|
When using this resource, please read FragKrag's opening post. The Tech Support forum regulars have helped create countless of desktop systems without any compensation. The least you can do is provide all of the information required for them to help you properly. |
On February 15 2012 04:55 Medrea wrote: I dont think anyone expected Ivy bridge to be cheaper >.>
I just don't understand why everyone says wait for Ivy Bridge is all. It will be a very minor, if any better bang for buck than a 2500k, 2600k or whatever someone is after.
|
It wont make Sandy bridge cheaper thats for sure.
The big deal about Ivy bridge is not its stock performance, but the other perks that a die shrink brings to the table. Like reduced power consumption, and heat.
|
Yeah I wouldn't think it's particularly worth waiting for unless your current computer is already pretty decent or you're getting a laptop. Some of those Ivy Bridge ULV parts at 17W TDP should beat an FX-8150 in Cinebench single-threaded, and probably in many games...though of course GPU options are limited on a laptop.
|
On February 15 2012 04:59 Medrea wrote: It wont make Sandy bridge cheaper thats for sure.
The big deal about Ivy bridge is not its stock performance, but the other perks that a die shrink brings to the table. Like reduced power consumption, and heat.
I agree it will have higher clock speeds, but I really doubt they'd do something like they did with the 2500k where they gave it a stock speed of 3.3/3.8Ghz and allowing to overclock to 5.5-6Ghz (A possibility with a die shrink). They will just raise the base clock speed, and it will still be priced at the same performance/price ratio as SB.
I've never had problems with a desktop on my desk or below my desk because of heat. I've never seen a processor fail because it runs warm (not hot). All that will happen is saving maybe $5 a year on your electricity bill, and run a few nanodecibels quiter? Which it probably wont since the GPU is the main source of noise.
|
On February 15 2012 05:11 Myrmidon wrote: Yeah I wouldn't think it's particularly worth waiting for unless your current computer is already pretty decent or you're getting a laptop. Some of those Ivy Bridge ULV parts at 17W TDP should beat an FX-8150 in Cinebench single-threaded, and probably in many games...though of course GPU options are limited on a laptop.
That's the one thing I am very excited about... to see the battery life of an i3 laptop. I have one right now, it does all my day to day tasks very well, as well as the improved intergrated gpu which let it play lots of games on lower settings.
|
Oh right.
The performance increase of the IGP is supposed to be huge.
|
Competition dictates pricing. AMD prices the 7970 as high as it is because it outperforms a GTX 580. Supply is limited and demand for such a part is there so pricing it lower would just be stupid.
2500k is 3.3 / 3.7 with a max multiplier of 57. Ivybridge is increasing it to 63.
|
On February 15 2012 05:18 skyR wrote: Competition dictates pricing. AMD prices the 7970 as high as it is because it outperforms a GTX 580. Supply is limited and demand for such a part is there so pricing it lower would just be stupid.
2500k is 3.3 / 3.7 with a max multiplier of 57. Ivybridge is increasing it to 63.
The pricing makes sense for the companies, all I'm saying is the lack of competition is really hurting the consumer because Nvidia just doesn't have an answer right now, and there has been very little released about the 600 or 700 series GPUs, so in the meantime the profit margin for them is very high.
I'm curious, does a die shrink mean clock for clock the performance will be exactly the same? Because if that's the case I'm sure they'd probably just create a die shrinked 2500k, clock it 500Mhz or so higher, and stick on an extra price of $30-40.
|
Die-shrinks come with slightly higher instructions per clock.
|
Erm. I dont think Ivy bridge is getting a change to its execution model with maybe a change to its AES but SSE4 I think and all that is untouched. Processors can only run one instruction at a time per toggle of the clock by definition anyway (more in cases like hyperthreading) and instructions vary in length (words, complex or simple etc etc). Unless you are talking about instructions per time due to less stalling. In which case Intel is hoping for 20 percent here but I doubt they will get it.
The transistor count is being raised from 1 billion to 1.4 billion.
And Ivy bridge is also introducing Intels new revolutionary Tri-gate technology, which lowers power consumption in both on and off states. It also improves response time of the gate so propagation delay when switching states is lowered. Which means higher clocks at lower voltages down the road. Remember that discussion about whose changing the market more, AMD with multi-core or intel with better fabrication? This is what i was referring to.
And yeah laptop longevity looks extremely promising with the new technology. We might finally have laptops that we can qualify as true mobile devices. Instead of now where we just have devices that are mobile enough to make it from power source to power source.
|
When Intel just does a die shrink, usually they make a few other tweaks as well that improve performance a little. It's not the die shrink itself improving performance but the tweaks and better integration here and there. Once you know the bottlenecks of an architecture a little better or have a little more time for fine tuning, improvements can be made while maintaining pretty much the same architecture. See here: http://www.anandtech.com/show/4830/intels-ivy-bridge-architecture-exposed/2
edit: on a quick read, the change list is pretty small in terms of architecture tweaks. I don't see where 20% more performance is coming from, like some have predicted. Or even 10% or whatever more sane things people are saying. It'll depend on the workload, but we'll see.
|
|
doesnt the die shrink make the manufacturing cheaper too? It should be priced below Sandy Bridge in a few months probably...
|
No, die shrinks dont do that. Raw material differences between die shrunk architectures are zero or virtually zero. The fabrication process involved to ram two elements closer and closer together without running into quantum tunneling problems or simply fucking up just makes the whole process even more expensive.
