These are a few things I was curious about. Anyone have answers to these questions. Appreciate it.
 
1. Why was Harpertown given 16GB of RAM and Barcelona 8GB of RAM in the recent benchmarks?
 
2. Why didn't they use a mother board that supports HT 3.0 and dynamic power management for the individual cores? i.e Socket 1207+. I realise that a lot of people will go the route of flashing their bios and popping in the new Barcelona - BUT we have new features on this chip that some of us are curious to see in action just in case we are buying new and not upgrading.
 
3. Would it make sense for data center to upgrade Clovertown with Harpertown (15% performance gain) or Santa Rosa with Barcelona (80-100% performance gain)? Would a company with AMD servers go for Harpertown for that extra 10-20% performance gain on SINGLE CPU applications.
 
4. Why does Intel have the performance crown When we are comparing 2.0Ghz and 2.5Ghz Barcelona to 3.0 Clovertown and Harpertown? Why not compare 2.0Ghz Clovertown/Harpertown to 2.0Ghz Barcelona? Compare 3.0Ghz Barcelona with 3.0Ghz Harpertown in January/Febuary 2008.
 
5. Also are these same companies going to upgrade to Harpertown when a new socket and infrastructure is comin out in 2008 from Intel - yet in 2008,2009 you just keep dropping in new AMD CPU's and flashing bios? Wouldn't this be cheaper especially if AMD keeps Intels performance advantage at 10%.
 
6. Why would anyone care about a 4-5.0Ghz Harpertown that would consume more power than California? Isn't the real bread and butter in the 2.0-3.0Ghz range for both companies? In fact, aren't the lower clocked CPU's the best sellers since the strike the best balance?
 
7. Why didn't they show us a multi-CPU setup? 2 Harpertowns (8 cores) vs. 2 Barcelona's (8 cores)? 4 Harpertowns vs. 4 Barcelona's? I believe 8 Harpertown Cores behave like 6.6 Cores and 8 Barcelona Cores behave like 7.2 Cores.
 - So at what point would AMD CPU's be untouchable?  
 
8. Why hasn't anyone overclocked their Barcelona's - do they?
 
9. On the benchmarks done with the old 1207 mother boards where bios was flashed, doesn't memory for Barcelona run slower than if they used new boards???? That is - even though they used DDR 667 it was running at 533 on the old boards whereas on new boards it would run at 667. Is this correct? I believe it is something to do with independent power for memory controler on the CPU. Is this right?
 
10. Is AMD gunning for the 3.0Ghz mark on the production side - any Quad CPU that misses the 3.0Ghz mark by way of a faulty Core becomes a 3.0Ghz tri-core? Presumeably most problems involve one core not working when they push the envelope on the Quad - so Tri-cores are clocked pretty high. Anyone else heard this?

people with AMD servers that go for Harpertown will aslo have to get FB-DIMMS

I believe they have compared the cpu's at close to the same frequency as well.  
 
Regardless of how the tests have been performed, I personally think that AMD is now competitive again in the performance arena.  

http://www.youtube.com/watch?v=6ipfxIgD8DU

geez weskurtz81,  
You make it sound like AMD is pickin' up the rear.  
 
http://www.news.com/AMD-surpasses- [...] 39522.html

enigma,  
 
Sales are not the problem.  ASP is the problem.  They really NEED to raise CPU prices.  I want AMD to raise prices.  I want them to stay in business and to continue to make fast cpu's for me to buy.  
 
The fact is, AMD has been brining up the rear.  They have been losing a ton of money over the past couple of quarters, and they cannot continue to do so.  They need to be able to raise CPU prices in order to compete in the long run, otherwise they will not have any money to compete with.

 
Are you an AMD shill?
 

 
Perhaps they couldn't find bigger capacity DDR2 modules at such a short notice?
 

 
Are they commercially available?
 

 
Depends on their application list.
 

 
Umm... why AMD had performance crown when it had Opteron at 2GHz versus Prescott at 3GHz? Or in other words, why frequency suddenly matters and it didn't when AMD was ahead?
 

 
Judging by SkullTrail demo, Harpertown at 4GHz is not impossible in that timeframe so why bother waiting, AMD will be trounced anyway.
 

 
You know that for sure? Also, Harpertown can be just plugged in into existing server boards if I heard correctly. It will surely work fine until the Nehalem comes out.
 

 
Err... because Intel can sell it cheaper?
 

 
8 cores you can see in SkullTrail demo, as for Barcelona, by looking at tri-core announcements it seems that AMD still hasn't produced two working quad-core chips yet, so they couldn't test
 

 
I know this one it's simple... if they don't hit the shelves nobody could touch them
 

 
You are simply expecting too much. It barely works at 2GHz
 

 
They are but it is an evasive target if your gun is small
 
/joke off
 
Seriously, what do you expect, Intel has just announced that they made first functional 32nm SRAM chip with 1.9 billion transistors -- every single one of them works, and AMD can't even manage to get 4 out of 4 cores working.

