Just going to call this right now

[BallaTheFeared]

10-20GHz computers in 2015

lol my chip runs at 20GHz, 5GHz x 4.

 

[Chiropteran]

Are serious? How FX8150 performed in that test is irrelevant, my point was about load power usage.

And that is my point as well. The majority of tests where FX-8150 is badly beaten are not full load. You can't use a full load power test to show single threaded power usage. You also can't use a single threaded performance test to show full 8 threaded performance.

I've already disputed your claim. The graph I posted shows the FX-8150 25% faster than the 2500k. The graph wasn't cherry picked either, I picked THE SAME TEST as you used.

The ball is in your court now, if you think you can prove that the FX-8150 is using that much power in a single threaded game or application please post the graph.

Spoilers: it doesn't.

 

[BallaTheFeared]

The graph I posted shows the FX-8150 25% faster than the 2500k. .

Yes 25% faster, while using 60% more power.

 

[sm625]

Some levels of intel cache are 3 times faster. We're getting close to orders of magnitude gaps in performance here. AMD supposedly specializes in gpu but they cant even use their gpu do what quicksync does. We're going to see more and more specialized sections of the cpu die that perform one-off functions very fast, like Quicksync. I've been arguing for an SSD controller for half a decade now. A javascript engine and a custom adobe flash renderer would be other obvious choices. The goal is to be able to run a web browser without waking up the cpu. There is no reason for the cpu cores to be awake if all you're doing is scrolling a webpage that was loaded 38 seconds ago. Intel has the resources to work with these companies and create dedicated silicon for them. AMD doesnt.

 

[Vesku]

"custom adobe flash renderer "

Kill it with fire.

Also, where is the 13 and 14 inch platform HP? Forget selling own branded RAM, AMD needs to push decent specs for their mobile chips.

 

[Khato]

The ball is in your court now, if you think you can prove that the FX-8150 is using that much power in a single threaded game or application please post the graph.

http://www.tomshardware.com/reviews/fx-power-consumption-efficiency,3060-12.html

According to the "Test Setup and Benchmarks" page of that review, the single-threaded efficiency 'benchmark' was a run of the following applications in order:

We ran the applications for the efficiency test in the following order: Adobe Acrobat, Winzip, iTunes, Lame, 3ds Max. Not surprisingly, it performs worse in the single-threaded efficiency metric than the multi-threaded one. (SNB is at 9.6Wh/22Wh for the single/multi threaded efficiency tests while the FX-8150 is at 25Wh/44Wh, so only 2x more power instead of 2.5x.)

 

[pelov]

The goal is to be able to run a web browser without waking up the cpu. There is no reason for the cpu cores to be awake if all you're doing is scrolling a webpage that was loaded 38 seconds ago. Intel has the resources to work with these companies and create dedicated silicon for them. AMD doesnt.

I'm guessing that can be mitigated by core parking and smart schedulers. The issue here would be that this isn't a silicon problem but rather involves the OS as well. It is a great idea, though.

As far as Quicksync goes, it isn't exactly Intel's baby but rather Intel's licensed baby. AMD has always been hesitant to reach out to other tech companies for help whereas Intel hasn't shied away from that approach. Seeing AMD license the resonant clock mesh tech and buying up SeaMicro does signal that they have been changing in that respect, which is great for them because although they might spend more money licensing that stuff it's still cheaper in the end because you didn't pay for R&D.

They've both realized that they're not the only show in town anymore and are being forced to get a little creative with what they have to offer. An x86 license is nice but it isn't that impressive nowadays

 

[IntelUser2000]

... you are aware that quick sync is for ENCODING video, not DECODING it, right? in terms of multimedia playback (aka DECODING video), AMD's IGP eats Intel's alive.

Majority of the benefits QuickSync offers is due to its decoding capability: http://www.tomshardware.com/reviews...d-app-nvidia-cuda-intel-quicksync,2839-7.html

You can see small benefits by enabling hardware encoding, but huge benefits enabling hardware decoding.

 

[frozentundra123456]

Honestly, even as someone who buys/supports AMD, the only reason it seems like AMD is catching up and Intel has plateaued is because Intel is getting to the point where it's hard to improve on something that's already so good, while AMD has a long ways to go. It'll be interesting though if AMD ever actually caught up again, but that's being hopeful. Intel has to compete with themselves so honestly I don't really care about the whole "AMD needs to be around or intel will blah blah". I just like having more options. The integrated graphics on AMD setups are looking awesome compared to Intel's....so that's nice.

I think this is just kind of the life cycle of a product. The large jumps in performance become increasingly hard to achieve as the product matures. I dont overclock so IB is not a big disappointment to me. It will have marginal CPU performance increase, quite a bit better graphics, and use less power. It seems more of an advance vs SB in the mobile sector, while basically a side-grade on the desktop where power usage and the IGP are not so important.

The biggest problem that I see is that if the chip runs as hot as early leaks indicate, I wonder how adversely this will affect laptops. It doesnt do much good to have longer battery life if the laptop runs excessively hot. This could also indicate a problem with Intel for further die shrinks and getting into tablets/phones.

I dont want to get too much into the AMD vs Intel argument, because a lot of people seem to favor one or the other blindly. If IVB has as many problems as it seems, AMD may have a chance to gain some ground though. AMDs plans sound good, but lets wait for the actual products to see if they deliver on their targets. It will also be interesting to see whether Haswell comes out on time and if Intel can get back to making major performance increases or if they have indeed plateaued out.

 

[PlasmaBomb]

Is that the power consumption of the CPU only? Cause my 2500K draws 100W in prime95 @ 4.5GHz according to HW monitor.

