Sign in with
Sign up | Sign in
Your question

AMD vs. AMD...

Last response: in CPUs
Share
June 2, 2007 4:06:20 PM

http://www.overclockers.com/tips01161/

Clock for clock dual Barcelona (compared to a 4P Opteron) only has a 5% performance advantage in floating point and 15% in integer?

More about : amd amd

June 2, 2007 4:37:57 PM

1) We can discuss AMD-released info until we're Intel-Blue in the face, but it doesn't mean squat until there's benches.

2) If this is the case, it's time to pull out Don Meredith to sing his old Monday Night Football song: Turn out the lights, the party's over...
June 2, 2007 5:05:35 PM

Oh man. For AMDs sake K10 better do more than +10% on the Opteron C-f-C.

This would certainly explain AMDs total refusal to release real benchmarks.
Related resources
June 2, 2007 5:31:12 PM

Also bear in mind K10 will have to compete with Penryn class CPUs. With another +10% IPC and a 10-20% clock boost, I dont see K10 pulling ahead of the Penryn chips.

AMD will have a competitive product but not a performance advantage that would allow them to dictate pricing like they did with K8.
June 2, 2007 5:35:28 PM

Quote:
Also bear in mind K10 will have to compete with Penryn class CPUs. With another +10% IPC and a 10-20% clock boost, I dont see K10 pulling ahead of the Penryn chips.

AMD will have a competitive product but not a performance advantage that would allow them to dictate pricing like they did with K8.


Forget Penryn, if K10 only has a 10%+ IPC over K8... it won't look pretty next to Kentsfield and its server variants.

But I do agree, we need real benches.
June 2, 2007 6:01:09 PM

Quote:
Also bear in mind K10 will have to compete with Penryn class CPUs. With another +10% IPC and a 10-20% clock boost, I dont see K10 pulling ahead of the Penryn chips.

AMD will have a competitive product but not a performance advantage that would allow them to dictate pricing like they did with K8.


Until Pentyn is actually out, which is in and of itself as much of a mystery as K10, we should probably refrain from making performance assumptions. Even though there is far more information available for penryn than K10, until retail versions are in the hands of 3rd party testers, penryn perfroms better than nothing. For all anyone really knows, penryn might scale to 45nm like K8 scaled to 65nm. Granted, unlike K8, C2D is being tweaked at the same time its going to the new node, but until someone other than Intel tests it, it has no advantage.
June 2, 2007 6:08:00 PM

Proof positive that Al Qaeda recruits from the AMD boardroom.

Four guys plan for over a year to blow up some fuel tanks at JFK and they can't pull it off.








Oh, alright. Thank goodness that they were caught! Terrorism is nothing to joke about.

As has been said, discussing performance figures that have not been independently verified is a fool's game. However, if these figures are anywhere near accurate, we can conclude that there is a very real reason why the real benchys aren't being released! Hector is trying to line up his plastic surgery and mansion in Paraguay first! :twisted:
June 2, 2007 6:10:55 PM

Hector doesnt have to worry too much....he relly hasnt said a lot. Henri on the other hand, and all the nameless "bobs" from AMDs various departments, well, thats another story
June 2, 2007 6:17:13 PM

Quote:
Hector doesnt have to worry too much....he relly hasnt said a lot. Henri on the other hand, and all the nameless "bobs" from AMDs various departments, well, thats another story


Yeah, but you know that unless you're in the US Cabinet, you can't just claim that you didn't know what your underlings were up to! :lol: 
June 2, 2007 9:26:52 PM

Quote:
Hector doesnt have to worry too much....he relly hasnt said a lot. Henri on the other hand, and all the nameless "bobs" from AMDs various departments, well, thats another story


Yeah, but you know that unless you're in the US Cabinet, you can't just claim that you didn't know what your underlings were up to! :lol: 

Why should plausible deniability be the exclusive right of politicians?
June 2, 2007 10:05:03 PM

Quote:
Why should plausible deniability be the exclusive right of politicians?


Because that is their constitutionally-assured right and privilege! Let Democracy Ring! :lol: 

Quote:
*sigh* oh well maybe next year. :roll:


Time to start the AMD joke mill:

What do you call a K10 owner? Barcelonely.

K10 is a K9. Woof.

Is that a ski hill or AMD's stock chart?

What do you get when you merge ATI and AMD? A video company without vision and a computing company that doesn't compute.

K10's L2 stands for Loser.


