Processor Archetecture and Processors?

What is the advantage of Sandy Bridge over Nelham?

Which is better: Amd Phenom II vs. Xenon vs. Intel i7 vs. Intel i5?

Also when will we see any Phenom or Xenon lappy's

 

uh...

sandy bridge isn't out yet and intel is tight lipped about it until q4 2010
nehelam is a desktop platform
xenon is a server-type cpu and not in laptops at all
new phenoms are hexcore and not in lappys until 2011


...???

 

oopps...thats why I asked...some post are confusing...please elaborate

 

the xeons use 1366 socket, which is similar to core i7s in clevo's D900F. Eurocomm and avadirect offers xeon CPUs in D900F.

 

oh....is the xeons power hungry? Idk 6 core lappys sound so nice

 

Some of the xeons have tdps lower than the 980x. Ofcourse tdp doesn't equate power consumption.

 

Xeons are the same silicon as any other desktop or laptop nehalem, which are all the core i processors. You need to look behind Intel and AMD's naming conventions to get at what the difference (or lack thereof) is among their various offerings. Wikipedia is a good start for researching such information.

 

Xeon is a desktop CPU. The 3xxx Xeons are exactly like i7s except that they use ECC memory. The 5xxx Xeons are different in that they have 2 QPI links, allowing you to actually put 2 Xeon 5500s in parallel on a same motherboard IIRC(assuming everything supports it of course).

Phenoms are already present in laptops, just not the 6 core kind. They stack and compete with i3s and i5s.

As for "which is better", you really should specify if you're talking about laptops or desktops as well as what uses you're comparing them in.

 

Nehalem is the microarchitecture, and i3/i5/i7s in both laptops and desktops, as well as some Xeons, Celerons and Pentiums, use the Nehalem architecture.

 

Yeah thanks I'm really not 100% sure about the whole platform/architecture/whatever thingies Intel and AMD use >.>

Well quite a few Xeons are for the most part the same as i7s so if i7s are Nehalem microarchitecture than quite a few Xeons will be as well.

 

See Thread for Sandy Bridge info.

 

If that's all Sandy bridge goingg to offer, the westmere version of nehalem seems to be as good as it will get for a while in the enthusiast market. I wonder when the 8-core w/ hyperthreading's coming. According to wikipedia SB should offer quadcore as mainstream.

 

Quad core goes mainstream in 2013 when Sandy Bridge gets shrunk to 22nm and is named Ivy Bridge. And Sandy Bridge will have big power savings over current Arrandale's. Also you didn't read the whole thread, there are a lot of improvements.

 

for the most part, the end user neither does not car nor do they 'see' any difference....

 

But they might care about Widi being able to stream 1080p, faster graphics, more laptops with USB 3.0, and 3D support...

 

I was talking about enthusiast market. I can see the power savings and all that, but that mostly matters to the mainstream.

 

I don't really like the term "enthusiast market". It's quite poorly defined, on the whole.

I have an MSI GX640 which I'd call an enthusiast laptop, and yet I definitely care about battery life. Mine gets 3 hours, but if a few years down the track I see a laptop that has a high-end GPU with switchable graphics, and a CPU that is powerful and yet has rather low power consumption, I'll jump right on that.

 

really? :laugh:

hmm i guess so

lackofchesse got it. also Forever_Melody say the cpu's would be use for data manupilation, science/engineering computation, 3D modelling/analysis, gaming, etc. Stuff like that.

Ahh Sandybridge looks fine...um whats replacing USB 3.0 again? I like how small it is, it has its own integrated graphics core..which is nice.

Really?

I guess....

yeah true dat...wait the integrated graphics gives u 3D support?

yup a do it all laptop

 

Xeons are targeted for workstations and servers, not necessarily desktops. Also ability to use ECC and/or registered memory is not universal. There is really a plethora of xeons available and few generalizations can really be made about them as a whole as they run the whole spectrum that Intel's consumer processors do.

 

Lightpeak is.

 

ahh it has the potential to transfer 100Gb/s of information, but at the start it will be ~10GB/s. I believe it uses an light instead of electromagnetic and currents. So I'm pretty sure there going to have some sort of inducer and analyzer from the port to the device?

 

Just note that Gb is not the same as GB, since the little "b" generally means bit as opposed to "B" for byte = 8 bits.

The figure is 10Gbit/s demonstrated bandwidth, which is around twice as much as USB 3.0.

 

:sigh: The basic principle behind them is the same. One end switches on and off really fast and the other end reads it, so both would have an "inducer" and "analyzer."

 

hmm? i felt like sandy bridge was a big jump from westmere, considering it's a new architecture + using socket 1155, which will require a new motherboard
"The top-of-the-line dual-core mobile Sandy Bridge is expected to be 20% faster than the top-of-the-line Arrandale. [5]" (wikipedia)

and from what i gather from wikipedia, the high performance SB processors will all be quad core

every year there's either a change in microarchitecture or the processor shrinks, from what i gather, a change in microarchitecture is more about performance, while a shrink is more about lowering power consumption and producing the chips at a lower price

so.. the next "big" change is SB, and the one after that is expected to happen in 2012/13

 

But I'm talking about enthusiast part, i.e. the i7-980x. There's 6-core >3GHz now. So a quad should be nothing. IIRC wikipedia mentioned appearance of 8-core cpus. But if the 8-core isn't coming then like I said, the 6-core westmere is going to be top of the line for a while.

