year will be Phenom X4 20550 and Phenom X4 20350, with their AM2+ and DDR2 support, these two 45nm Deneb CPU are only available for a short period (last order at 09Q2), for AM2+ mobo users to update AMD’s latest CPU.

In 2009 Q1 and Q2, all Deneb CPU will shift to AM3, supports DDR3. also in 09Q1 Propus will show up, they will become Phenom X4 16×00. 45nm triple core CPU Heka and Rana will become Phenom X3 14×00 and Phenom X3 12×00.


Thought it might be of interest ...

More about : deneb december

a b à CPUs
September 23, 2008 10:06:46 PM

more numbers = better????
lol, AMD marketing sucks almost as bad as CPU
September 23, 2008 10:30:40 PM

Ed from overclockers.com

 rocessors&Itemid=4263" rel="nofollow" target="_blank">http://overclockers.com/index.php?option=com_content&vi...

Quote:
What's probably the most practical piece of information in the article, though, is none of the abovementioned, but that AMD isn't going to make these DDR2 Denebs for very long, just a few quarters. Hence, they aren't going to make very many of them. Hence, they're going to be relatively expensive. It some ways, these chips resemble the initial socket 754 Hammers.

Perhaps the biggest puzzle is why these chips exist at all. If AM3 chips are supposed to run on AM2+ motherboards, why make AM2+ only chips? Perhaps the reason for that is that AM3 chips were supposed to have both a DDR2 and DDR3 memory controller. Maybe AMD has changed its mind about that, maybe they can't afford to spare the die space. If you say, "Well, AMD will save die space by not including a DDR3 controller in the initial Denebs," well, sure, but they'll save a lot more die space in total by dropping a DDR2 controller from what will be many, many more AM3 Denebs.

I think we're going to find out that AMD has dropped the dual-memory controller idea from AM3, and these AM2+ chips are AMD's way of making it up to current AMD users. The only other real possibility is that AMD is really having problems with the DDR3 controller and is getting what they're sure will work out the door. They might both be correct.
Related resources
a b à CPUs
September 23, 2008 10:37:14 PM

BadTrip said:
more numbers = better????
lol, AMD marketing sucks almost as bad as CPU

lol at youself too. They are already at 9950, where are you gonna go from there?? It's not like they are changing the whole name of the cpu like intel with that core i7 thingie that doesn't really mean anything... :D 
a b à CPUs
September 23, 2008 10:47:22 PM

well that was marketing also. I dont need to have a sub par CPU.
a b à CPUs
September 23, 2008 10:48:05 PM

maybe to have sub par bbs skills, lol @ myself
September 23, 2008 11:18:08 PM

I wonder if the AM2+ versions just mean that they want to ensure compatibility with existing bios? I'd expect most manufacturer's to update their bios, but that might not happen if AMD wants to move to AM3 and DDR3 for most Phenoms.

Xbitlabs mentioned Propus quads and Rana triples without L3 cache, they sound like budget models designed for AM2+ upgrades, whereas Deneb and Heka triple cores with L3 cache sound like they're designed for AM3.

Looks like I'll stick with my triple core for a year or two until this all pans out.
September 23, 2008 11:54:33 PM

I think having Deneb working on the AM2 boards are for there loyal customers.Deneb will more shine with DDR3 than 2.Just my pennys
September 24, 2008 1:06:18 AM

Kari said:
lol at youself too. They are already at 9950, where are you gonna go from there?? It's not like they are changing the whole name of the cpu like intel with that core i7 thingie that doesn't really mean anything... :D 


I think the problem stems from (and Intel is guilty of this as well) the questionable, confusing, and arbitrary numbering system that both use. I have not seen a single Athlon/Phenom/Core processor (outside the FX/ Opteron line) that ends in anything other than a useless 0 placeholder. They all pretty much end in 00 or 50. Huge gaps in the numbers permeate both systems with full knowledge that they will NEVER be filled in. Both companies are just in love with really really big numbers even if they mean nothing. I think this may have started with Intel and the moronic idea that having a higher hz number on a processor (Pentium 4) was more important than actual performance.