Need we continue...
June 2, 2007 10:33:54 PM

Quote:
http://www.overclockers.com/tips01161/

Clock for clock dual Barcelona (compared to a 4P Opteron) only has a 5% performance advantage in floating point and 15% in integer?


I woud tend to believe David Kanter and realworld tech rather than Ed.
June 2, 2007 10:43:29 PM

I don't make any comparisons without benchmarks....
June 3, 2007 12:31:02 AM

Baron, I don't know if anyone else would pitch this to you, so I will.

Would you consider contacting Jack and settling your differences?

There is no way that THG would be the same without YOU or HIM. I'm aware that you two have been going at it pretty fast and furious, but I'm absolutely certain that deep down there is some mutual respect and that is something to build on.

Baron. Please take this under consideration.

Thanks.
June 3, 2007 4:40:11 AM

If the numbers are real, I will be really disappointed with the integer performance. :( 
June 3, 2007 7:14:27 AM

How come there are only 3 tests showing that AMD blowing away Intel? Please correct me if I am wrong but I remember reading somewhere that there are more than just the 3 bench marks showing Intel beating AMD (3DMark, power consumsion, etc.) Should beating someone 3/12 consider winning or should it mean that AMD perform exceptionally well in these 3 tests but not overall?
June 3, 2007 7:59:58 AM

Quote:
K10 is a K9. Woof.
lame
Quote:
What do you call a K10 owner? Barcelonely.
a bit lamer

Quote:
K10's L2 stands for Loser.
really lame

Quote:
Proof positive that Al Qaeda recruits from the AMD boardroom.

Four guys plan for over a year to blow up some fuel tanks at JFK and they can't pull it off.
Oh, alright. Thank goodness that they were caught! Terrorism is nothing to joke about.
fucking gay
So why did you say it?

Quote:
Need we continue...


Yeah if you can actually find ones that are funny and that make sense... te he
June 3, 2007 7:05:39 PM

Excuse my rudeness last night Captain Crunch (may I call you that? it sounds cooler) I was a bit out of line. What's done is done, all have a good day!
June 3, 2007 7:11:37 PM

You can call me anything, just don't call me happy that Jack's gone. :cry: 

No rudeness perceived at all and no need to apologize, but thank you anyway. This forum is all about people agreeing or disagreeing with each other. Forumz members should never let it get to the point where extremely valuable posters are driven out.
June 3, 2007 7:16:18 PM

Yeah, I agree, I asked in the other thread, answer in either or... Why did Jack take off?
June 3, 2007 7:22:09 PM

Quote:
This forum is all about people agreeing or disagreeing with each other.

I disagree with that.







:twisted:
June 3, 2007 9:46:40 PM

Quote:
AMD vs Intel

http://www.overclockers.com/tips01162/

It seems that AMD is still leading :D 

in server apps.
in other words, we might see another landslide victory on the desktop and mobile for Intel. :roll:
June 4, 2007 1:10:30 AM

Quote:
http://www.overclockers.com/tips01161/

Clock for clock dual Barcelona (compared to a 4P Opteron) only has a 5% performance advantage in floating point and 15% in integer?


I woud tend to believe David Kanter and realworld tech rather than Ed.

Was Ed the one that wrote that very poorly thought out piece about the convertible bond being bad for AMD shareholders because it would dilute the stock (failing to get the very simple obvious point the AMD stock would have to double first!)?

I too went to the Dailytech piece, which was a lot more informative.
June 4, 2007 3:22:24 AM

Quote:
Hector doesnt have to worry too much....he relly hasnt said a lot. Henri on the other hand, and all the nameless "bobs" from AMDs various departments, well, thats another story


Yeah, but you know that unless you're in the US Cabinet, you can't just claim that you didn't know what your underlings were up to! :lol: 

Gee, I thought they spent a lot of time claiming the fifth; of bourbon, scotch, tequila, vodka. Isn't that what they keep in the "Cabinet"?
June 4, 2007 3:45:01 AM

Quote:
AMD vs Intel

http://www.overclockers.com/tips01162/

It seems that AMD is still leading :D 


That is 2P servers vs. 4P servers. Then it is 2P vs. a single CPU. Go work out the prices on those things. The Xeon 5355 they used are about $1200 each. The Opteron 8222SE costs well over $2000 each.
June 4, 2007 3:51:17 AM

Quote:
Also bear in mind K10 will have to compete with Penryn class CPUs. With another +10% IPC and a 10-20% clock boost, I dont see K10 pulling ahead of the Penryn chips.

