AMD Shanghai/Deneb Review Thread

[iocedmyself]

They've been working on this for years and tape out had to of been at least over a year, I'm sure they know what their process can do by now. I think that it's too close to release to debate so I guess we will see.

Yes they had been working on agena for a couple years as well, and TLB kinda crushed it in most peoples eyes, even after that issue was fixed the chips still ran quite hot and the 750SB ACC solution was put into the mix. But still the chips ran quite hot even though they were expected to do clock higher and run much cooler.

I tend to agree with UJ's assessment of "better to promise less and deliver more" so they don't have another Agena fiasco, but then again you have to consider it could just have been a way to get the jump on Intel. Intel really wouldn't have reason to blink at another 140w chip by AMD regarded mainly as a die shrink of Agena, but they might have something to say about a

• 50-60w 3ghz cpu with 20%+ ipc
• clocks easily to 4ghz on air with 90-100w TDP putting power consumption at half of their 3.5 ghz cpus.
• 5ghz+ clocks attainable with water, higher with cold
• Backwards compatible with existing platforms, making a serious upgrade $200-300 instead of $800-1000 whether it's for gamers, servers, workstations, or power minded consumers
• Already will support the next round of motherboards with DDR3, so the few hundred buck you spent on your cpu will last you longer than 4 months.

As pretty much all of the phenom II's have been rated at 125-140w i tend to think it was in preperation for the 6 core desktops, which seem more likely to hit closer to that mark

Then you have the mobile angle to consider. Could be the mobile barton days making a come back. 60w may be high for a laptop part, but drop the voltage to .7-.8v and clock down to 2ghz and you're looking at more like a 25-35w part that is high performing and battery friendly. As i can do 3.3ghz at 1.0875v i tend to think the voltage could be dropped a fair amount below that to stay at 2ghz or clock raised at my estimated volts.

Intel won't have much of an answer to that, can you picture a nehalem making it into a laptop anytime soon? Perhaps an atom chip catching up with performance? Wow, even as i typed that my sides hurt from laughter.


@sundancerx

Yes i love the irony of the good luck i have when my black cat is near while i overclock. He must feel bad about almost killing the chip As amusing as it would be to have Intel spin it to give credit to the kitty, i half expect the intel labs will start keeping their own animals near when testing chips. I can see the advertisements now.

A shot of a gerbil wearing a tiny sweater reading "Intel", with "Intel inside" in friendly font below.

But thank you for the applause if i'm not mistaken that alone puts me at top Air clock speed, stable or not

[gallag]

Yes they had been working on agena for a couple years as well, and TLB kinda crushed it in most peoples eyes, even after that issue was fixed the chips still ran quite hot and the 750SB ACC solution was put into the mix. But still the chips ran quite hot even though they were expected to do clock higher and run much cooler.

I tend to agree with UJ's assessment of "better to promise less and deliver more" so they don't have another Agena fiasco, but then again you have to consider it could just have been a way to get the jump on Intel. Intel really wouldn't have reason to blink at another 140w chip by AMD regarded mainly as a die shrink of Agena, but they might have something to say about a

• 50-60w 3ghz cpu with 20%+ ipc
• clocks easily to 4ghz on air with 90-100w TDP putting power consumption at half of their 3.5 ghz cpus.
• 5ghz+ clocks attainable with water, higher with cold
• Backwards compatible with existing platforms, making a serious upgrade $200-300 instead of $800-1000 whether it's for gamers, servers, workstations, or power minded consumers
• Already will support the next round of motherboards with DDR3, so the few hundred buck you spent on your cpu will last you longer than 4 months.

As pretty much all of the phenom II's have been rated at 125-140w i tend to think it was in preperation for the 6 core desktops, which seem more likely to hit closer to that mark

Then you have the mobile angle to consider. Could be the mobile barton days making a come back. 60w may be high for a laptop part, but drop the voltage to .7-.8v and clock down to 2ghz and you're looking at more like a 25-35w part that is high performing and battery friendly. As i can do 3.3ghz at 1.0875v i tend to think the voltage could be dropped a fair amount below that to stay at 2ghz or clock raised at my estimated volts.

Intel won't have much of an answer to that, can you picture a nehalem making it into a laptop anytime soon? Perhaps an atom chip catching up with performance? Wow, even as i typed that my sides hurt from laughter.


@sundancerx

Yes i love the irony of the good luck i have when my black cat is near while i overclock. He must feel bad about almost killing the chip As amusing as it would be to have Intel spin it to give credit to the kitty, i half expect the intel labs will start keeping their own animals near when testing chips. I can see the advertisements now.

A shot of a gerbil wearing a tiny sweater reading "Intel", with "Intel inside" in friendly font below.

But thank you for the applause if i'm not mistaken that alone puts me at top Air clock speed, stable or not

I am willing to bet money that no Intel or AMD quad cpu will be running at 4ghz and only pulling 90-100w tdp this side of 2010. And 20% ipc?

[TL1000S]

I am willing to bet money that no Intel or AMD quad cpu will be running at 4ghz and only pulling 90-100w tdp this side of 2010. And 20% ipc?

He did not relate those 20% to something.
Could be Barton you know.
Probably compared to Agena.

[G0ldBr1ck]

Those temps and the TDP that it hints at are very shocking. Even if your calculations are way off and the TDP was at 140 at those clocks and volts it would be amazing. This definitely hints at the possibility of some really high clocked FX models in the near future. Good Job iocedmyself

[gallag]

He did not relate those 20% to something.
Could be Barton you know.
Probably compared to Agena.

But deneb is not 20%ipc+ agena. Seriously, People are really setting expectations high, If I believed every thing that I am reading Deneb will oc to at least 4ghz on air, be 20% faster per clock and only consume 90watts and hit max 40c

Say its true and deneb only has 95tdp @4ghz how many hundred's or thousand's percent better is the performance per watt of deneb over agena? What are you guys expecting from a die shrink? The only review so far, which people are happy enough to link to for showing the strengths of deneb (which there are many) also show that power consumption is not so good.

[imsochobo]

All theese calculations are done on the silver chips (3.5 ghz for 110 W) isnt a impossible statement.
140 W for 4ghz, surely its possible if the frequency doesnt mess too much with the numbers.

If i was amd.
FX models 3.5 GHZ
Higher HT link.

The higher ht link benefits alot, and doesnt increase the TPD all too much.
I didnt see any increase in temps here, while 50 mhz increases temp alot( water+phenom 1 not 2)

But depends if amd wants that war, mhz war, its given nothing for intel @ the last war, and surely, pure mhz isnt the aim.
Higher performance through HT link increase.
Revisions will fix little by little probaly, its a quite new architecture, and will surely get a little bump in next revision.

[G0ldBr1ck]

All theese calculations are done on the silver chips (3.5 ghz for 110 W) isnt a impossible statement.
140 W for 4ghz, surely its possible if the frequency doesnt mess too much with the numbers.

If i was amd.
FX models 3.5 GHZ
Higher HT link.

The higher ht link benefits alot, and doesnt increase the TPD all too much.
I didnt see any increase in temps here, while 50 mhz increases temp alot( water+phenom 1 not 2)

But depends if amd wants that war, mhz war, its given nothing for intel @ the last war, and surely, pure mhz isnt the aim.
Higher performance through HT link increase.
Revisions will fix little by little probaly, its a quite new architecture, and will surely get a little bump in next revision.

Where did you come up with the Higher HT link having large benefits? do you mean NB/IMC/L3

[Uncle Jimbo]

They've been working on this for years and tape out had to of been at least over a year, I'm sure they know what their process can do by now. I think that it's too close to release to debate so I guess we will see.

The Barton processors originally came out with a max 'rating' of 2600. A year later that was up to 3200. I don't know how much you understand about semiconductor manufacturing, but the process tuning to get yields of certain parts is not the same as the chip design. So if they want to shift yield balance, they might do minor tweaks to increase yields of higher speed parts. If, on the other hand, they see a bigger volume need for low power chips, and there is a market for more than the bin split gives, they may adjust the process to increase those yields.

I think AMD is more interested in tuning the process for low power than for high speed.

[ajaidev]

Quite rite Uncle Jimbo, AMD high end processors have always been tuned base level architectures with minor differences and as low TDP rise as possible but Intel high ends such as i7 965 are more hand picked silicon wafers.

Also as said AMD's are lower clocked so that they eat less energy and that means that for normal consumers will be happy with it...!!

[demonkevy666]

this is the first time AMD is lunching NEW CPU @ 3.0ghz!!

they've never done that,before. clearly this a change for AMD

they usually start off with the lowest bin they got.

and oh look lower volts at same speed

Seems to undervolt rather well..

2.8ghz at 1.184 volts

[imsochobo]

Where did you come up with the Higher HT link having large benefits? do you mean NB/IMC/L3

Tested it, seen test of it, and it yield big improvements in games, especially graphic intensive games.
Since its a link between some stuff, what im hinting for, is link between chipset cpu and graphics card.

It yield about 5% from 2000 to 2400 mhz maybe, and well, quite good when you talking performance improvement, the PH2 isnt rated for more than 5% faster than PH1 clock for clock, and thats actually alot for cpu's.

Maybe not so much that core 2 had over the pentium 4, but, well well. still alot.

[gallag]

and oh look lower volts at same speed



2.8ghz at 1.184 volts

Stable? If so that's pretty impressive.

[Astennu]

Tested it, seen test of it, and it yield big improvements in games, especially graphic intensive games.
Since its a link between some stuff, what im hinting for, is link between chipset cpu and graphics card.

It yield about 5% from 2000 to 2400 mhz maybe, and well, quite good when you talking performance improvement, the PH2 isnt rated for more than 5% faster than PH1 clock for clock, and thats actually alot for cpu's.

Maybe not so much that core 2 had over the pentium 4, but, well well. still alot.

Kendsfield > Penryn and Penryn > Nelhalem had bigger improvments.

The first had a 5-15% performance boost in most cases. The Second goes from -5% up to + 40%.

I was hoping for a bigger performance boost for PHII. So they could at least beat Penryn clock for clock. Now They only closed the gap. A Q9550 will be on par with the 940 BE.

[Oldguy932]

The performance increase you're seeing from a higher ht-link is actually because of the higher nb speed. For the ht link to be at 2.6ghz, the nb has to be at 2.6ghz or higher. Try looking at results between 2.6ghz for nb and ht-link and then also 2.6ghz nb and 2ghz ht link. There will be very little difference if at all from changing the speed of the ht link, nb has a much more dramatic effect on performance than that.

[Astennu]

The performance increase you're seeing from a higher ht-link is actually because of the higher nb speed. For the ht link to be at 2.6ghz, the nb has to be at 2.6ghz or higher. Try looking at results between 2.6ghz for nb and ht-link and then also 2.6ghz nb and 2ghz ht link. There will be very little difference if at all from changing the speed of the ht link, nb has a much more dramatic effect on performance than that.

Indeed i did those tests with my X3. With 2.2 NB and 2.2 HTT and 1.2 GHz HTT. Could not find big changed in 3DMark 2001 SE scores. And if there is a test that wil notice those things its 3Dmark 2001. Because its cpu + system limited.

[demonkevy666]

Kendsfield > Penryn and Penryn > Nelhalem had bigger improvments.

The first had a 5-15% performance boost in most cases. The Second goes from -5% up to + 40%.

I was hoping for a bigger performance boost for PHII. So they could at least beat Penryn clock for clock. Now They only closed the gap. A Q9550 will be on par with the 940 BE.

what is with every one using the york field that has 12mbs ? would not the 6mb enough?

4mb > 2mbs

[knopflerbruce]

Stable? If so that's pretty impressive.

That's about the same as my old Opty 175 (2.84ghz @ 1.19v). Perhaps these K10's really scale with voltage. I haven't played around with quads much yet, the only chip I have is a crapoy, old B0 Agena ES that runs at 1.6ghz... so i have no clue, really. But 2.8 @ 1.18v doesn't seem THAT good to me. I'm sure ANY F3 Windsor can do that.

[gallag]

[*]50-60w 3ghz cpu with 20%+ ipc [*]clocks easily to 4ghz on air with 90-100w TDP putting power consumption at half of their 3.5 ghz cpus.[*]5ghz+ clocks attainable with water, higher with cold[*]Backwards compatible with existing platforms, making a serious upgrade $200-300 instead of $800-1000 whether it's for gamers, servers, workstations, or power minded consumers [*]Already will support the next round of motherboards with DDR3, so the few hundred buck you spent on your cpu will last you longer than 4 months.[/LIST]

As pretty much all of the phenom II's have been rated at 125-140w i tend to think it was in preperation for the 6 core desktops, which seem more likely to hit closer to that mark

Then you have the mobile angle to consider. Could be the mobile barton days making a come back. 60w may be high for a laptop part, but drop the voltage to .7-.8v and clock down to 2ghz and you're looking at more like a 25-35w part that is high performing and battery friendly. As i can do 3.3ghz at 1.0875v i tend to think the voltage could be dropped a fair amount below that to stay at 2ghz or clock raised at my estimated volts.

Intel won't have much of an answer to that, can you picture a nehalem making it into a laptop anytime soon? Perhaps an atom chip catching up with performance? Wow, even as i typed that my sides hurt from laughter.


@sundancerx

Yes i love the irony of the good luck i have when my black cat is near while i overclock. He must feel bad about almost killing the chip As amusing as it would be to have Intel spin it to give credit to the kitty, i half expect the intel labs will start keeping their own animals near when testing chips. I can see the advertisements now.

A shot of a gerbil wearing a tiny sweater reading "Intel", with "Intel inside" in friendly font below.

But thank you for the applause if i'm not mistaken that alone puts me at top Air clock speed, stable or not

3.85ghz @ 170w
http://www.xtremesystems.org/forums/...d.php?t=213644
And crazy thing is the temps where reported as 47c and I dont think I need to tell anyone here that a cpu burning 170w is going to be that cool on air, Faulty temp probes on deneb?

[Mr.BSEL]

The review is up at Hexus

http://www.hexus.net/content/item.php?item=16757&page=1

[Oliverda]

The review is up at Hexus

http://www.hexus.net/content/item.php?item=16757&page=1

"Sorry, but the URL you followed to get to this item is invalid."

They hitted by AMD commando.

[gallag]

The review is up at Hexus

http://www.hexus.net/content/item.php?item=16757&page=1

got a few pages in and it seems they took it down

[Mr.BSEL]

Yeah it looks like they took it down

[gallag]

Yeah it looks like they took it down

did you read it all?

[ajaidev]

Dam the review is down, AMD ninja cum commando's killed the servers hdd and they had to do a restore.

[demonkevy666]

Dam the review is down, AMD ninja cum commando's killed the servers hdd and they had to do a restore.

I read it lol I must be a ninja it said I7 till rules high end, and all alone up there. Deneb beats the Q9300 and puts a good fight, higher power consumption for platform.

wonder how they got higher power consumption...