Die shrinks are more expensive, not less expensive.
Im not sure where people are getting this idea that raw materials in a computer chip go down as you die shrink. Can someone link a source? Ill settle for a source that says die shrinking reduces fabrication cost in general.
Honestly its not even about the raw materials, we fucking use silicon with just about everything. Its mainly the R&D and fabrication process.
|
On February 15 2012 08:33 Medrea wrote: No, die shrinks dont do that. Raw material differences between die shrunk architectures are zero or virtually zero. The fabrication process involved to ram two elements closer and closer together without running into quantum tunneling problems or simply fucking up just makes the whole process even more expensive.
Die shrinks are more expensive, not less expensive.
Im not sure where people are getting this idea that raw materials in a computer chip go down as you die shrink. Can someone link a source?
I'm not an expert on this subject but when you read various tech news, the manufacturers say that 'because of the die shrink, the manufacturing process will be cheaper' and so on. For GPU, CPU, SSD and basically everything. In short term, this is not always true due to the factories having to adapt to the shrink, various issues from the process and so on, but because the chisel is smaller it will eventually be cheaper overall to manufacture. But maybe someone more knowledgeable could clarify?
|
For just a die shrink, the size of the die is...shrunk, so smaller. Assume that the production cost of fabricating on a wafer is the same for the new process and the yields as a percentage are also the same (both are not true, but with time it almost becomes that way), and that the wafers are the same size. So we're assuming the cost per wafer is the same to the manufacturer.
Smaller die size means you get more dies per wafer so less production costs per working chip you get.
The refined silicon wafers they buy aren't that cheap (super-duper purity of anything is never that cheap), and neither is the fabrication process and time it takes to handle a wafer. Over time the savings per chip should compensate for the cost of the new fabrication plant or retooling, and so on. Actually, that's not really the case as the R&D and plant costs are huge. Bleeding edge die shrinks these days are more for competitive advantages in power consumption and heat. That's why AMD wants to shift focus towards selling stuff that doesn't emphasize the manufacturing process--provide types of solutions others don't, rather than produce products that straight-up compete with Intel.
|
Noob question alert, whats a die shrink?
|
Im going to need clarification since Intel's chip area has never dropped below 75mm squared. Since intel is cutting out the same square area of silicon wafer for every single chip, how does a smaller die area reduce material usage?
On February 15 2012 08:50 Josh_rakoons wrote: Noob question alert, whats a die shrink?
When we talk about die shrink we are talking about how close various elements on the chip are allowed to be. Smaller spaces allow current to travel through the processor faster, as well as a bunch of other benefits.
Overall die size rarely goes down, only up.
|
Various CPU die sizes: (edit - updated link with Sandy Bridge E for more comparisons) http://www.anandtech.com/show/5276/intel-core-i7-3820-review-285-quadcore-sandy-bridge-e
I forget if there was a die size for Ivy Bridge revealed, but AT had an estimate earlier:
While we don't have a close-up die shot of Ivy Bridge (yet), we do know its approximate die size and the layout should be similar to the Sandy Bridge die as well. Anand estimated the die size to be around 162mm^2 for what appears to be the quad-core die (dual-core SNB with GT2 is 149mm^2, and even with the more complex IGP we wouldn't expect dual-core IVB to be larger). That's a 25% reduction in the die size when compared with quad-core SNB die (216mm^2). A 22nm quad-core SNB die would measure in at 102mm^2 with perfect scaling and assuming all the logic/architecture is the same; however, scaling is never perfect and we know there are a few new additions to IVB, so 162mm^2 for IVB die sounds right. Transistor wise, IVB counts in at around 1.4 billion, a 20.7% increase over quad-core SNB.
http://www.anandtech.com/show/5174/why-ivy-bridge-is-still-quad-core
|
Alright Ill grant that then.
But the technological advancements needed to actually squeeze elements always add a higher cost than the actual material you are cutting the chip out of.
Either way it seems like total die area isnt a concern. Once we get more space freed up it seems like we (as in Intel only at this point I guess) just seem to shove more shit directly inside the CPU.
I really dont like AMD's new focus. We need someone to give Intel a run for its money otherwise we have absolutely no pricing competition at all.
EDIT: Now that I think about it, are we sure that a smaller used silicon die area translates to a smaller overall wafer size? I can't imagine Intel cutting these things right to the edge if you know what I am saying.
|
|
|
|