Intel's Core 2011 Mobile Roadmap Revealed: Sandy Bridge Part II - AnandTech :: Your Source for Hardware Analysis and News
Memory bandwidth could be quite imortant again. The 2620M could perform nicely with 2 GPU cores and DDR3 1600 support.
-
Meaker@Sager Company Representative
-
+rep.
Nice info. I find it strange that Huron River supports 1600 MHz and desktop Sandy Bridge doesn't, all previous information pointed towards 1333 MHz. But things might change still.
I also don't fully understand why are two graphic cores needed. Maybe for power saving purposes.
Also I wouldn't trust that road map past Q1 2011. I am sure that in Q3 2011 there will be a Sandy Bridge refresh, the Q3 mobile refresh is now an almost Intel tradition. -
Meaker@Sager Company Representative
Two graphics cores make sense. One is giving desktops around the performance of a HD5450 (above in some, below in others).
Each core work together and are made up of 8 units. So two gives 16 units.
Now depending how this goes it could be 0-100% faster (depending how much is duplicated and how much memory bandwidth limited).
But we could see ALL sandybridge laptops able to play ALL modern games at 1366x768 at low-med details.
With 2 cores I could see memory bandwidth being an issue so DDR3 1600 could really help.
Not only that but they turbo independtly. The CPU and GPU will turbo to balance to give the best performance for each game.
That sounds like a great way of getting performance from laptops. -
Come to think of it, Intel and AMD (which has the same problem with Llano and its successors) could beat the bandwidth issue by simply using more memory channels. Even assuming they stay with DDR3-1600 (and they probably won't -- DDR3 is getting faster) all they need to do is use 3 channels to get 38.4 MB/s. With 4 channels, we're well into high end territory with 51.2 MB/s. Three channels will also help offset the fact that the GPU is gobbling up system RAM. -
Meaker@Sager Company Representative
A few Problems with the above.
1. More memory channels = more pins on the CPU = larger sockets = less space in the machine and more legs (difficult to make, intel use pads on all the high pin count chips, but legs on laptop chips to make them more secure).
2. More memory channels = more dimms, 4 channels = at least 4 slots, all of which need to be populated. More space and more expensive. Most laptops can stack 2, 3 would need a thicker machine.
3. More memory dimms = more power consumed.
4. More memory channels/dimms = more complicated PCB routing, could easily mean that more layers needed to be added racking up the cost.
Considering all the above and the broad target market for the platform it would not make sense from Intel's perspective.
Now if graphics performance is memory bandwidth limited and good, maybe manufacturers could play around with memory ratios and get DDR3 2000 in there?
Of course with the new BCLK changes and locking etc, that might not be possible. -
tilleroftheearth Wisdom listens quietly...
Just a note that the AnandTech preview processor is now thought to have 12 EU, not 6 as first believed. -
Question here for all you notebook geeks
I was planning to purchase the Asus G53 with I7 740m and the GTX 460m. But bad news came and 3D, Widi and Sibeam was all excluded from the G53...
So the only update i get from buying that laptop is 300 MHz faster CPU, GTX 460m istead of GT 240m and 8GB instead of 4GB. Which i think is not very much.
So im thinking of waiting until a new Asus laptop with Sandy Bridge comes along.
1. What are the main benefits of Sandy Bridge compared to Nehalem?
2. Will all new laptops with Sandy Bridge be available with DDR 1600? Or will it be available? Any noticable differences with 1600 compared to DDR 1333?
3. Will there be any new SSDs from Intel around when the first laptops with SB is released?
4. When will the first laptops with SB be available if SB is released around CES 2011?
Thank you -
tilleroftheearth Wisdom listens quietly...
1). See first post for a preview of what is to come.
2). I think it will be DDR3 1333, but you'll notice no difference on a notebook (unless you want to overclock).
3). Yes, be expecting the G3's to be available in full force.
4). If I'm remembering correctly, the first SB notebooks will begin to show up in Q4 2010 - so, the wait is not as long as it might otherwise seem.
Hope this helped. -
Quick second question here: G3 compared to G2, whats the main difference? Any noticable speed boost? -
tilleroftheearth Wisdom listens quietly...
The higher speed RAM will simply have higher latencies to go along with it so the actual benefit to the notebook user will be minimal, if any.
G3 vs. G2 - that is the question of the hour! I think instead of an increase in theoretical numbers, we will see a nice speed up in actual realworld use - not to mention the huge (double to triple) capacity gains of up to 600GB.
Note though that with all SSD capacity increases, their is a noticeable speed boost too (with most SSD's) - but we'll only know for sure when Anand and/or others get them and put them through their paces. -
That i7-2720qm looks like it will rock
I can't wait to see how it performs.
-
Meaker@Sager Company Representative
However yes for a machine using dedicated graphics there would be 0 difference between 1333 and 1600 ram.
Of course you can get low latency 1600 kits from kingston. -
That is amazing, and the upgrades to the current i5's and i7's dual cores seem impressive. (the ram will probably be the most notable speed increase, seeing as how most users will never need that kind of processing performance). However, even moderate gaming should be supported by the discrete GPU's that those units are being shipped with...
However, i am very interested in what kind of battery life an 8 / 9 cell laptop would get using that Core i7 2920XM... or even the QM's... -
Well i have to get DDR 1600 anyways. Cant get new technology: new CPU sandy bridge, new awesome GPU, new G3 SSD and not new RAM?! That is just wrong.
-
tilleroftheearth Wisdom listens quietly...
Cloudfire, you really don't believe (me) that the RAM will have little actual effect, huh?
