Sign in with
Sign up | Sign in
Your question

More Deneb Leaks

Tags:
Last response: in CPUs
Share
August 20, 2008 12:43:22 PM

http://xtreview.com/addcomment-id-6100-view-AMD-45NM-de...

Looks like better clocking capabilities and gobs of L3 cache have had some effect on gaming.

http://www.itocp.com/thread-12164-1-1.html

In Chinese. Lots of pretty pictures, thermal testing, and a couple of benchmarks. Looks like it runs fairly cool under load at 3.2 Ghz (42.9C).

However... the big thing to notice is the Lavalys Everest benchmark. The memory bandwidth scores are beyond bad. Thoughts?



More about : deneb leaks

August 20, 2008 1:27:19 PM

Malovane said:
http://xtreview.com/addcomment-id-6100-view-AMD-45NM-de...

Looks like better clocking capabilities and gobs of L3 cache have had some effect on gaming.

http://www.itocp.com/thread-12164-1-1.html

In Chinese. Lots of pretty pictures, thermal testing, and a couple of benchmarks. Looks like it runs fairly cool under load at 3.2 Ghz (42.9C).

However... the big thing to notice is the Lavalys Everest benchmark. The memory bandwidth scores are beyond bad. Thoughts?


I would have to say AMD has a bit more work left, at least the electrical draw is down.

Word, Playa.
August 20, 2008 1:58:27 PM

spud said:
I would have to say AMD has a bit more work left, at least the electrical draw is down.

Word, Playa.


Seemed ok. Lets see how this is going become, or if it is just FUD.
Related resources
August 20, 2008 2:19:21 PM

a joyful experience, learning that in fact, Shanghai is maybe going to reach core 2 standards..... (I somehow doubt it will reach penryn or will match conroe OCing but...)
August 20, 2008 4:08:58 PM

spud said:
I would have to say AMD has a bit more work left, at least the electrical draw is down.

Word, Playa.


Going by WinRAR bench, and Memory latency in Everest I'd say the TLB fix is enabled. Either through BIOS or Vista SP1. WinRAR numbers should be about 4 to 5x what they had. At least they used a decent bios on the k9a2 plat, same bios I use.
a c 127 à CPUs
August 20, 2008 6:10:49 PM

^I was going to say the memory looked quite low too. Coompared to what I have seen from current B3 Phenoms it looks too low and I would expect that to increase a bit.
August 21, 2008 1:02:58 AM

Who cares about Deneb? Intel finally has an IMC.
a b à CPUs
August 21, 2008 9:39:45 AM

BaronMatrix said:
Who cares about Deneb? Intel finally has an IMC.


...
August 21, 2008 10:19:07 AM

amdfangirl said:
...


Hey it was a somewhat decent pun. But if you didn't like it much ill quote a sensationalist reviewer:

Quoting Hexus

Quote:

As a summary, we know that the monolithic, 45nm-based Nehalem microarchitecture has an integrated memory controller - supporting tri-channel DDR3-1,600 RAM for the desktop - QuickPath Interconnect, simultaneous multithreading, and a three-level cache hierarchy with a large pool shared amongst all cores.


This is a much better pun. Even if i tried i couldn't do better. You gotta love Kool Aid.
a b à CPUs
August 21, 2008 11:27:32 AM

I hope everybody's being sarcastic
August 21, 2008 11:45:44 AM

I like lime
a b à CPUs
August 21, 2008 12:39:32 PM

Kool ... I have an IMC too.

Should be good news to the faithfull sitting on their AM2 mobos ... a nice injection of extra power there.

Hopefully they can get them out into the market soon.

Man that Larrabee article on the front page was **** ... the articles at THG are reaching new lows in tech journalism.