It's Idontcare's data and the y axis is clearly labeled as power consumption at the wall (ie. total system draw from wall socket before taking into account PSU efficiency). Hope that helps

 

[Khato]

The biggest problem that I see is that if the chip runs as hot as early leaks indicate, I wonder how adversely this will affect laptops. It doesnt do much good to have longer battery life if the laptop runs excessively hot.

There's a difference between actual silicon temperature right in the middle of the densest, hottest spots on a chip and the heatsink/surface/exhaust temperature. Given the same cooling, a chip that draws less power will result in lower heatsink/surface/exhaust temperatures. However, given the same cooling, a fictitious 10W 10mm^2 would have a silicon delta temperature compared to the heatsink 5.3x greater than a 30W 160mm^2 IVB despite the heatsink/surface/exhaust delta temperature compared to ambient being 3x less.

All evidence thus far points to the change in thermal resistance due to the smaller process/die size being the cause for high temperatures on IVB when overclocked. I do wonder what possible solutions Intel might be researching to overcome this issue as the only alternative is to not shrink logic as much as would otherwise be possible (some evidence of this on IVB as an individual core's logic appears to only go from ~18mm^2 on SNB to 14mm^2.)

 

[IntelUser2000]

(some evidence of this on IVB as an individual core's logic appears to only go from ~18mm^2 on SNB to 14mm^2.)

The cores are actually ~11mm2. Based on earlier pictures it ends up being high end of 13mm2, but that's not the real pic: http://images.anandtech.com/doci/4798/IMG_0257_575px.JPG

This is how Ivy Bridge really looks like: http://forums.anandtech.com/showthread.php?t=2234017

 

[Khato]

The cores are actually ~11mm2. Based on earlier pictures it ends up being high end of 13mm2, but that's not the real pic: http://images.anandtech.com/doci/4798/IMG_0257_575px.JPG

Quite true. Now I'm trying to remember where I heard the 14mm^2 figure. Regardless, that's more in-line with expectations at getting ~66% of theoretical scaling.

This is how Ivy Bridge really looks like: http://forums.anandtech.com/showthread.php?t=2234017

Nice guesses regarding the IVB GT1 layout and HSW possibilities, though they aren't exactly correct. Sadly, I expect that we'll have to wait until fall IDF to get a glimpse of HSW GT3.

 

[IntelUser2000]

Nice guesses regarding the IVB GT1 layout and HSW possibilities, though they aren't exactly correct. Sadly, I expect that we'll have to wait until fall IDF to get a glimpse of HSW GT3.

Of course.

Ivy Bridge won't be that much different, they have limited ways to cut. I suspect Haswell might be much more off.

 

[DominionSeraph]

Majority of the benefits QuickSync offers is due to its decoding capability: http://www.tomshardware.com/reviews...d-app-nvidia-cuda-intel-quicksync,2839-7.html

Which is only relevant in encoding. Playback is at 1x.

 

[Smartazz]

It's possible, but Ivy Bridge is probably going to be a hit in the mobile space even if it's not too great for enthusiasts.

 

[TuxDave]

If AMD can beat out Haswell or Broadwell core for core...

... then they clearly are much better at their job than I am.

 

[Arachnotronic]

If AMD can beat out Haswell or Broadwell core for core...

... then they clearly are much better at their job than I am.

You work for Intel?

 

[Suspicious-Teach8788]

Yes 25% faster, while using 60% more power.

lol.

It's actually 71% more power.

Or how about compare to the 2600k where its 1% slower but uses 47% more power? You know what kind of comparison that sounds like? It's like when you play some game and some idiot cheats, but you still beat him at the game. That's how bad of a beatdown this is.

 

[Chiropteran]

http://www.tomshardware.com/reviews/fx-power-consumption-efficiency,3060-12.html

Thanks, proves my point. Singled threaded power usage for FX-8150, 133W. A far cry from the absurd 405W figure posted by the guy who started this line of discussion. FX uses a lot of power, sure. 405W, hell no.

Yes 25% faster, while using 60% more power.

Thanks for the backup. Faster is faster. It uses more power, this isn't news to anyone. The common misconception that it uses 400W while still falling behind is just hyperbole.

 

[IntelUser2000]

Which is only relevant in encoding. Playback is at 1x.

Right, but the point is decoding isn't worse compared to AMD. If anything, it might even be better.

 

[IntelUser2000]

Thanks, proves my point. Singled threaded power usage for FX-8150, 133W. A far cry from the absurd 405W figure posted by the guy who started this line of discussion. FX uses a lot of power, sure. 405W, hell no.

Of course, that's system power consumption. FX-8150 probably runs close to its TDP at stock, its hard to tell exactly with overclock since other components may use more power(more I/Os processed, more memory used, etc).

 

[Rifter]

No, just no. The fact that AMD resorted to months and months(years?) of lies(JF's IPC is better claims to the bitter end, among other lies) shows that AMD is in no way ready to challenge intel at anything. Thats not going to change in the short term. Intel just has to large of a fab advantage, the only way AMD can compete is if the use intel fabs or buy there own that are on the same level as intels.

 

[moriz]

Right, but the point is decoding isn't worse compared to AMD. If anything, it might even be better.

It isn't. Intel's video device capabilities is pretty rudimentary compared to AMD's.

 

[Zap]

I just think Intel's becoming stagnant, at least compared to AMD, who are trying their hardest to fight back. IVB on mainstream should've been a bigger improvement -- either more cores, more cache, etc.

IVB is going according to plan.

Intel Tick Tock Model

Basically it is risky to do BOTH a process shrink along with new designs. So, Intel separates the two. Sandy Bridge was a new design. Ivy Bridge was a process shrink.