32nm Arrandale (Laptop) in 1Q10
- Thread starter 21stHermit
- Start date
Impressive, although unfortunately most OEMs will probably jack up the price to upgrade from 2.4 to 2.5 to more than $40.
I hope the news about new wireless is true too. These new cards should be final N spec. I am assuming both the 6200 and 6300 will support 5ghz.
I'm starting to see what everyone means now though about the grain of salt.
I hope the news about new wireless is true too. These new cards should be final N spec. I am assuming both the 6200 and 6300 will support 5ghz.
I'm starting to see what everyone means now though about the grain of salt.
Unlike fud/theinq/BSN, digitimes can actually be fairly well relied upon when it comes to aligning your expectations with reality when they say things like this:
The sources in Taiwan are usually pretty good, the problem is that the sources inside the notebook makers can only communicate what they are being told by their management and by Intel.
The trick of course is actually getting Intel to commit to a release date prior to the release date.
If you don't have that in writing then you can't legitimately say Intel (or AMD for that matter) was actually late, all you can claim is that the notebook makers in Taiwan were wrong and Intel shipped later (or earlier) than they were projecting when talking to digitimes staff.
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,787
- 136
Yea I agree. So far as we know, the release date for Clarkdale/Arrandale hasn't changed at all. If the product was supposed to release at Q1 2010, but got forwarded to Q4 2009 and got
"delayed" to Q1 2010, is it really delayed or not?
I mean, delays are what we call for NV30, Tukwila/Dothan/Prescott.
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
Hmm. Well pricing info is interesting. Still find it hard believe we know basically everything but performance numbers and a release date and they are saying Q1 and its JULY. Maybe its a fab issue (IE they dont have the 32nm production capacity to supply the market after launch).
Im curious where the 2 cores -> 1 core switchover is for turbo mode. Do they only 1 extra multi for dropping to a single core, 2? 3?. We know base speeds and turbo speeds, just not the switchover point.
Im curious where the 2 cores -> 1 core switchover is for turbo mode. Do they only 1 extra multi for dropping to a single core, 2? 3?. We know base speeds and turbo speeds, just not the switchover point.
What I had said in the other thread is that they brought the date forward, axed havendale, then delayed it back, which I immediately (over)reacted to with suspicion. If there was no havendale it would have been different. I admit I don't know the entire timeline, so I will reserve judgement.
and lol@nv30. That was the second longest running joke in the gaming industry, right after Duke Nukem 4ever. The GeForce Vacuum cleaner(we laughed about it back then, but now almost all high end cards are as big as the 5800. Of course, unlike the 5800, they aren't outperformed by the single slot versions
).Isn't it about threads, not speed?
Seems to me if you are running 3 or more threads it has to be in dual core mode and can't use turbo. If you have 2 or 1 threads it would automatically process in turbo.
Having said that, if Arrandale turbo is like Clarkdale, then it's a joke, only one multiplier step: i.e. if it is a 12X Multiplier then turbo give you 13X. Certainly not worth the added expense for me.
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
The info from goto-san shows significant gains from turbo mode for arrandale (and clarkfield).
http://img190.imageshack.us/img190/4264/capturehuv.png
And turbo and HT aren't mutually exclusive. AFAIK turbo relates to physical cores in use, not threads in use.
edit: I was bored and expanded the chart http://img213.imageshack.us/img213/310/capture2fdj.png
http://img190.imageshack.us/img190/4264/capturehuv.png
And turbo and HT aren't mutually exclusive. AFAIK turbo relates to physical cores in use, not threads in use.
edit: I was bored and expanded the chart http://img213.imageshack.us/img213/310/capture2fdj.png
Lets say you have a 2ghz dual core with a 3ghz turbo.
When you are using multihtreaded apps, you have, essentially, 4ghz of processing power. When you have a single threaded app, you get 3ghz of processing power. Multicore is more efficient than single cores with high clockspeeds, but Turbe mode is there for situations where you are unable to take advantage of that.
Heres a question: if Intel released a 3ghz single core CPU with HT for 65% of the cost of a 2ghz dual core with the same TDP, would you buy it?
:thumbsup: :beer: Now that's value-add posting! Thanks for taking the time to compile it and doubly so for then sharing it with the rest of us
:thumbsup:
Consider putting more column headers, even if you have to abbreviate and use a legend. You included the power for some of the laptop IGPs, that power is part of the TDP correct?
:thumbsup:
Agreed.
If you have a dual core CPU (Arrandale or Clarkdale) with HT, then you can have max 2 threads running on one core before you must bring on a second core. Whenever the second core is on then turbo is off, hence my thread logic for turbo.
If I'm wrong, please someone explain.
