AMD Shanghai/Deneb Review Thread

[Lightman]

I've just did a quick Vantage run on my Phenom setting it to 3GHz CPU, 2277MHz NB, 1110MHz RAM 5-5-5-12-18 2T, HTT 207MHz and got:
CPU Test1: 1263
CPU Test2: 15.99

Figure it out

[spursindonesia]

I dont know, maybe intel thinks they dont need to drop prices.

Take a look around, how many guys already have 45nm quad ready 775 boards here.

When the 940 is ~225€ (incl tax) you still need a board for the PII. A decent 790GX board is ~100€ which would put the whole package to ~325€.

A Q9550 is ~290€, so its cheaper to just drop in a Q9950 into a 775s board then to buy a 940 and a board.

The only people that really benefit from the 940 are the quys that already have a AM2(+) board or come from older platfroms (S939 or 775 not compatible with Quads) or buy a whole new pc.

For my case and my country context, going for a platform change isn't that hurtful, since my X38 board -DFI LanParty UT X38-T2R has a pretty good secondhand price, relatively comparable to a new, highend 790GX board. And ofcourse, supporting the underdog should they have a competitive product always carries a premium in my decision makin process.

[mrcmtl]

I've just did a quick Vantage run on my Phenom setting it to 3GHz CPU, 2277MHz NB, 1110MHz RAM 5-5-5-12-18 2T, HTT 207MHz and got:
CPU Test1: 1263
CPU Test2: 15.99

Figure it out

nice, although can u run it at 3ghz/1.8ghz NB/800mhz ram? just to be 100% fair

[keithlm]

nice, although can u run it at 3ghz/1.8ghz NB/800mhz ram? just to be 100% fair

And timings of 6-6-6-18-24 if you really want to compare it to the "leaked" Deneb result.

[JumpingJack]

Was hoping for great competition.

I think you are going to get your wish ... AMD has been clear that they are aiming for the sweet spot (middle of the curve), Deneb appears to be hitting right there, so AMD will be able to increase ASPS, offer up great performance at the price they are aiming for....

This is 10x better overall (not performance comparative, but competitive comparative) than Agena.

Jack

[dinos22]

Ugh, that benchmark with 4x4GB sticks, DDR800 6-6-6 and 1.8Ghz NB... What

This ain't really reliable

[EDIT]
Prolly OT, but I just got a mail saying my pre-order is expected to arrive 18-12-2008! WTF!? 'n awesome

prepare for a disappointment

i doubt they would allow distributors to sell before official launch`

[dread77]

PhenomII x4 940 on air cooling @ 3,8...

[Ghostbuster]

PhenomII x4 940 on air cooling @ 3,8...

Why blank out the voltage? The Turkiye one didn't...

[Jacky]

I think you are going to get your wish ... AMD has been clear that they are aiming for the sweet spot (middle of the curve), Deneb appears to be hitting right there, so AMD will be able to increase ASPS, offer up great performance at the price they are aiming for....

This is 10x better overall (not performance comparative, but competitive comparative) than Agena.

Jack

Are you really sure? I believe it's a big improvement over K10, yes, but the K10.5 die still is huge and I don't know how good the new process yields? I don't know if they can improve overall margins *that* much. I'd guess margins for desktop ++, 2P server --, others flat. Just a guess.

[duploxxx]

PhenomII x4 940 on air cooling @ 3,8...

what's up with that ht link, the 940 isn't at 1800 at all it should be much higher, or is this still under NDA

[leoy]

At 3.8 it is absoloutely stable... on air...
And the temps are really well...
From oktabits forum...

[bedlamite]

940 NB will be 1,8 GHz.
945 will be 2400 MHz...

[informal]

940 NB will be 1,8 GHz.
945 will be 2400 MHz...

Hmm are you 100% sure?
1.8Ghz is way too low,even 9850 65nm part has 11% faster NB.

[bedlamite]

That's what I've heard today from some sources
But I'll see it with my own eyes soon

[G0ldBr1ck]

Hmm are you 100% sure?
1.8Ghz is way too low,even 9850 65nm part has 11% faster NB.

Maybe they just want there to be more OC headroom for marketing reasons. A 50% HT oc sounds better than 20%.

lol, actually that sounds pretty stupid but might have something to do with it.
It would also help create another reason for people to upgrade to AM3 945 otherwise not much to gain from 940 to 945 upgrade.

[Macadamia]

940 NB will be 1,8 GHz.
945 will be 2400 MHz...

Aww. I was about to ask if these were going to be ESes.

The last time AMD sent out ESes they didn't match the L3 clock at production (but the ES was clocked higher, 2.2Ghz IIRC)

[informal]

That's what I've heard today from some sources
But I'll see it with my own eyes soon

I dunno man.Take a look at the previous 65nm and released 45nm Shanghai parts. The Core clock : NB clock ratio never went above 1.3.
For example: Phenom 9350 model has a ratio of 2.3/1.8=1.27; Phenom 9850 2.5/2=1.25; Phenom 9950 2.6/2=1.27;Shanghai 2384/8384 model has a ratio of 2.7/2.2=1.22

Now compare those ratios to unreleased Phenom II chips ratios: Phenom II 920 2.8/1.8=1.55 (way too high ratio);Phenom II 940 3/1.8=1.66 (way way to high ratio) double .

It is entirely possible desktop parts will receive these NB clocks,but in the light of Shanghai parts,i'm very skeptical...

What would be "normal" NB clocks if we follow the trend seen in Phenom I and Shanghai? It would go like this:

Phenom II 920 model : 2.8/1.22=2.3Ghz in best case or 2.8/1.27=2.2Ghz in "worst" case.The 2.2Ghz would be my bet following the rule of integer NB multies,Shanghai NB clocks and 1.27x ratio seen from Phenom 9950.So 2.2Ghz for 920 model.

Phenom II 940 model : 3.0/1.22=2.45Ghz in best case or 3.0/1.27=2.36.This time i would say the NB would probably be between these two frequencies, following the rule of 200Mhz increments and integer multies. So 2.4Ghz for 940 model.

Like i said previously,maybe the retail parts will indeed have 1.8Ghz NB clock so the above "math" is pointless.But even so,it makes sense to me.

[Oliverda]

940 NB will be 1,8 GHz.
945 will be 2400 MHz...

I think that it is believable. The higher NB clocks for DDR3 is more necessary than for DDR2.

http://www.bit-tech.net/news/2008/11...ed-for-ocing/1

"Theo also claims AMD's latest AM3 Phenom II CPUs generate 16GB/s from dual channel DDR3 1,333MHz. While this is clearly less than Intel's triple channel malarkey with Core i7, we've seen in our initial Core i7 review it does little to further performance over dual channel."

[duploxxx]

I dunno man.Take a look at the previous 65nm and released 45nm Shanghai parts. The Core clock : NB clock ratio never went above 1.3.
For example: Phenom 9350 model has a ratio of 2.3/1.8=1.27; Phenom 9850 2.5/2=1.25; Phenom 9950 2.6/2=1.27;Shanghai 2384/8384 model has a ratio of 2.7/2.2=1.22

Now compare those ratios to unreleased Phenom II chips ratios: Phenom II 920 2.8/1.8=1.55 (way too high ratio);Phenom II 940 3/1.8=1.66 (way way to high ratio) double .

It is entirely possible desktop parts will receive these NB clocks,but in the light of Shanghai parts,i'm very skeptical...

What would be "normal" NB clocks if we follow the trend seen in Phenom I and Shanghai? It would go like this:

Phenom II 920 model : 2.8/1.22=2.3Ghz in best case or 2.8/1.27=2.2Ghz in "worst" case.The 2.2Ghz would be my bet following the rule of integer NB multies,Shanghai NB clocks and 1.27x ratio seen from Phenom 9950.So 2.2Ghz for 920 model.

Phenom II 940 model : 3.0/1.22=2.45Ghz in best case or 3.0/1.27=2.36.This time i would say the NB would probably be between these two frequencies, following the rule of 200Mhz increments and integer multies. So 2.4Ghz for 940 model.

Like i said previously,maybe the retail parts will indeed have 1.8Ghz NB clock so the above "math" is pointless.But even so,it makes sense to me.

i agree with you, as far as i am aware in latest docs its 2,2 and 2,4 depending on speed, even shanghai was launched with parts on 2,2 and 2,2 so i don't see why it would drop again to 1,8 unless we are talking about future desktop E and server HE parts.

[Oliverda]

i agree with you, as far as i am aware in latest docs its 2,2 and 2,4 depending on speed, even shanghai was launched with parts on 2,2 and 2,2 so i don't see why it would drop again to 1,8 unless we are talking about future desktop E and server HE parts.

Maybe AMD want to leave some headroom for improvement on AM3.

[informal]

I think that it is believable. The higher NB clocks for DDR3 is more necessary than for DDR2.

http://www.bit-tech.net/news/2008/11...ed-for-ocing/1

"Theo also claims AMD's latest AM3 Phenom II CPUs generate 16GB/s from dual channel DDR3 1,333MHz. While this is clearly less than Intel's triple channel malarkey with Core i7, we've seen in our initial Core i7 review it does little to further performance over dual channel."

I think Theo was speaking about AM2+ version .
Also i think DDR3/AM3 Denebs will not need higher NB clocks more than DDR2/AM2+ ones.
In any case,the NB clocks never crossed the 1.3 ratio mark like i've shown in my previous post.Not even on Shanghai 45nm parts(which are socketF,not F+/HT3.0).

[AbelJemka]

Link to new CPUZ:
http://www.xtremesystems.org/forums/...d.php?t=209806
Link to new Hall of Fame :
http://valid.canardpc.com/records.php
Link to world record frequency K10:
http://valid.canardpc.com/show_oc.php?id=448873

See also :
http://valid.canardpc.com/show_oc.php?id=452168

[Ghostbuster]

Poser....

If the AMD demo can show Phenom II running Crysis at 5GHz, then why we haven't seen any Phenom II 5GHz+ validation yet?

[AbelJemka]

First i don't think this guy have right to valide CpuZ cause of NDA.
Second i just post link i make zero assumption contrary to u.
if u can conclude base on this link that AMD can't do 5 GHz... you are very good, very, very good

[informal]

First i don't think this guy have right to valide CpuZ cause of NDA.
Second i just post link i make zero assumption contrary to u.
if u can conclude base on this lin that AMD can't do 5 GHz... you are very good, very very good

Exactly.Plus,if I had to guess,I'd say all of those submitted clocks are done with air/wc.