August 21, 2008 12:44:23 PM

They should just admit theyre reaching, when they simply dont know. Its nothing more than what we can see on the forums. Until we see more about Larrabee, we will simply be left in the dark. That goes for Toms writers. Just because theyre interviewing 1 person, I can get a better perception after reading 3 or 4 interviews from various people about it. So to me theyre reaching, and should simply state it
August 21, 2008 12:52:57 PM

JAYDEEJOHN said:
I like lime


Pîe is better, in Soviet Russia you can eat your pie and keep it to.
August 21, 2008 12:56:17 PM

radnor said:
Pîe is better, in Soviet Russia you can eat your pie and keep it to.


But its not just your pie - its everyones ;) 
a b à CPUs
August 21, 2008 1:00:24 PM

Reynod said:
Kool ... I have an IMC too.

Should be good news to the faithfull sitting on their AM2 mobos ... a nice injection of extra power there.

Hopefully they can get them out into the market soon.

Man that Larrabee article on the front page was **** ... the articles at THG are reaching new lows in tech journalism.


*nods*

Maybe we can give THG some sunshine. Maybe some lollipops too...

Would be good for my 690G too...

What am I thinking about again?
August 21, 2008 1:11:13 PM

I like pie, even keylime pie, but for puters, Ill take a slice o that super pi please
a c 127 à CPUs
August 21, 2008 1:20:39 PM

Reynod said:
Kool ... I have an IMC too.

Should be good news to the faithfull sitting on their AM2 mobos ... a nice injection of extra power there.

Hopefully they can get them out into the market soon.

Man that Larrabee article on the front page was **** ... the articles at THG are reaching new lows in tech journalism.


Remeber though ray, this is still up to the mobo makers. I would expect more AM2+ mobos to support Deneb than AM2 though. So for some they will be waiting only to be let down cuz their mobo makers decide not to update the BIOS for it.

Deneb is hard to judge without any reliable sites testing it yet. But good to see they are on their way.
a b à CPUs
August 21, 2008 1:26:05 PM

Amiga500 said:
But its not just your pie - its everyones ;) 


Yes, everyone in Russia should have some. I'll take the fruitarian pie tho, I'm gonna bet nobody wants to eat that!
a c 127 à CPUs
August 21, 2008 1:27:01 PM

^Screw pie. Its all about the Cheese Cake.
a b à CPUs
August 21, 2008 1:48:13 PM

Cheese Cake cannot be good for my moral or weight =P

I'm standard weight now don't want that to change!
a c 127 à CPUs
August 21, 2008 1:50:57 PM

^But its SOOOOOOOOO GOOD!!!!!!

Creamy and smooth and taste sweet and like heaven. Its what I get instead of a birthday cake on my b-day.
August 21, 2008 1:51:32 PM

amdfangirl said:
Cheese Cake cannot be good for my moral or weight =P

I'm standard weight now don't want that to change!


This thread is now about Cheese cake !! Don't go off-topic !!! :kaola: 
a b à CPUs
August 21, 2008 1:54:13 PM

^ Here's a thought: Let's turn this into a thread were I nag about animal rights?

jimmysmitty said:
^But its SOOOOOOOOO GOOD!!!!!!

Creamy and smooth and taste sweet and like heaven. Its what I get instead of a birthday cake on my b-day.


I can control my cravings... I think... better get another apple...
a c 127 à CPUs
August 21, 2008 1:56:57 PM

If AMD offered a cheesecake with every CPU they would sell more. But then again the cake has always been a lie.....

I'll give you some of my experience fangirl. Its dependant per person on how food affects you. I myself can eat a large pizza from Pizza Hut in one setting yet I can't break 155 pounds. Other people eat one and they gain 10 pounds. You just have to find your balance.

Fruits are great but after a while you need to experience all the goodness food has to offer.
a b à CPUs
August 21, 2008 2:05:43 PM

1KG of junk food = 750G of fat into my body. I know. I've tried 1KG of Twix...

You know me. I'm supposed to be a fruitarian that also eats Twix.
a c 127 à CPUs
August 21, 2008 3:37:14 PM

Mmmm.... Twix Cheese cake..... that sounds good too...
August 21, 2008 4:12:36 PM

jimmysmitty said:
Mmmm.... Twix Cheese cake..... that sounds good too...