The Phenom line starts at 8xxx (granted I think there was going to be an Phenom X2 line starting in the 6000 range that was dropped). Is it really that surprising they ran out of 4 digit numbers when you skip at least the first 5000 useful ones? So next step is the 1x,xxx range right? Nope, they want a really big number, and those 1s just don't bring across the awesomeness of AMD's new line of processors, so they are going strait to the 2x,xxx range! The FX (and the Opterons before 2007) seem to be the only lines that had any sanity in the numbering.

Intel is going to release the Core I7 and there is strong evidence that it will start at 9xx. Only time will tell how out of control Intel will let this series of number get. Perhaps they got wind that AMD Deneb processors were going to lead with a 2 and they wanted to drive home the greatness of their chips so they started with a 9. But clearly 20550 > 965 so intel may want to strike back with the Core I7 99965 just so AMD knows who is boss.
September 24, 2008 3:07:02 AM

Wake me when reliable sources are out.
September 24, 2008 6:21:20 AM

Dacrath, there is definitely a Spinal Tap kind of logic to model numbers, you know, an amp where the dial goes to 11 beats one where the dial goes to 10.

AMD/ATI sort of cleaned things up with their GPU numbers (now it's Nvidia who's confusing and rebranding), but that sort of logic hasn't arrived with the CPU's yet.

Phenom's ending in 50 were nice. It allowed us to know which were B3's without having to look at specs. Xbit labs reported that there will be dual core's in the 45nm Phenom generation, but no single cores.

There will be a confusion of quads, triples and duals with L3 cache and without, but with different core star names (Deneb, Propus, Hexa, Rana)

We'll see where Intel goes with their nomenclature.

seboj, by the time reliable sources are out, so will the CPU's from both AMD and Intel. It's fun to speculate based on leaked roadmaps and engineering sample specs. That's part of the reason most of us are here on this board.
a c 127 à CPUs
September 25, 2008 9:44:34 AM

Kari said:
lol at youself too. They are already at 9950, where are you gonna go from there?? It's not like they are changing the whole name of the cpu like intel with that core i7 thingie that doesn't really mean anything... :D 


Big difference. Dneb is just a die shrink and a bit of improvment. Core i7 is a new CPU all together. Thats why its Core i7 instead of Core 2 still. Penryn kept Core 2 because it wasa die shrink/improvements.

Reynod said:
The numbering system sounds annoying.

Just like the P4 system they come up with at the end.

Luckily the processors were so bad we just ignored the whole episode.

http://www.intel.com/products/processor_number/chart/pe...


Bleh. I like the old system. Pentium 4 3.2GHz. Easier to understand. And Intels current system is ok. But this new one is annoying sounding.

And they seem to be promising 3GHz. Lets hope they deliver on that this time.

And whats with the 2 types of each chip?
a b à CPUs
September 25, 2008 12:51:08 PM

Expreview has Kuma benched - power looks a bit high ... like a Phenom with two cores disabled.

September 25, 2008 5:26:17 PM

Dacrath said:
Is it really that surprising they ran out of 4 digit numbers when you skip at least the first 5000 useful ones? So next step is the 1x,xxx range right? Nope, they want a really big number, and those 1s just don't bring across the awesomeness of AMD's new line of processors, so they are going strait to the 2x,xxx range!



:pt1cable:  :pt1cable:  :pt1cable:  Way too funny for me.
a c 127 à CPUs
September 26, 2008 12:59:29 AM

Ouch. Same frequency as a Brisbane core and uses 20w more? Could it be because of the disabled cores?
September 26, 2008 1:11:05 AM

Reynod said:
Expreview has Kuma benched - power looks a bit high ... like a Phenom with two cores disabled.


I clicked off the site because of the porno ads on the sidebar. I guess they think everyone interested in tech is a guy who can't get women so he gets addicted to porn? Sad that.

Anyways, I can't wait to see reviews on sites that don't have ads like that. If a Kuma is just a quad with two cores disabled, then it's not worth it. A triple I can see, both AMD and Intel disabled cores to create single core CPU's, but two cores with four core thermals?

Personally, I think that both AMD and Intel should drop dual cores as fast as the market allows. Ideally, we should see triples at the low end, if one core needs disabling with Deneb or Nehalem because of yields, but the mainstream should be four cores. Then developers will finally listen and optimize their software.

I noticed the quote didn't post correctly. Hope it does so in the future. It's the following from George MacDonald, the 19th century fantasy writer and Christian minister who influenced many fantasy writers and also C.S. Lewis:

I find that doing of the will of God leaves me no time for disputing about His plans.
September 26, 2008 1:13:35 AM

jimmysmitty said:
Big difference. Dneb is just a die shrink and a bit of improvment. Core i7 is a new CPU all together. Thats why its Core i7 instead of Core 2 still. Penryn kept Core 2 because it wasa die shrink/improvements.



...

Was there something wrong with Core3?
September 26, 2008 1:32:05 AM

B-Unit said:
Was there something wrong with Core3?


:lol: 

7 is bigger than 3. That's how marketing works. We should all appreciate marketing. :na: 

Remember Nigel from Spinal Tap: 11 is better than 10.
September 26, 2008 2:52:36 AM

yipsl said:
Personally, I think that both AMD and Intel should drop dual cores as fast as the market allows. Ideally, we should see triples at the low end, if one core needs disabling with Deneb or Nehalem because of yields, but the mainstream should be four cores. Then developers will finally listen and optimize their software.


I think I am going to disagree. While my Tim Allen nature compels me to purchase a quad core, most people will not benefit in any way by having more than a dual core. You suggest that developers should optimize their software. I agree if, and only if, it makes sense. How much faster can you really see Word, Excell, solitaire, minesweeper or a plethora of other programs actually running? After being loaded from the hard drive into memory all of them, no matter how fast the processor, run identically. Even Firefox has been coded to display websites almost as fast as the data is downloaded with the latest release. The time and effort to multi-thread these seems pointless (this may change for the web browsers depending on the direction the Internet takes). In fact, the total futility of optimizing many programs is why so much bloat code exists. The time and money saved in the development process more than offsets the paltry ms or two gained in tedious optimizations. Drive and ram space is also become quite prolific.

Some of the largest bulk buyers of PCs are businesses and government. My place of employment runs mainly XP on Pentium 3s. The custom software is limited by network connectivity and server load, not processor speed. Information is displayed as near to instantaneously after download as to make no difference. Other than dropping a few seconds off the boot time a quad Core I7 would do nothing. Dealing with large databases, like most governments and businesses, no update of this software or change in the GUI will ever test its processing power like a Photoshop render or game trying to recreate physics. But then, no one in my entire building will ever run either (at least not while at work).

Further, I am firmly convinced with rising electrical costs the majority of businesses would much rather have a dual core and save on the power bill. This also spills over into the notebook market, where performance is regularly traded away to save the ever precious watt. I do agree that the disabling of 2 cores to create a dual core from a quad while still having the quads power envelope is sketchy. Both AMD and Intel need a dedicated dual core solutions with lower TDP compared to their quad core offerings or find a way to eliminate the current going to unused cores (although that still seems like a waist of silicone).

With this continued trend to try and cut power bills and extend battery life is why I think AMD's fusion and Intel's inevitable cpu/gpu combo chip will be lucrative when released.

This is not to discount the need for more cores/computing power in certain applications. Games continue to push the technological edge, graphic artist, CAD programs, movie editing, and especially server applications can always use a power boost. Like most people on this board I am drawn to more power. I just don't see a compelling argument for the masses needing anything more than a dual core. But then, I am not sure what the Windows 7 requirements are yet. MS seems to drive computer upgrades in more people than game developers could ever hope.

But thats just my opinion
September 26, 2008 3:45:11 AM

I also think dual cores might be enough, however, there was a time when 64Kb of memory was more than you could ever use....and I picked up a 128Mb flash drive for a buck yesterday. I think that technology will keep moving forward, and that we'll keep getting more cores, and software will (slowly) utilize those cores.
September 26, 2008 5:24:32 AM

I no longer see the masses as business or government, and they simply don't need to upgrade for Word or Excel. The masses are people who are finally doing what they couldn't do on the multimedia PC's from a few years ago.

