The Intel Ivy Bridge (Core i7 3770K) Review: AnandTech | Hacker News Reader

The Intel Ivy Bridge (Core i7 3770K) Review: AnandTech

anandtech.com

50 pointsamartya91626 comments

"Classic feature segmentation is alive and well with Ivy Bridge. In the quad-core lineup, only Core i7s get Hyper Threading - Core i5s do not. When the dual-core Core i3s show up in the coming months they will once again do so without support for turbo boost. Features like VT-d and Intel TXT are once again reserved for regular, non-K-series parts alone."

This makes shopping for computers with Intel CPUs incredibly frustrating. It's impossible to tell what a CPU supports without getting the full product name and plugging it in http://ark.intel.com/

Quite disappointing gains considering the move to 22nm, 3d transistors, and being delayed (from Jan).

I could live with the overclocking/heat issues if it meant more undervoltage headroom- but this doesn't seem to be the case either.. odd. Perhaps the process will improve over time? Load power is down, but the 3770K is only where the 2500K was.. yet to see where the 3570K lies. GPU increases are welcome but not thrilling.

I was waiting for this for an upgrade, but i'm left unsure 3570k vs 2500k. Think i'll wait for some more info on clocks vs volts/temps.

You have to say it's a no-brainer for anyone considering a non-K cpu however; cheaper, faster, better GPU, less power. And they don't lose virtualisation, either.

Methinks the point is to survive & solidify the transition to the new manufacturing tech before going whole-hog on capability improvements. Such is the nature of Intel's "tick-tock" development cycles: get fundamental and boring changes down cold before showing what awesomeness can come of them.

But a new process is at least supposed to provide 30% more performance even without awesomeness, not 10%.

Let's check that claim for the 65nm -> 45nm transition of the Core2Duo.

According to

http://en.wikipedia.org/wiki/List_of_Intel_Core_2_microproce...

The fastest 65nm Core2Duo with 65W TDP was the 3.0GHz E6850, released in July 2007. The first batch of 45nm CPUs was released in January 2008, the fastest one being the 3.17GHz E8500. A whopping 6% clock speed improvement, though admittedly the cache size grew from 4MB to 6MB. Bus speed stayed the same. I doubt that resulted in a 30% performance improvement overall, 10-15% seems more likely.

They followed up with some significantly faster CPUs a few months later, and I suspect they will do so with Ivy Bridge, too.

Transistor density went up about 60%. Power per performance is down about 20%. Performance per clock is up about 5-10% for some loads. Idle power is mostly unchanged because it's dominated by leakage in modern CPUs and that gets worse with process shrinks (note that Ivy Bridge isn't showing much change in idle consumption despite being a smaller die). Really, except for the 3.5 month slip, this is a pretty typical new process release.

You don't get a Sandy Bridge "wow factor" release with a die shrink. It used to be that they didn't do a big product reveal at all. The P5 core saw (I think) three die shrinks over its lifetime but was always just marketed as the X MHz Pentium.

More or less what the rumors were predicting. slightly faster CPU, significant power reductions, significantly better GPU.

Exactly, and the review nails it when it says that "It's a bridge to 22nm, warming the seat before Haswell arrives". Intel is being fairly conservative (later in the review, the real estate on chip to push the graphics further is mentioned), but the fact that all the improvements you list come at no additional costs for the user - despite having a brand new fabrication process - is nice.

Haswell's Fused Multiply-Add is the thing which excites me the most. I remember seeing that on the roadmap when it came out, and wondering how I'd live until 2013. It's weird that 2013 is next year now.

Really? There are lots of architectures that have that already, and really when you're using the vector unit you're going to be limited more by bandwidth than by execution resources. I hear that it will let you have more precision in the intermediate state, though, which some scientific computing people will care about.

Here's a very good summary of the changes in the GPU http://www.realworldtech.com/page.cfm?ArticleID=RWT042212225...

Ouch, this is horrible news for Intel they are ~10% faster than the year old CPU and use about 20% less power. The horrable GPU still get's dominated by 80$ GPU's even though it uses a large chunk of the 300$ CPU's transistor and power budget.