Heretic !!! This reminds me of Homer J. Simpson. Fattening !!!!
a b à CPUs
August 21, 2008 4:25:19 PM

cheescake it is !!!

August 21, 2008 4:31:55 PM

Malovane said:
http://www.overclock.net/amd-cpus/356662-amd-phenom-45n...

4 Ghz in the mid 50C range on stock cooling, so they say. Not bad for an early engineering sample.



Wow, that was supposedly a C2 Deneb so there may be a 3GHz Shanghai since those have more stringent requirements. I was hoping they would't release at C0. if that's what's in the oven now they will launch Shanghai in October with systems hopefully by early December.

And what's really amazing is that they aren't using HiK yet. Imagine what the D2 chips will be like. CPU-z does suck at voltages so I assume that 1.22V is not correct. It would be great if it was though 4GHz at 1.2V :pt1cable: 
August 21, 2008 4:32:08 PM

Talking about animal rights, my cat just pointed at the screen to your post, and said....all your puters belong to us. Im sure he meant you
August 21, 2008 4:45:13 PM

BaronMatrix said:
Wow, that was supposedly a C2 Deneb so there may be a 3GHz Shanghai since those have more stringent requirements. I was hoping they would't release at C0. if that's what's in the oven now they will launch Shanghai in October with systems hopefully by early December.

And what's really amazing is that they aren't using HiK yet. Imagine what the D2 chips will be like. CPU-z does suck at voltages so I assume that 1.22V is not correct. It would be great if it was though 4GHz at 1.2V :pt1cable: 



The CPU-z was wrong, check the overdrive voltages.

1.47V or something.




I'll hold fire on this - after all, we did see 3 gig Phenoms around this time last year and we know how that turned out.
a c 102 à CPUs
August 21, 2008 4:56:54 PM

jimmysmitty said:
^But its SOOOOOOOOO GOOD!!!!!!

Creamy and smooth and taste sweet and like heaven. Its what I get instead of a birthday cake on my b-day.


Cheesecake is a very versatile dessert. I had seven cheesecakes at my wedding instead of a standard wedding cake. It was AWESOME.
August 21, 2008 5:05:59 PM

Amiga500 said:
The CPU-z was wrong, check the overdrive voltages.

1.47V or something.




I'll hold fire on this - after all, we did see 3 gig Phenoms around this time last year and we know how that turned out.



We all know why AMD didn't pull the trigger on 3GHz. It would have been 150W+, totally unacceptable for a CPU. 1.47V is not bad for a CPU at 2.3GHz stock. I can see C3 or D0 being able to do 3GHz at 125W.
August 21, 2008 5:18:20 PM

A lot of people think the 4GHz screeny is faked, something about the fonts being slightly different or something. I haven't really looked into it too closely though.
August 21, 2008 5:22:50 PM

BaronMatrix said:
We all know why AMD didn't pull the trigger on 3GHz. It would have been 150W+, totally unacceptable for a CPU. 1.47V is not bad for a CPU at 2.3GHz stock. I can see C3 or D0 being able to do 3GHz at 125W.


Hmmm, how much difference would the extra cache make...


1.47 ain't all that good, power consumption is proportional to voltage squared.

But, then, the machine I'm posting on (Q6600) is at 1.312V stock.


So (assuming an identical CPU resistivity - unlikely) the 45nm would have just over 25% more power draw.
August 21, 2008 5:24:13 PM

In soviet russia drugs do YOU!

So with this chip the design changesintel made use of are
AMD's
IMC
Hypertransport (no it's not! no it's not! it's quick connect! new name! make it different!)
Native-quad core die - more than a year after amd
x86-64 instructons - because why should they bother when 64bit isn't used
independent core clock/voltage control - oh yeah well so can we!

with a splash of Gigabytes CIA dynamic clock control - because it helps justify our price!

Smaller die, updated cache...and 3 way memory linking tossed in with some good ol' reverse engineered features.

