Turion vs Sonoma

[pcchen]

In that article, it is also mentioned that there's a "J" alphabet before the S-spec on the box. They think it means that these new Pentium M also supports NX bit.

About Alviso... I don't know what you mean about "leaked." There are a lot information about what Alviso is.

 

[ANova]

Yes, the J suffix means they support NX.

 

[pcchen]

Sharp unveils Sonoma based notebook
News Release
(Both in Japanese)

In the news release, Sharp confirmed that the new Pentium Ms do support NX bit. These notebooks use i915GM chipset.

 

[Bjorn]

Tom's have done some tests with the new Alviso chipset for the Pentium M, and apparently , there won't be any large gains with the faster FSB and DDR2 memory.

http://www.tomshardware.com/hardnews/20050119_110843.html

Real world performance gains from increased FSB speed, PCI Express and DDR2 memory will be limited, as first tests in our test lab revealed.

That was actually what i suspected also since the Pentium M has such a large cache to compensate for the slower FSB.

 

[T2k]

YEah and also to have a speedy memory at hand, instead of the slow (high latency) system memory.

Well... shouldd I say just as I suspected? This was my first, topic starter post here:

AMD is getting ready its new, uber-mobile chip, called Turion whereas Intel is busy upgrading current Dothan to already widely used faster FSB and memory architecture.

Question is whether this upgrade will be enough to stand against AMD's new part? They are wroking on this Turion for a very long time now, rumours say it's going to be some overkill stuff - I guess we can be sure it's got integrated memory controller as well as 64bit etc...

It seems we have an answer already, unless AMD will screw up (unlikely) like Intel did this time...

 

[Accord1999]

It seems we have an answer already, unless AMD will screw up (unlikely) like Intel did this time...

What answer, it doesn't look like Turion (which looks like it's just a low voltage A64) will arrive till May with specified power consumption levels that are still too high to compete with Dothan.

 

[T2k]

It seems we have an answer already, unless AMD will screw up (unlikely) like Intel did this time...

What answer, it doesn't look like Turion (which looks like it's just a low voltage A64) will arrive till May with specified power consumption levels that are still too high to compete with Dothan.

Well... what I heard it's quite different story...

 

[Bjorn]

It seems we have an answer already, unless AMD will screw up (unlikely) like Intel did this time...

What answer, it doesn't look like Turion (which looks like it's just a low voltage A64) will arrive till May with specified power consumption levels that are still too high to compete with Dothan.

AMD has specified their dual core 1.6 GHz CPU's at 30W under full load. And their current 90 nm A64's consume about 31W at load (3500+). So i'm thinking that they will be able to compete rather well with the Dothan in that department.

 

[pcchen]

Actually power under load is not the most important issue of a mobile CPU. It usualy more about how much power it can save under sleep/semi-sleep mode. For example, Sonoma's TDP under 800MHz and low vcc is about 10.8W. In deep sleep mode Sonoma can reduce its max power dissapation to a bit more than 1 W. AMD may still have some to catch up since Intel has worked so much on Pentium M cores.

 

[Bjorn]

Actually power under load is not the most important issue of a mobile CPU. It usualy more about how much power it can save under sleep/semi-sleep mode. For example, Sonoma's TDP under 800MHz and low vcc is about 10.8W. In deep sleep mode Sonoma can reduce its max power dissapation to a bit more than 1 W. AMD may still have some to catch up since Intel has worked so much on Pentium M cores.

Tom's tests indicated that the power usage under idle was 3W for the 90 nm A64's.

http://www.tomshardware.com/cpu/20041115/pentium4_570-20.html

 

[Fox5]

Tom's have done some tests with the new Alviso chipset for the Pentium M, and apparently , there won't be any large gains with the faster FSB and DDR2 memory.

http://www.tomshardware.com/hardnews/20050119_110843.html

Real world performance gains from increased FSB speed, PCI Express and DDR2 memory will be limited, as first tests in our test lab revealed.

That was actually what i suspected also since the Pentium M has such a large cache to compensate for the slower FSB.

Or maybe the cpu just isn't powerful enough to make use of the extra bandwidth, pentium m is p3 based right? In fact, I've even seen benchmarks were a 1.4ghz p-m performs exactly the same as a 1.4ghz p3(and both lose handidly to an athlon at 1.4ghz), yet the p-m has both the memory bandwidth and cache advantage.(the one used had a 1MB cache) The cache may just help with programs that branch very heavily, as games appear to be about the only applications where pentium ms can compete with the very top of the line ahtlon 64s and pentium 4s.

BTW, how did tom's hardware measure the power draw of the cpus?

 

[Accord1999]

Actually power under load is not the most important issue of a mobile CPU. It usualy more about how much power it can save under sleep/semi-sleep mode. For example, Sonoma's TDP under 800MHz and low vcc is about 10.8W. In deep sleep mode Sonoma can reduce its max power dissapation to a bit more than 1 W. AMD may still have some to catch up since Intel has worked so much on Pentium M cores.

Tom's tests indicated that the power usage under idle was 3W for the 90 nm A64's.

http://www.tomshardware.com/cpu/20041115/pentium4_570-20.html

And that 2W difference at idle can represent 10-20% better battery life. And x86-secret.com measured a 2GHz Dothan 755's full load at 16W as well.
http://www.x86-secret.com/index.php?option=articles&Itemid=3&topid=2#

x86-secret also has the rumored TDP of the Turions:
http://www.x86-secret.com/?option=newsd&nid=830

At those TDPs, the Dothans are clearly superior in performance/watt.