PS: I am saying this as someone that was going to upgrade just for the sake of it. But, even though the cost is irrelevant I don't think it's worth the time to swap motherboards. And if you read their comments section I am not the only one.

It is great news for those of us shopping for laptops. They will have better battery life and decent GPU performance, as well as USB3.

I would be shocked if the battery life increased by more than 3% because idol is almost identical and CPU's don't eat that much power to begin with. If you actually use the GPU your better off with a cheaper CPU and a dedicated graphics card from a power, cost, and performance standpoint.

Anyway, as long as Intel has such a significant performance lead and still charges reasonable prices it's mostly moot. I just wish they where still hungry for more.

The GPU might be over a generation behind in terms of performance, but it's embedded so it has a smaller footprint, a lower cost, and a lower power demand.

Most applications require little more than a basic GPU. Facebook isn't going to run any faster with 3,000 shaders on hand.

I'm using a circa 2007 core 2 duo. I genuinely see no improvement in CPUs since then in price/performance.

Is there a reason I should upgrade my kit that I'm missing?

In terms of raw CPU power, the i5/i7 was a big jump over the Core 2 Duo (both in going to quad core and general efficiency). This latest development, though? Not so much. A high end i5 or i7 is the best bang for the buck right now.

I'm still using a summer 2006 one! I think next year I'll be able to both rationalize and afford a complete upgrade of my desktop - I was waiting for it to be worthwhile to drop a large amount on a stupidly crazy gaming desktop, so a top of the line haswell-based cpu, a reasonably high-end graphics card or two, an SSD and bucketloads of RAM... yeah!

(We'll see if that actually comes to pass.. more likely I'll end up replacing my laptop instead of my desktop, despite that the desktop is what I use to play games and its starting to show its age now.. reasonable graphics card, ancient everything else)

On Mac laptops there's not much reason. You probably have an Nvidia GPU, which is still a bit faster than Intel ones and supports OpenCL. An upgrade would probably get you more max memory, and maybe twice the CPU performance, but it's not as amazing as a five year difference used to be.

I put an SSD + 2GB memory into my early-2006 Core "1" Duo mac mini, and it's almost as usable as my 2010 Macbook (hybrid SSD/HD with 8GB) when doing light tasks (browsing, note-taking, iTunes).

Yes, the SSD is only SATA1, but that's still 4-6x faster on synthetic benchmarks.

I couldn't find all the numbers for any 2007 mainstream notebook cpu, but here is a comparison between 3720QM (an upper mainstream, 2012) and P8400 (upper mainstream 2008):

    Model           3DMark06 CPU    Cinebench R10 32Bit Single    Cinebench R10 32Bit Multi    SuperPI 1M*    SuperPI 32M*    wPrime 32
    3720QM              6470                4810.5                    16822                        11             594                12.9
    P8400               2014                2526.9                    4700.6                       22.1          1212                34.2
                         321%                 190%                     358%                        50%            49%                 38%

So you would see 2x performance in single threaded, 4x performance in multithreaded and probably way better battery life due to stuff like TurboBoost (and getting a new battery :-)

Thanks for this - appreciated :)

Perhaps I should take a look at a newer machine! Battery life is what is super-important to me.

That's an impressive die shrink as well. Seems like Intel can make the 4 core Ivy in about the same die area as the 2 core Sandy. Considering the 4 cores are going to sell for quite a bit more that has to be great for margins.

Great for them, what does it to for the end consumer?

I'm guessing this is the natural tilt after they (effectively) defeated AMD using their slimy tactics through out the last decade. The customer always loses when the monopoly decides breaking the law (and paying off the judges + any fines) is more profitable than competing fairly.

I can't wait for Intel's mobile reckoning to occur. I still don't see their tech being anywhere near energy efficient to compete with ARM. It's a matter of time before Apple, Google, and Microsoft have effectively cut Intel out of the processor monopoly.

Related discussion:

http://news.ycombinator.com/item?id=3879473

Will there be a version without the GPU?

Reply on news.ycombinator.com