Yes, duals are still good, but they acheive that by high clock speeds. The sort of quad or triple the alleged masses need is a 65 watt processor. No need for gaming to be the generative principle on these type of boxes.

A Phenom 9150E is the sort of quad that needs to arrive for the big box store multimedia PCs:

http://www.newegg.com/Product/Product.aspx?Item=N82E168...

I'm sure there will be something equivalent out in the Deneb and Nehalem generations. Both AMD and Intel claimed that more cores didn't mean higher clocks, but games that barely supported 2 cores, plus marketing demands, drove up thermals and clocks.

That will change. There will always be the need for a powerhouse quad or triple, but there's also the need for a quad or triple that uses less power and is more efficient in HTPC and multimedia PC's. They can also handle light gaming too, especially as IGP's improve with hybrid designs.

Dual cores are only enough when they're clocked higher. More cores will be used, and sooner than you think.

Intel claimed today that they saved the world 2 billion in power costs. How much did they lose the world during the days of Netburst?
a c 127 à CPUs
September 26, 2008 11:22:42 AM

B-Unit said:
Was there something wrong with Core3?


No I didn't see anything wrong with it. But considering how much of a change Nehalem is compared to Conroe even Core 3 doesn't truly fit the profle since Core 3 would mean the same CPU just enhanced. In fact Core 2 was basically Core 1 with higher IPC and better clock speeds/OC ability. Nehalem on the other hand is very different. It has a IMC, uses QPi and has a larger L3 cache.

Still the number system for Phenom has changed for Deneb, or so it seems, and it doesn't make sense as to why they jumped over the 10K or hell just used the X in front like ATI did for their next step after the 9800 series.
September 26, 2008 6:52:09 PM

YES! I loved the X800 series. Totally the best naming. What's bigger than a 9..... X! Or, they could put an exclamation mark after each number, like they did with pluses (Athlon 64 3000+). That'd be awesome. Seriously, the marketing boys need to start thinking outside the box. I want to see the

Ph3n0m X4 X9958!

(Note: 9958! ~= 3.1 X 10^35491, Windows calculator doesn't want to figure X9958!)
September 26, 2008 8:04:39 PM

yipsl said:
I clicked off the site because of the porno ads on the sidebar. I guess they think everyone interested in tech is a guy who can't get women so he gets addicted to porn? Sad that.

Anyways, I can't wait to see reviews on sites that don't have ads like that. If a Kuma is just a quad with two cores disabled, then it's not worth it. A triple I can see, both AMD and Intel disabled cores to create single core CPU's, but two cores with four core thermals?

Personally, I think that both AMD and Intel should drop dual cores as fast as the market allows. Ideally, we should see triples at the low end, if one core needs disabling with Deneb or Nehalem because of yields, but the mainstream should be four cores. Then developers will finally listen and optimize their software.

I noticed the quote didn't post correctly. Hope it does so in the future. It's the following from George MacDonald, the 19th century fantasy writer and Christian minister who influenced many fantasy writers and also C.S. Lewis:

I find that doing of the will of God leaves me no time for disputing about His plans.


They can't always control every ad their ad provider puts in their, I've seen pr0n ads slip into perfectly reputable sites like even neowin.net had it happen. And I don't see any pr0n ads. It is a very in depth benchmark though that is worth looking at.
September 27, 2008 3:23:10 AM

Okay, the ads aren't there now. I tried it yesterday and they popped up on the sidebar.

I wonder if it was an ad hijacking? Ads can get redirected, as can sites. Site hijacking to a porno link supposedly happened to Tom's Hardware under the old management, but I'm glad I missed it.

I'll read the review now before I head off to work at the data center.

Just scanned it quickly, it's not that bad, but I do agree with them that it looks like it's a Phenom quad with 2 cores disabled. I know AMD has to make some money, but this they should not do. Hope they can get yields up and retire the duals.

IMHO, only one core should be disabled. A dual core with those thermals is a a bit too much.
a c 127 à CPUs
September 27, 2008 3:47:22 AM

yipsl said:
Okay, the ads aren't there now. I tried it yesterday and they popped up on the sidebar.