Seriously, 3 way memory? Do they think it sounds kinky? They have been on their soap box for 5 years saying how 64bit is nothing more than a marketing strategy, along with the rest of the chips features, but they have focused only on 32bit performance, in 32bit OS's where the system is limited to 2 gigs application accessable memory.

They are like a guy at the play ground trying to get someone in his van with candy...."no no, don't pay attention to what was said 30 seconds ago...i'm not a stranger, you know me, i want to give you this yummy candy"

Btw, when larrabee is another 6-9 months away from offering an 8 core, 2ghz card that delivers UP TO 256 Gigaflop/sec (16 flop/sec per clock/ per core) haven't made a good IGP, let alone actual gaming gpu, and larrabe hinges not only on being coded specially for the card, but their own driver optimization...

being able to deliver performance on par with single core 4870's 1200 gigaflop/sec in the end of 2010 with either a 24 core 3ghz card, or 32 core 2 ghz card...it's a joke. It's based on the logic of, "dur, if a gpu is 20 times faster than a cpu, then 20 cpu's should be as fast as a single gpu."

20% even 30% performane gain still does not make nehalem worth the 200-500% price priemium for a system.
August 21, 2008 5:41:34 PM

Amiga500 said:
Hmmm, how much difference would the extra cache make...


1.47 ain't all that good, power consumption is proportional to voltage squared.

But, then, the machine I'm posting on (Q6600) is at 1.312V stock.


So (assuming an identical CPU resistivity - unlikely) the 45nm would have just over 25% more power draw.



Ummm, the circuits are smaller so the resistance is less. That means less power draw not more. V = IR so I = equals V\R. That's a simple circuit and I don't feel like digging for the 3 phase CPU equation but in any case power goes down as the features get smaller.
August 21, 2008 5:48:02 PM

radnor said:
Heretic !!! This reminds me of Homer J. Simpson. Fattening !!!!


Mmmm Doughnut....Doh, my heart stopped....oh there it goes again.

I don't know, that 4ghz Screen doesn't look faked, though I doubt it would validate, or run prime stable. May be possible though. On the other hand, being able to push a 2.3ghz cpu up to 3.4 is pretty impressive, especially for an AMD chip. Think by the time we see revision c2 or c3 AMD might have a pretty capable chip. Now on the other hand, we need to figure out what is going on in Vista that is shafting all Phenom core processors regardless of stepping. My performance actually went up some more with Xp Sp3. But every time I see people post phenom bench's that use Vista sp1 they look abismally lower than they should. They should at least be on the same playing field, especially with SP1 and definitely in 64bit versions it should actually be better. But instead, according to most of the benchmarks I've seen posted, I'm running between 2 and 4 times faster on a lot of benchies than quite a few people that are running on vista with the same chip.

in other news...Mmmm Pie, I likes me some Pecan Pie. Pie pie pie pie piiieee.
August 21, 2008 5:48:13 PM

iocedmyself said:
In soviet russia drugs do YOU!

So with this chip the design changesintel made use of are
AMD's
IMC
Hypertransport (no it's not! no it's not! it's quick connect! new name! make it different!)
Native-quad core die - more than a year after amd
x86-64 instructons - because why should they bother when 64bit isn't used
independent core clock/voltage control - oh yeah well so can we!

with a splash of Gigabytes CIA dynamic clock control - because it helps justify our price!

Smaller die, updated cache...and 3 way memory linking tossed in with some good ol' reverse engineered features.

Seriously, 3 way memory? Do they think it sounds kinky? They have been on their soap box for 5 years saying how 64bit is nothing more than a marketing strategy, along with the rest of the chips features, but they have focused only on 32bit performance, in 32bit OS's where the system is limited to 2 gigs application accessable memory.