The quad could be 20% faster for, what I assume, clock for clock, but if core #s would remain at 6 at most, then I doubt anything new will be way faster than the i7-980x. But we'll see.

AMD's supposed to release Llano or what ever they call it, their version of HT CPUs. Maybe that'll push Intel and make them come out with something better than what they've shown so far.

 

Llano will not have hyper-threading. Llano is basically a current Phenom cpu combined with an ATI 5000 series graphics unit, plus some optimizations. Bulldozer, which comes later in 2011 for notebooks, will have hyper-threading.

 

None of which have anything to do with the current or expected generation of processors.......

 

One of the noble gasses man, tis Xeon, and yes it is in one or two DESKTOP motherboard/calibre boxes that can barely be called notebooks, and are insanely ridiculous to begin with, i think the D900C can take Xeons.

 

Never mind, delete.

 

Oh dang. It's going to be even worse for enthusiast then.

 

Don't obsess over hyperthreading. Intel themselves say that HT on two cores is not nearly as efficient as conventional threading on four cores.

There is little enough consumer desktop code out there that runs properly multi-threaded, much less worrying about whether or not it can run in four or eight threads.

Stuff that is intended to run on servers like databases and apps servers, that is another matter. But consumer desktop code is way behind in taking clear advantage of the multi-thread capabilities already available.

 

oh wow, yeah for supercomputers and stuff. Hey is it better to have 4 logical cores (2 physical and 2 threads) or 4 logical cores (4 physical, 0 Threads)?

 

4 physical cores is obviously better than 2 Hyper-Threaded physical cores when it comes to performance. However, the hardware required for HT on a core is much less than for a whole extra core, so an extra HT core is less expensive and consumes less power than a whole physical core. I'm having trouble finding numbers for the extra transistors and power consumption that come with HT, though.

 

it's cool take time, because aparently a Quad core i7 has more processing power vs. a Core 2 Quad of same clockspeed...how ? Aren't i7 just faster?

 

Well, for a start, there's HyperThreading in the Core i7s while C2Q's don't have it. That's the biggest difference. Besides that, there are architectural differences that act in favour of the i7, such as the on-die memory controller.

 

oh yeah...forgot that.

 

Next time if you see spam, don't quote it, report it please thank you.

To those who did report it, much appreciated,
Johnny T
NBR Moderation Team

 

oh okay..thanks

 

i3 and i5 have hyperthreading too.......

 

Take a P4 for example. When there were only single core P4's out there, performance was decent for the time. But the NetBurst architecture had major efficiency issues, per-clock. When a program would put information through the processor some operations would have to wait until others were done, in order to be processed properly, and this let to a HUGE amount of wasted clock cycles.... THIS is where Hyper-threading came along. Intel provided a crutch to make up for the in-efficiency in the NetBurst design, allowing all the clock-cycles to be utilized, thereby appearing double threaded. Hyper-threading isn't TRULY multi-threading. It's just using the left over clock cycles that the current core isn't using, and that's also why performance doesn't scale linearly with the amount of threads present. Personally, to me it's just cleaver marketing on Intel's account. In a sense it does make things more efficient, but only to the programs that are specifically written to take advantage of the way it works.

 

hmm nice.

yes very clever..SO how does the AMD phenon X4 and X6 stack up to intel? They don't use HT..so do they suffer problems of unused processing cycles?

 

Hyperthreading has shown up to 30% improved performance in some extreme circumstances. Usually it is around a 10-15% improvement but sometimes it has no effect or even decreases performance.

The fact that hyperthreading doubles the number of cores the system sees is just a side-effect of it and really has no bearing on anything.

 

oh so it doesn't really boost anything that will blow things away its more like it might but not noticeable in real world apps?

 

Basically the processors with HT are optimized processors. Like proven the increase in performance is not much because you can't split in 2 a real core, you are just using whatever extra clocks are in there waiting (and a processor is more than clocks, you have cache memory and so on which are not split in 2). In any case an increase in performance is always good.

 

It's not exactly spare clocks that are the issue - with modern superscalar architecture, the CPU is actually capable of executing more than one instruction at once, but in some situations two threads can use this capability better than just one.

 

Thanks for the link and clarifying my lack of precision... this HT technology seems more complex than simple adds.

EDIT: I finished reading the link. It was quite enlightening. Thanks again!

 

Yea, lol. I was just keeping it simple, and I didn't mention that HT has been basically over-hauled for the 45nm and smaller CPU's. HT almost made the P4's worse.

 

interesting indeed, but really multi-core processors are a smarter, cheaper way of emulating a super scalar processor. Interesting indeed. So really HT does give a performance boot due to it running serial instructions but two instances of it. I believe intel killed AMD right now

 

That's a well known fact if you consider performance and only performance.

 

And this is only in one small market. If Intel truly was much better and AMD went out of business, you'd see a huge decline from Intel since it would have no competition to push it.