AMD will have a competitive product but not a performance advantage that would allow them to dictate pricing like they did with K8.


Until Pentyn is actually out, which is in and of itself as much of a mystery as K10, we should probably refrain from making performance assumptions. Even though there is far more information available for penryn than K10, until retail versions are in the hands of 3rd party testers, penryn perfroms better than nothing. For all anyone really knows, penryn might scale to 45nm like K8 scaled to 65nm. Granted, unlike K8, C2D is being tweaked at the same time its going to the new node, but until someone other than Intel tests it, it has no advantage.

Actually, I must disagree. Intel has shown benchmarks on well known applications, on idential chipsets and memory showing that Penryn is faster than Conroe.

http://www.tgdaily.com/images/stories/article_images/ID...

On some ,the difference clock for clock is pretty small (a few %) on others huge (40%) - Looks like 15% overall - might be 12, might be 17, but I am pretty sure it's more than 10, and less than 20 (Unless the p35 and x38 chipsets are WAY better than the old 975s, but so far while they seem better, they are not WAY better, at least not yet)

The huge unknown question, perhaps even still unknown to Intel, is how will this thing ramp clock wise?

We have seen Conroes on special cooling go deep into the 4.00+ ghz range so we know the design gets there, the question is will the thermals let it go that high?

Both AMD and Intel have shown amazing capacity to tweek a process over time to lower the thermals. Athlon started as single core 110 watt parts, and ended up at under 45 watts for dual cores. Conroe started out at 90 watts for Dual core and now has sub 50 watt quads.

How far along Intel in with this tweeking process, and how far it can ultimately go, is the $6.4 billion question.
June 4, 2007 4:02:31 AM

Quote:

Actually, I must disagree. Intel has shown benchmarks on well known applications, on idential chipsets and memory showing that Penryn is faster than Conroe.

http://www.tgdaily.com/images/stories/article_images/ID...

On some ,the difference clock for clock is pretty small (a few %) on others huge (40%) - Looks like 15% overall - might be 12, might be 17, but I am pretty sure it's more than 10, and less than 20 (Unless the p35 and x38 chipsets are WAY better than the old 975s, but so far while they seem better, they are not WAY better, at least not yet)

The huge unknown question, perhaps even still unknown to Intel, is how will this thing ramp clock wise?

We have seen Conroes on special cooling go deep into the 4.00+ ghz range so we know the design gets there, the question is will the thermals let it go that high?

Both AMD and Intel have shown amazing capacity to tweek a process over time to lower the thermals. Athlon started as single core 110 watt parts, and ended up at under 45 watts for dual cores. Conroe started out at 90 watts for Dual core and now has sub 50 watt quads.

How far along Intel in with this tweeking process, and how far it can ultimately go, is the $6.4 billion question.




Quote:
until retail versions are in the hands of 3rd party testers, penryn performs better than nothing


Intel has show benchmarks....which has as much validity as AMDs POV 'benchmark' debacle.
June 4, 2007 5:14:55 AM

They were Intel benchmarks in controlled environments. At least they allowed Hexus, Anandtech etc. to run the benchmarks, everything was disclosed. Intel hid nothing. AMD has been real secretive with the K10 results though. They haven't let any of the major tech sites run a benchmark and post the results.

What I would have liked to have seen from Intel was Wolfdale and Yorkfield machines running at the same clockspeed as the current high end Conroe and Kentsfield parts. That way we could have seen the difference in IPC.
June 4, 2007 9:29:49 AM

Quote:
They were Intel benchmarks in controlled environments. At least they allowed Hexus, Anandtech etc. to run the benchmarks, everything was disclosed. Intel hid nothing. AMD has been real secretive with the K10 results though. They haven't let any of the major tech sites run a benchmark and post the results.

What I would have liked to have seen from Intel was Wolfdale and Yorkfield machines running at the same clockspeed as the current high end Conroe and Kentsfield parts. That way we could have seen the difference in IPC.


Yep, Anandtech tested Penryn (under Intel's supervision), way back in April: http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx...

AMD on the other hand refuses to release any meaningful numbers. If people are starting murmur that K10 is broken, whose fault is that?
a b à CPUs
June 4, 2007 10:43:37 AM

Quote:
http://www.overclockers.com/tips01161/

Clock for clock dual Barcelona (compared to a 4P Opteron) only has a 5% performance advantage in floating point and 15% in integer?


That article really shows a whole lotta nothing...
!