They are like a guy at the play ground trying to get someone in his van with candy...."no no, don't pay attention to what was said 30 seconds ago...i'm not a stranger, you know me, i want to give you this yummy candy"

Btw, when larrabee is another 6-9 months away from offering an 8 core, 2ghz card that delivers UP TO 256 Gigaflop/sec (16 flop/sec per clock/ per core) haven't made a good IGP, let alone actual gaming gpu, and larrabe hinges not only on being coded specially for the card, but their own driver optimization...

being able to deliver performance on par with single core 4870's 1200 gigaflop/sec in the end of 2010 with either a 24 core 3ghz card, or 32 core 2 ghz card...it's a joke. It's based on the logic of, "dur, if a gpu is 20 times faster than a cpu, then 20 cpu's should be as fast as a single gpu."

20% even 30% performane gain still does not make nehalem worth the 200-500% price priemium for a system.



Now that was funny. It does sound like Intel though. I tell people they'r like the little kid who gets mad and tries to take the ball home. Now it turns out it's not even their ball (X86 CPU). I too am interested to see Larrabee fulfill the claims. I mean to say their first GPU will be as fast as 4000 or GT200 is like wow, you can barely draw Aero or HD and you've had what 5 years.

But anyway, I may end up haing an even better studio setup if Deneb is timely. Next month is upgrade time finally.
August 21, 2008 5:53:03 PM

BaronMatrix said:
Ummm, the circuits are smaller so the resistance is less. That means less power draw not more. V = IR so I = equals V\R. That's a simple circuit and I don't feel like digging for the 3 phase CPU equation but in any case power goes down as the features get smaller.



R = (rho * L ) / A

R = resistance
rho = resistivity (material property)
L = circuit length
A = circuit cross sectional area


With a smaller node:

A goes down - R goes up
L goes down - R goes down

A will go down more than L will go down = net resistivity increase.
August 21, 2008 7:24:21 PM

Amiga500 said:
R = (rho * L ) / A

R = resistance
rho = resistivity (material property)
L = circuit length
A = circuit cross sectional area


With a smaller node:

A goes down - R goes up
L goes down - R goes down

A will go down more than L will go down = net resistivity increase.



Not a material scientist but it's seem counter-intuitive that smaller circuits use more power. Why bother using smaller circuits?
August 21, 2008 7:30:00 PM

Well, I wouldn't want to be the one to try overclocking a tube system to 4 Ghz.
August 21, 2008 7:36:34 PM

Resistance goes up. Like friction in a way (actually just like friction) Pushing the same amount of water thru a smaller hose needs more pressure. Same thing.
August 21, 2008 7:36:35 PM

Mathos said:
Mmmm Doughnut....Doh, my heart stopped....oh there it goes again.

I don't know, that 4ghz Screen doesn't look faked, though I doubt it would validate, or run prime stable. May be possible though. On the other hand, being able to push a 2.3ghz cpu up to 3.4 is pretty impressive, especially for an AMD chip. Think by the time we see revision c2 or c3 AMD might have a pretty capable chip. Now on the other hand, we need to figure out what is going on in Vista that is shafting all Phenom core processors regardless of stepping. My performance actually went up some more with Xp Sp3. But every time I see people post phenom bench's that use Vista sp1 they look abismally lower than they should. They should at least be on the same playing field, especially with SP1 and definitely in 64bit versions it should actually be better. But instead, according to most of the benchmarks I've seen posted, I'm running between 2 and 4 times faster on a lot of benchies than quite a few people that are running on vista with the same chip.

in other news...Mmmm Pie, I likes me some Pecan Pie. Pie pie pie pie piiieee.


Here's the solution: http://www.xtremesystems.org/forums/showthread.php?t=18...

Vista SP1 always enable the TLB fix, no matter the stepping of the Phenom.
a c 127 à CPUs
August 21, 2008 7:48:02 PM

iocedmyself said:
In soviet russia drugs do YOU!

So with this chip the design changesintel made use of are
AMD's
IMC
Hypertransport (no it's not! no it's not! it's quick connect! new name! make it different!)
Native-quad core die - more than a year after amd
x86-64 instructons - because why should they bother when 64bit isn't used
independent core clock/voltage control - oh yeah well so can we!

with a splash of Gigabytes CIA dynamic clock control - because it helps justify our price!