 

[Fox5]

Actually power under load is not the most important issue of a mobile CPU. It usualy more about how much power it can save under sleep/semi-sleep mode. For example, Sonoma's TDP under 800MHz and low vcc is about 10.8W. In deep sleep mode Sonoma can reduce its max power dissapation to a bit more than 1 W. AMD may still have some to catch up since Intel has worked so much on Pentium M cores.

Tom's tests indicated that the power usage under idle was 3W for the 90 nm A64's.

http://www.tomshardware.com/cpu/20041115/pentium4_570-20.html

And that 2W difference at idle can represent 10-20% better battery life. And x86-secret.com measured a 2GHz Dothan 755's full load at 16W as well.
http://www.x86-secret.com/index.php?option=articles&Itemid=3&topid=2#

x86-secret also has the rumored TDP of the Turions:
http://www.x86-secret.com/?option=newsd&nid=830

At those TDPs, the Dothans are clearly superior in performance/watt.

AMD had 35w 130nm athlon 64s, you'd think that they could get much lower by handpicking the best 90nm athlon 64s. Or maybe they've decided that they're low enough and they'll just go for lowering cost and increasing supply now.

 

[Bjorn]

And that 2W difference at idle can represent 10-20% better battery life. And x86-secret.com measured a 2GHz Dothan 755's full load at 16W as well.
http://www.x86-secret.com/index.php?option=articles&Itemid=3&topid=2#

Perhaps, but we're talking about a desktop CPU. And they mention 1W at "deep sleep mode". Is that the necessarily the same as idle in the case of the A64 ?

x86-secret also has the rumored TDP of the Turions:
http://www.x86-secret.com/?option=newsd&nid=830

At those TDPs, the Dothans are clearly superior in performance/watt.

So the Turions would use more power (at least when comparing the 2GHz versions) then the desktop 90 nm A64's ? I somehow doubt that.

 

[Mariner]

Another thing to consider is that there are reports of some people considerably undervolting their A64s successfully. I've seen reports of people running their 90nm A64s at 2Ghz with a voltage of just 1.1V as compared to the standard 1.4V! This would decrease power consumption markedly.

I expect Turion to be little more than the forthcoming 'E' revision which reputedly has strained silicon to further reduce power consumption. That's just my guess, however!

I do consider both the Pentium M and A64s to be excellent chips. It's just a pity that the Pentium Ms are so expensive at the moment.

 

[pcchen]

It is essential to lower down the voltage while reducing the frequency to reduce power. Sonoma uses 0.988V when running at low frequency mode (800MHz).

I think AMD will be able to come close to the power consumption of Dothan/Sonoma with Turion, but probably won't be able to do better. Many parts of Banias core are redesigned for lower power consumption, and I doubt that AMD has enough resource for a major core redesign.

But if Turion is that better than current AMD's mobile offering, I think it will have a nice effect on the pricings of Pentium M processors, making good notebooks more affordable.

 

[Mariner]

I think as standard the current revision of A64s run at 1Ghz and 1.1V when in "Cool-n-Quiet" mode. They are capable of this speed at much lower voltage, however - some chips can manage it as low as 0.85V!

Edit: Actually, I've just noticed that the 0.85V figure mentioned was achieved on a 130nm Newcastle Athlon64. I suppose it is possible that the 90nm chips could run at even lower voltage at the same speed.

I'd expect a 'tweaked' CnQ in the Turion when it finally appears. As you mention, we need some competition in this area to push down prices!

 

[Sxotty]

Now that some info is available on these, I am thinking that they would make some killer chips on the desktop front. I have a 2600mobile in my machine that runs @ 3200 speeds below normal desktop voltage. I have clocked it at 2.6Ghz, but it was too hot (my room that is).

Anyway these new mobile chips could make great desktop overclockers...near 3Ghz would be groovy for

AMD Turion 64 mobile technology models ML-37, ML-34, ML-32, ML-30, MT-34, MT-32, and MT-30 are priced at $354, $263, $220, $184, $268, $225 and $189 respectively, in 1,000-unit quantities."

 

[{Sniping}Waste]

Here is a first look at Turion.
http://xtremesystems.org/forums/showthread.php?t=55174

You can buy Oakviles now. A 2800+ Oakvile 35W is about $120.
http://www.memoryexpousa.com/amdatmo28754.html

 

[swaaye]

I think as standard the current revision of A64s run at 1Ghz and 1.1V when in "Cool-n-Quiet" mode. They are capable of this speed at much lower voltage, however - some chips can manage it as low as 0.85V!

Edit: Actually, I've just noticed that the 0.85V figure mentioned was achieved on a 130nm Newcastle Athlon64. I suppose it is possible that the 90nm chips could run at even lower voltage at the same speed.

I'd expect a 'tweaked' CnQ in the Turion when it finally appears. As you mention, we need some competition in this area to push down prices!

I use RMClock to do just this on my Clawhammer notebook. It will run 800Mhz @ .85v and full 1.80Ghz at 1.25v. Quite a nice drop from .95/1.4 default.