It would be 'just wrong' if you didn't have to pay a premium for the nominally higher spec'd modules. -
And i do have the money for it, so... -
the only thing ill care about with ssd's is if they get significantly cheaper.
-
Just like with the current i7 quad cores, there is a difference in L3 cache between the Core i7 2720QM and the Core i7 2820QM.
Is the difference in L3 cache just plain marketing from intel to make us pay another 400-500€ more for an unnoticable increase in real life performance, or will it really matter in the future when your laptop is at the end of its life cycle and you are squeezing every bit of power out of your cpu?
And if it really does make it a difference, for what applications (for example: will it only benefit people who use heavy programs like architectural designing, or will it also affect gamers) ? -
Also, you're extremely unlikely to see significant difference between two high-end processors with 8MB and 6MB cache -- the problems usually arise with CPUs that are cache-starved even when they're released (e.g. AMD's Athlon IIs and mobile Phenom IIs or Intel's Celerons). -
Meaker@Sager Company Representative
Well phenoms just love cache lol.
-
Here is a couple of good articles you could read. Judge for yourself if it's worth the money.
Memory Scaling on Core i7 - Is DDR3-1066 Really the Best Choice? - AnandTech :: Your Source for Hardware Analysis and News
Core i7 Memory Scaling: From DDR3-800 to DDR3-1600 : Which RAM Speed Is Best For Core i7?
I know the articles are for desktops, but the information is also relevant to Laptop's as well. -
-
tilleroftheearth Wisdom listens quietly...
Although GPU tasks do benefit (in benchmarks), as long as the min frame rates are above a certain threshold (lets say 30) the real world effect is essentially non-existant.
Given that it costs over 2x the money (the money spent on 'better' RAM) to get the 'benefits' (2-20%, depending on which app is used) the money is usually better spent on a faster cpu and/or more RAM. -
-
Meaker@Sager Company Representative
The effect does reduce as you approach the screens refresh rate (60hz or 60 fps in most cases) but in particularly in stressful circumstances (lots of explosions) it would help reduce slow down even if you are beyond 60fps.
We are talking this specific case using the integrated graphics of course. The rest of the time it makes not very much difference.
Some 1600 memory:
Newegg.com - Kingston HyperX 4GB (2 x 2GB) 204-Pin DDR3 SO-DIMM DDR3 1600 (PC3 12800) Laptop Memory Model KHX1600C9S3K2/4GX
Is $106 so not that bad.
For a bit more and if you need the OCing head-room there are also these:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820104192&Tpk=SODIMM -
Except most people who want to game don't run integrated graphics, so for the majority, memory speed won't affect their performance.
-
Meaker@Sager Company Representative
-
Thanks for the info!! I like the boost of the base frequency of new i7s. No more 1.6/1.7Ghz i7s!! Looks like the i7 2720QM will the successor of the current i7-720/740m, it should bring the current extreme i7-920QM to its knees!! Q1 2011 will be a great time to buy a new laptop!!
-
-
I have a question about the sandy bridge architecture
I read that it won't be possible to overclock the cpu so much except for the K-series as everything will be linked
so my question is :
because of this link, overclocking "everything" just 2-3% via Base Clock adjustments could it give an improvement as good as overclocking several components but one after another separatly like we do in the current architecture
like for instance overclocking the gpu, the gpu being the bottleneck then overclocking the memory then the cpu or overclocking only the gpu or any other combination for a total of 6-9%? (if taking into account only three components and if we suppose that every component in the sandy bridge architecture fit correctly with each other at start)
it's like 6-9% increase in this example like parallel vs serie
Is this correct or not? -
I believe that's incorrect. I don't see how a 2-3% cpu/gpu overclock would be more effective than traditional overclocking of the cpu multi, fsb, mem etc. Traditional overclocking will be limited to the K series. Overclocking on anything else isn't going to be the same as what we currently know overclocking to be.
-
Doesn't Turboboost sort make overclocking irrelevant for all but the hardcore? Especially on a laptop.
-
Meaker@Sager Company Representative
Not exactly. You have frequency range for all chips, going up as you pay more.
The lower priced chips will probably still reach those higher frequencies but are locked by intel. -
At least for my desktops, I know I'll probably be buying AMD from now on, unless these Intel chips massively outperform them for the same price, to the point where overclocking the AMD chip can't compensate.
At least, if this move by Intel does anything to increase AMD's market share (and I'm fairly certain it will), then they probably won't do the same thing in the future. -
OMG
This is gonna be one looooooong period of 4 months waiting for a laptop with Sandy Bridge. I wanted to buy the G53 from Asus and was all psyched about it, until i realized it is waaay more value waiting for SB.
Please someone ship me some strong super secret drug i can take that put me out for 4 months -
I sure hope desktop sandy bridge will have decent overclocking capabilities (even if you eventually have to rely on the multiplier) or it would be a big let down. It's already a shame that it seems not all of them will feature HT.
-
I wouldn't expect so much from overclocking unless you plan on getting an unlocked multiplier CPU.
-
Is it confirmed that hybrid graphics and discrete GPU's won't mix on the mobile chipsets?
-
sandy bridge K-serie overclocked to 4.9Ghz
Intel overclocks Sandy to 4.9GHz - Processors - Processors - Fudzilla -
How awesome is this:
Intel's Sandy Bridge processors have a remote kill switch
More sandy bridge mobile info (clocks, gpu unit clocks and amounts)
Discussion in 'Hardware Components and Aftermarket Upgrades' started by Meaker@Sager, Aug 30, 2010.