I wonder if it was an ad hijacking? Ads can get redirected, as can sites. Site hijacking to a porno link supposedly happened to Tom's Hardware under the old management, but I'm glad I missed it.

I'll read the review now before I head off to work at the data center.

Just scanned it quickly, it's not that bad, but I do agree with them that it looks like it's a Phenom quad with 2 cores disabled. I know AMD has to make some money, but this they should not do. Hope they can get yields up and retire the duals.

IMHO, only one core should be disabled. A dual core with those thermals is a a bit too much.


They probably cannot enable the third core. These are probably a collection of ones that had two bad cores and I thought it would have high thermals because a tri core at the same speed as a quad wasn't much cooler. This leads me to believe that the "dead" cores are still sucking some power out of the system even though they are turned off.
a b à CPUs
September 27, 2008 4:32:06 AM

Maybe the core should be physically served?
a c 127 à CPUs
September 27, 2008 8:21:53 AM

amdfangirl said:
Maybe the core should be physically served?


I doubt even that would help unless they can find a way to close the areas that are causing the leakage.
September 27, 2008 8:23:15 AM

Deneb (probably before December)

Deneb x4 /45nm /socket AM2+ /DDR2-1066 /2.8 ghz, 3.0 ghz
only available for upgraders WITH am2+ mobos until Q2/09 (as a courtesy)

January 8, 2009 = RELEASE @ major trade show
Deneb 45nm /socket AM3 - new socket due to DDR3mem up to 1366

Exact details not known esp. re IMC (internal memory controller). The dual memory controller idea is not known.
It is NOT known whether AM3 chips will work in AM2+ boards. Latest info, this is NOT recommended(?)

There are many and varied reports = rumour and speculation.
Much discussion at forums at:
www.amdzone.com

ALSO = 45nm Shanghai Server release Q4/08.

And yes - the model names and numbers are changing, since this chip is a significant change from Phenom.

Mobo samples are in the works.

see forums at THE ZONE for more info.
www.amdzone.com
a c 127 à CPUs
September 27, 2008 8:27:05 AM

^Why would AMD be stupid enough to actually go back on their word that a AM3 chip would work in AM2+ mobos? That would be the dumbest thing I could think they could ever do considering they already talked about it.

Also, the numbers are a fine change but Deneb is NOT a major change. Its the same core just on a smaller process and more L3 cache. If it was a arch rework THEN it would be a major change, like K8 to K10 or Conroe to Nehalem. Those rae major changes for the CPU itself.

But Deneb is not a major change just a sort of refresh.

As for it before December, I doubt it. All signs hav pointed to a Feb 09 release so I doubt they will change it this late.
September 27, 2008 9:55:03 AM

jimmysmitty said:
They probably cannot enable the third core. These are probably a collection of ones that had two bad cores and I thought it would have high thermals because a tri core at the same speed as a quad wasn't much cooler. This leads me to believe that the "dead" cores are still sucking some power out of the system even though they are turned off.


I'm sure they have 2 bad cores, but they should be trashed instead of marketed as dual cores, especially if they can't get the thermals down. I'll "vote" AMD down -1 on this one.

I'm probably getting voted down by Intel fans like badtrip, jilted AMD lovers like TC or outright AMD marketing hacks like zoot. IMHO, AMD has their place and their tech isn't bad when yields are up and 3 or 4 cores are up to snuff. I'm just waiting for them to get better. As is, I really do prefer AMD/ATI chipsets and ATI GPU's, hence I'm into the total platform. If I weren't, I'd have waited for Nehalem to arrive at 2.66.
September 27, 2008 1:28:18 PM

ZootyGray said:
Deneb (probably before December)

Deneb x4 /45nm /socket AM2+ /DDR2-1066 /2.8 ghz, 3.0 ghz
only available for upgraders WITH am2+ mobos until Q2/09 (as a courtesy)

January 8, 2009 = RELEASE @ major trade show
Deneb 45nm /socket AM3 - new socket due to DDR3mem up to 1366

Exact details not known esp. re IMC (internal memory controller). The dual memory controller idea is not known.
It is NOT known whether AM3 chips will work in AM2+ boards. Latest info, this is NOT recommended(?)

There are many and varied reports = rumour and speculation.
Much discussion at forums at:
www.amdzone.com

ALSO = 45nm Shanghai Server release Q4/08.

And yes - the model names and numbers are changing, since this chip is a significant change from Phenom.

Mobo samples are in the works.

see forums at THE ZONE for more info.
www.amdzone.com


Hey tool :hello:  ,

No site pimping here. kthxbye.
September 27, 2008 1:36:03 PM

jimmysmitty said:
^Why would AMD be stupid enough to actually go back on their word that a AM3 chip would work in AM2+ mobos? That would be the dumbest thing I could think they could ever do considering they already talked about it.


Because its a marketing strategy to keep AMD users from switching to Intel. If you know that your best product can't even beat your competitor's mainstream product, what can you do to keep your market share? That's right, by offering a direct upgrade path from current generation to future generation.

But of course, its really difficult to design circuits to support different pin layouts from components to components (DDR2 vs. DDR3 for example). I suspect AMD's really gambling big on Deneb at the moment. If Deneb still does not perform on par to Nehalem and mainstream Yorkfield, its very likely their AMD base will start jumping ship to the competitor.

Quote:

Also, the numbers are a fine change but Deneb is NOT a major change. Its the same core just on a smaller process and more L3 cache. If it was a arch rework THEN it would be a major change, like K8 to K10 or Conroe to Nehalem. Those rae major changes for the CPU itself.

But Deneb is not a major change just a sort of refresh.


But you know, some people still refuse to believe that there is a possibility AMD will be out of the mainstream CPU business. And they will likely spin as much as they can to keep people from going to Intel.

You're right. Deneb is just like Conroe to Penryn. In fact I reckon that due to the process technology AMD uses, we'll probably see a lot less change from Phenom to Deneb than Conroe to Penryn. At least Intel can keep their clockspeed high :p .

As for it before December, I doubt it. All signs hav pointed to a Feb 09 release so I doubt they will change it this late. said:
As for it before December, I doubt it. All signs hav pointed to a Feb 09 release so I doubt they will change it this late.


Here is the funny part. This guy (ZootyGray?) believed that Deneb will come before December, when NDA doesn't even expire until Jan of 09 :lol:  :lol:  :lol: 
a b à CPUs
September 27, 2008 1:57:23 PM

jimmysmitty said:
^Why would AMD be stupid enough to actually go back on their word that a AM3 chip would work in AM2+ mobos? That would be the dumbest thing I could think they could ever do considering they already talked about it.

Also, the numbers are a fine change but Deneb is NOT a major change. Its the same core just on a smaller process and more L3 cache. If it was a arch rework THEN it would be a major change, like K8 to K10 or Conroe to Nehalem. Those rae major changes for the CPU itself.

But Deneb is not a major change just a sort of refresh.

As for it before December, I doubt it. All signs hav pointed to a Feb 09 release so I doubt they will change it this late.

Deneb IS a arch rework no I don't think that it will be on par with Nehalem.
The fast ones might match a Q9550.
a b à CPUs
September 27, 2008 2:09:55 PM

^ Deneb is more a die shrink plus a few...
a c 127 à CPUs
September 27, 2008 3:24:44 PM

jj463rd said:
Deneb IS a arch rework no I don't think that it will be on par with Nehalem.
The fast ones might match a Q9550.


It is not a rework. A rework would take more than just a year. It would take a lot of time and money that AMD does not have. Deneb is Phenom @ 45nm with a larger L3 cahce. To compare Penryn was just Conroe @ 45nm with a larger L2 cache. Well really in truth Penryn was more of a change than Deneb is considering it was also a HK/MG process and included SSE4.1 instructions.

And Yips, you know what will happen if they change it and AM3 chips don't work on AM2+ mobos. You will have some that will get pissed off and change to whatever the best performance is at the time, and if its not AMD they wont be buying AMD. Only ones who will stick to AMD are the die hard fans.

I understand the complexity but if you can't gurantee it, don't ever say it will then change. Its best just to say that you know it will work on this and this and thats it. If you can get it to work on a older one then great. If not then no loss.
a b à CPUs
September 27, 2008 3:40:16 PM

jimmysmitty said:
It is not a rework. A rework would take more than just a year. It would take a lot of time and money that AMD does not have. Deneb is Phenom @ 45nm with a larger L3 cahce. To compare Penryn was just Conroe @ 45nm with a larger L2 cache. Well really in truth Penryn was more of a change than Deneb is considering it was also a HK/MG process and included SSE4.1 instructions.

And Yips, you know what will happen if they change it and AM3 chips don't work on AM2+ mobos. You will have some that will get pissed off and change to whatever the best performance is at the time, and if its not AMD they wont be buying AMD. Only ones who will stick to AMD are the die hard fans.

I understand the complexity but if you can't gurantee it, don't ever say it will then change. Its best just to say that you know it will work on this and this and thats it. If you can get it to work on a older one then great. If not then no loss.


Yes it is a rework.Not a major (complete) rework but there have been changes on the cores.
It is not just a die shrink with extra L3 cache.
September 27, 2008 4:53:07 PM

Can you please provide links to support that?
a b à CPUs
September 27, 2008 11:28:43 PM

Arch improvements don't really mean total redesign
a c 127 à CPUs
September 28, 2008 5:34:04 AM

jj463rd said:
I've read about it several months ago but didn't save any important links.
I did see one link but it does mention architectural improvements (vague though).

http://www.xbitlabs.com/news/cpu/display/20080828043213...


Quote:
Earlier this year AMD said that code-named Shanghai and Deneb microprocessors, which are based on the improved K10 micro-architecture, will be able to offer higher instructions per clock (IPC) throughput compared to currently available AMD Phenom and AMD Opteron processors, which should transform into higher overall performance per clock. Thanks to higher IPC and larger level-three cache (6MB instead of 2MB), the new processors are likely to offer considerably higher speed than existing quad-core chips by AMD.


This is not a arch rework. This is adding moire L3 that performs better. probably lower latency. As I said Deneb is not a rework of the arch itself. Just some enhancements to add IPC and a smaller process (45nm) which should help the yields on the die itself.
September 28, 2008 6:35:57 AM

Personally I think some architecture tweaking may be implemented. I really doubt that 3x the L2 cache would result in 10~15% improvement on clock-for-clock basis against Phenom.
a b à CPUs
September 28, 2008 7:31:31 AM

Well Here is a photo of a Phenom Die.
http://www.techpowerup.com/img/08-03-27/43261A_Phenom_D...
and I think that this is a photo of a 45nm Deneb Die.
http://www.amd.com/us-en/assets/content_type/Additional...
I have also seen this here
http://images.google.com/images?client=safari&rls=en&q=...
Still Core i7 will whomp on it but the 2.8 Ghz and 3.0 Ghz AM2+ Denebs might be a nice economical upgrade path for those with some AM2+ boards that support it.
September 28, 2008 11:52:18 AM

Well, if I have to buy an AM3 board and DDR3, then what's the point of not going 2.66 Nehalem? Other than the half joking comments I make about liking pins on my CPU, the main reason I've stuck with AMD has been their budget chipsets (690V, 690G and 780G) and the idea of a total platform.

My motherboard does support a 125 watt 9750, but not a 9850. It may or may not support either of the two DDR2 compatible Denebs. I'd hoped that AMD would have controllers for both types of memory worked out so I could put a 95 watt Heka triple core in, at least. That's the one with L3 cache, Propus quads and Rana triples won't have L3 according to leaked roadmaps -- what's the point of that other than cost?

Not sure what I'll do yet. Might just stick with this until the next architecture, but my wife wants an upgrade by February, though a 9600 would make her happy on her ASUS 690G. I'd just rather build her a Nehalem just for the experience. I'll be the die hard fan who waits for AMD's ship to come in from the Gulf of Arabia.

AMD does not need to lose the sales of upgrade fans, unless their partners make some really good deals on motherboard/CPU bundles at places like Newegg or Fry's.
!