"by looking at tri-core announcements it seems that AMD still hasn't produced two working quad-core chips yet, so they couldn't test"
 
lol. can't stop laughing!  Comment of the month :-)

levicki,
 
you accuse him of being an AMD shill, but you sound like an Intel shill.  

 
1: Find one benchmark that needs even close to 8GB and we'll talk about the importance of 16GB.
 
2: Those board aren't available yet. The Barcelona were, so they hhad to go HT2.0
 
4: Because freakin 3.0GHz Harpertown is available now while only 2.5GHz Barcelona is on AMD side.
 
5: Would company switch to AMD if they are to loose 10% (which is more like 30 with 3.0GHz by the way) performance with them?????
 
6: Who's talking about 4.5GHz Harpertown? Oh yeah... you biased AMD fanboy!!!!!!!!
 
7: Who fuXXing need 8 core anyway for his server? I'll tell you, about 1% of the market as oppose to maybe close to 20% for 2 and 4 S system.
 
8: I didn't see much overclocking test on Harpertown side neither. Oh yeah, they are both server cpu on server board with very limited OC capability... Fanboy don't see this usually...  
 
Didn't want to be rude, but you sound exactly like a biased fanboy, and that piss me off. Who care if Intel has the upper hand. Has long as AMD survive to get a better architecture out we don't have a problem, do we????

NightlySputnik
 
"Who care if Intel has the upper hand. Has long as AMD survive to get a better architecture out we don't have a problem, do we????"
 
That is all I care about, AMD doesn't have to be faster all the time.... they just need to compete and survive.  

Oh hell no, not another Intel supporter vs AMD fanboy forum...

 
They don't become untouchable. AMD CPUs don't scale past 8P. Harpertown and Clovertown are for DP servers only. Tigerton is Intel's MP server product (up-to 32P).
 
What's with the stupid 2ghz vs. 2ghz comment? You expect people just to ignore the 2.33, 2.66 and 3GHz Clovertown CPUs that are out there now until AMD has faster CPUs?      
 
FWIW, Harpertown goes to 3.16GHz and 3.2Ghz at launch in November.

Because Anand is a Intel paid pumper. Anand has always been using hardwares in favor of Intel.  
 

Same answer as above. Although the boards were not commercially available, there got to be a prototype. Why Anand didn't use the board is pretty clear.
 

Because Intel's predatory pricing, as well as some under-table deals will "force" most company into abandoning AMD platforms. As Hector said, if Intel didn't use this immoral tatics, AMD should already have 50% of the market share.
 

That's because the test was unfair. As Anandtech's article showed, Barcelona is generally 300Mhz faster than identically clocked Clovertown. This means Barcelona has the IPC advantage.
 

Exactly. This is why every companies should adopt AMD's platforms, as they are more future proof than Intel. On the other hand, with Nehalem debuting, a completely new socket and chipset need to be used, and that means tremendous extra cost.
 

Definitely true. AMD pretty much has the performance/ watt advantage with their Barcelona. Harpertown is going to be a lot hotter, since they're clocked higher, and that needed to be cooled with stronger air-conditioning. This also means higher costs.
 

I believe Harpertown cores will behave like 3 cores, while 8 Barcelona cores will behave like 8 core processors, due to native design.
 

To put it simply, none.
 

They certainly do, I believe. As Fuad wrote, Barcelona can be easily overclocked to 3.0Ghz.
 

The memory was actually running at 667mhz. However, if Barcelona has a newer board, DDR2-800 can be used. This means a solid increase in its performance, as AMD's CPU always needs to be paired with decent performing RAM.
 

 
According to [H]ardOCP, AMD are pretty proud of their yields. AMD's 65nm process has always been known for its high yield, with low leakage on the manufactured die. I believe they're just holding the ace in their sleeve, to retard the reaction of Intel. If AMD suddenly releases a 3.0Ghz, Intel will need to spend a decent amount of time ramping the clockspeed up. Since Intel's CPU has diminished return as clockspeed increases, a 3.0Ghz Barcelona should be able to outperform a 3.4Ghz Harpertown by a sizeable margin.

Mandrake,
 
Could you post info proving AMD doesn't scale well past 8p?  I have always been under the impression that it's Intel's who don't.  
 
wes

found some info.  How many HT links does K8 have?  If it's less than 3, Barcy should scale better over 8p.

 
Proof?
 

 
You do know that Anandtech had about two days to run the tests, don't you? Where are they supposed to get a prototype ES motherboard with two days to run all these tests?
 

 
This is hilarious. One minute AMD is claiming that Intel's 'predatory pricing' (pricing CPUs below cost, while magically making profit!) is causing AMD to lose a ton of money. The next they claim that Intel is responsible for high CPU prices and milking the market for profit!   (60bn in monopoly profits press release anyone?!)
 
AMD saying they are 'proud' of their yields means NOTHING. It's not proof. See this for yields information. (When you look at the 'supporting document', scroll to Figure 3-9. The Barcelona die-size is 280mm squared.)
 
Finally, the performance. There are other reviews besides the Anandtech article.
 
I could go on, but I have better things to spend my weekend doing.