Smaller die, updated cache...and 3 way memory linking tossed in with some good ol' reverse engineered features.

Seriously, 3 way memory? Do they think it sounds kinky? They have been on their soap box for 5 years saying how 64bit is nothing more than a marketing strategy, along with the rest of the chips features, but they have focused only on 32bit performance, in 32bit OS's where the system is limited to 2 gigs application accessable memory.

They are like a guy at the play ground trying to get someone in his van with candy...."no no, don't pay attention to what was said 30 seconds ago...i'm not a stranger, you know me, i want to give you this yummy candy"

Btw, when larrabee is another 6-9 months away from offering an 8 core, 2ghz card that delivers UP TO 256 Gigaflop/sec (16 flop/sec per clock/ per core) haven't made a good IGP, let alone actual gaming gpu, and larrabe hinges not only on being coded specially for the card, but their own driver optimization...

being able to deliver performance on par with single core 4870's 1200 gigaflop/sec in the end of 2010 with either a 24 core 3ghz card, or 32 core 2 ghz card...it's a joke. It's based on the logic of, "dur, if a gpu is 20 times faster than a cpu, then 20 cpu's should be as fast as a single gpu."

20% even 30% performane gain still does not make nehalem worth the 200-500% price priemium for a system.


WHat the hell are you talking about? Rambling about BS that can easily be changed for anything? Seriously just LET IT GO!!!!! Intel has been working on a IMC sine oh the 386 and has had a few out as well but never fully marketed it. AMD was the first to fully market the IMC. You lose the FSB when you move the MC to the CPU so what do you do? You create the fastest freaking connection to memory and other components you can and thus far Intel is smart using high bandwidth tri channel DDR3, I think you are mad AMD didn't think of it first.

But seriously the AMD argument is getting old. In the PC industry EVERYONE compies EVERYONE. ATI had a Unifide Shader arch first allowing for programmable shaders. Then nVidia adopted the same idea. Anyone complain there? No. SO just stop using the same argument thats just your BS opinion and preference.

$300 for the 2.66GHz CPU is not that bad, DDR3 is not as bad as people let on and I am sure the decent mobos will come out at about $200. So lets see what AMD will charge for their new CPU and IF you are lucky the mobo maker will decide to update the mobos so that Deneb is supported so people wont have to get a new mobo. If not you are SOL.

BaronMatrix said:
Now that was funny. It does sound like Intel though. I tell people they'r like the little kid who gets mad and tries to take the ball home. Now it turns out it's not even their ball (X86 CPU). I too am interested to see Larrabee fulfill the claims. I mean to say their first GPU will be as fast as 4000 or GT200 is like wow, you can barely draw Aero or HD and you've had what 5 years.

But anyway, I may end up haing an even better studio setup if Deneb is timely. Next month is upgrade time finally.


Seriously BM you are just a piece of work. You just bi*ch about the same thing consitently. Consider that Larrabee is NOTHING like their IGPs. That changes EVERYTHING.

God the BS that is spouted on a daily basis is just unbearable. This is supposed to be a thread about Deneb that changed to cheesecake and it just had to be ruined by a rambling of "But but but.... they had it first...." wah wah, boo hoo.
August 21, 2008 8:06:37 PM

Until theres more proof, its speculation. How many speculation threads have wandered? Hell, how many serious, good and informative threads have wandered? I agree, Larrabee will have to show me before I get excited for it. AMD has been billions of dollars more sucessful than Intel using a IMC, but who cares?
August 21, 2008 9:08:44 PM

JAYDEEJOHN said:
Until theres more proof, its speculation. How many speculation threads have wandered? Hell, how many serious, good and informative threads have wandered? I agree, Larrabee will have to show me before I get excited for it. AMD has been billions of dollars more sucessful than Intel using a IMC, but who cares?


Once more i stand by JDJ in this matter. The Cheese cake, homer simpson, pie, lime discussion was far more interesting.

And funny too.
!