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
your logic is fine, we just don't know if its implimented that way, it'll depend on the scheduler and probably what power mode, (HT being more efficient than waking the second core most likely).
I'll make some adjustments to the chart soon. It takes only a few seconds to post a new version but I literally posted that and then formatted my comp.
more headings: http://img34.imageshack.us/img34/1854/capturebvo.png
I'll make some adjustments to the chart soon. It takes only a few seconds to post a new version but I literally posted that and then formatted my comp.
more headings: http://img34.imageshack.us/img34/1854/capturebvo.png
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,787
- 136
I'm thinking for the SV versions the Turbo will be +2 for dual core. So a 2.66Ghz version will reach up to 2.93GHz in dual core operation. +3 would be too high as the max is +5, and its how Lynnfield's Turbo works with +5 max Turbo.
LOL. The single core would be at worst equal to the 2GHz dual core if the dual core doesn't have HT. Remember, multi cores have problems with cache coherency and conflicts, along with other things etc which prevent perfect scaling.
Yes, I would take the 3GHz any day.
You could be right of course but just saying that since HT doesn't add much thermals and power as activating more cores it might be purely core based.
So whats the technical reason why Intel doesn't produce something like this? It should be half the cost since you eliminate half the die, and smaller dies have higher yields so that should start to even out the cost of the cache and the overall package.
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
Nah. Went from RC to RTM of win7 x64.
Drizek: One word. Atom. The 1->2 core jump might be BE a huge leap, but it FEELS like a huge leap. I dont want to touch a single core ever again. Its like nails on a chalkboard every time I try.
Just noticed that Gulftown (the one thats cutoff) should have a NA under GPU power.
fixed, and added process tech. http://img269.imageshack.us/img269/7091/captureikq.png
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,787
- 136
We probably won't see single core in anything above Atom anymore. The comparison would have been more realistic with dual core 3GHz + HT vs quad core 2GHz sans HT.
In both cases, it comes back to the same question. Yes, a dual core is faster than a single and a quad is faster than a dual, but that doesn't change the fact that you are almost doubling the die space. This was always sort of "hidden" because when Intel went from single to dual pentium m, they did a die shrink and released new cpus at the same price, same TDP and same clockspeed with twice the cores. I'm just wondering if maybe more people would prefer to just get a CPU thats as good as the previous gen for half the price, especially if that also means you can get 40-50% higher clock speeds than the multicore version with the same thermals.
Theirs a gem in the chart. The new 45nm IGP has a power range of 8-10W, that's half of the 945 IGP and it's independent of CPU clock. If one attempts to OC Clarkdale then the power in the IGP will scale with that OC. Likely limiting the OC. Time will tell.
The new 45nm IGP will also be used in Atom Pine Trail. So that 8-10W will still be more than Atom's 4-8W in SC and DC, but a significant improvement over the 945 IGP.
ilkhan
Golden Member
- Jul 21, 2006
- 1,117
- 1
- 0
I added some shading to keep the columns separate. Hopefully this is the last version for a while.
http://img32.imageshack.us/img32/2179/capturexzq.png
http://img32.imageshack.us/img32/2179/capturexzq.png
IntelUser2000
Elite Member
- Oct 14, 2003
- 8,686
- 3,787
- 136
945GM is 7.0W btw.
And no, they aren't using the IGP in Clarkdale to Pine Trail. Pine Trail is likely going to use a faster GMA 500 or GMA 950.
And you forget that in Netbooks Atom is 2.5W and the 945 IGP is only 5W.
TDP for the socket? Yes.
TDP for the entire laptop unit? No.
edit: Very timely, Goto-san's latest chart on TDP CPU roadmap
tokie
Golden Member
- Jun 1, 2006
- 1,491
- 0
- 0
Why would the entire laptop unit NOT have a higher TDP? would it not even be only 1-2 watts higher because of what Intel has done?
The GPU would obviously not be able to be physically locked off since it contains the memory controller. So there must be at least a certain amount of spillover current into the GPU from memory controller, even if Intel GPU was not in use. That whole 45nm die would not be able to be shut off b/c of memory controller, and Intel couldn't laser-shut the GPU section off for each manufacturer who wished to use solely discrete GPU.
I think Intel really has Nvidia/AMD discrete chips by the balls here with what it has done...
The GPU would obviously not be able to be physically locked off since it contains the memory controller. So there must be at least a certain amount of spillover current into the GPU from memory controller, even if Intel GPU was not in use. That whole 45nm die would not be able to be shut off b/c of memory controller, and Intel couldn't laser-shut the GPU section off for each manufacturer who wished to use solely discrete GPU.
I think Intel really has Nvidia/AMD discrete chips by the balls here with what it has done...
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter