If you are waiting for Intel's new crop of desktop CPUs for 2013 (expected to be released March-June), sadly reports so far on engineering samples appear to show about a 10% improvement ... the focus of these CPUs appears to be more around on-board graphics and the mobile sector (less power consumption) with only minor improvements to CPU raw performance (which is what FSX likes). So if you're getting 30 fps now in FSX, moving to Haswell-DT may get you 33 fps.

 

 

Haswell-DT should overclock well also and hit the same 4.6-5GHz that current 3960X and 3970X can reach without much effort (and remain stable). But Haswell is 4 core (no 6 or 8 versions).

 

For flight simulations that can indeed utilize more CPU cores effectively, Haswell-DT may not perform as good as existing 3960X and 3970X 6 cores units.

 

However, what Haswell-DT's release will do is drive down the price of the 3960X and 3970X -- so if cost is your consideration then waiting may indeed be beneficial.

 

It looks like 2014 is the year when we'll see significant improvements in Intel's CPU performance. I'm NOT suggesting that 10% is bad either, just in relation to FSX it's not really going to provide much improvement. However, X-Plane may be able to utilize Haswell-DT better than FSX if X-plane is updated to use AVX2 instruction set.

 

My sources are Intel engineers (some friends that work on the on the Cache system of Intel's CPUs) and I'm passing this info along because it really doesn't violate any NDA that isn't for the most part already known.

 

Rob

I don't see how this is disappointing at all. Ivy Bridge was a great chip and many reported an improvement in FSX. If the TIM on the Haswells is good and the chip doesn't run too hot then it might be a great chip. Anyway, I wouldn't judge just yet...

 

And 30 to 33 FPS is not a bad improvement. Imagine you getting 28 FPS at and airport so that the VSync tweak doesn't work anymore and actually creates stutters. If you now get 30 FPS this is great. It's not just about 1 computer part, it's everything together. All little bits help, and every bit gives you more headroom. Imagine if memory, videocard and CPU improve performance by 10% seperately, then altogether you'll get a great improvement wouldn't you?

The only competition they've got from AMD is the on-board graphics and the mobile sector, so it's not surprising to see Intel focusing on these. Performance-wise, AMD is far behind and proves no competition for Intel.

Yeah we need AMD to become a good comptitor again like they used to be in 2007, then Intel has got something to fight for so we'll be getting better CPUs.

Yeah we need AMD to become a good comptitor again like they used to be in 2007, then Intel has got something to fight for so we'll be getting better CPUs.

 

AMD have already said that they'll stop competing with Intel over performance. It's clear that AMD will never dominate again.

If you are waiting for Intel's new crop of desktop CPUs for 2013 (expected to be released March-June), sadly reports so far on engineering samples appear to show about a 10% improvement ... the focus of these CPUs appears to be more around on-board graphics and the mobile sector (less power consumption) with only minor improvements to CPU raw performance (which is what FSX likes). So if you're getting 30 fps now in FSX, moving to Haswell-DT may get you 33 fps.

 

 

Haswell-DT should overclock well also and hit the same 4.6-5GHz that current 3960X and 3970X can reach without much effort (and remain stable). But Haswell is 4 core (no 6 or 8 versions).

 

For flight simulations that can indeed utilize more CPU cores effectively, Haswell-DT may not perform as good as existing 3960X and 3970X 6 cores units.

 

However, what Haswell-DT's release will do is drive down the price of the 3960X and 3970X -- so if cost is your consideration then waiting may indeed be beneficial.

 

It looks like 2014 is the year when we'll see significant improvements in Intel's CPU performance. I'm NOT suggesting that 10% is bad either, just in relation to FSX it's not really going to provide much improvement. However, X-Plane may be able to utilize Haswell-DT better than FSX if X-plane is updated to use AVX2 instruction set.

 

My sources are Intel engineers (some friends that work on the on the Cache system of Intel's CPUs) and I'm passing this info along because it really doesn't violate any NDA that isn't for the most part already known.

 

Rob

 

5-15% is all we ever get anymore WRT single thread performance, all the low-hanging fruit has already been picked from an engineering standpoint. Even the infamous Nehalem -> Sandy Bridge transition only brought 15-20% more IPC.

 

I disagree with your speculation that the followup to Haswell (Broadwell) to be released in 2014 will bring more performance than Haswell, since Haswell is a new architecture (tock) and Broadwell is just a die shrink annual refresh (tick). There's never been a case of a tick bringing more performance than a tock.

5-15% is all we ever get anymore WRT single thread performance, all the low-hanging fruit has already been picked from an engineering standpoint. Even the infamous Nehalem -> Sandy Bridge transition only brought 15-20% more IPC.

 

I disagree with your speculation that the followup to Haswell (Broadwell) to be released in 2014 will bring more performance than Haswell, since Haswell is a new architecture (tock) and Broadwell is just a die shrink annual refresh (tick). There's never been a case of a tick bringing more performance than a tock.

 

Didn't Ivy Bridge bring 5% to 15% performance increase over Sandy Bridge, clock per clock? I'm sure Broadwell will bring even better performance.

 

If not, there's always Skylake and Skymont.

Didn't Ivy Bridge bring 5% to 15% performance increase over Sandy Bridge, clock per clock? I'm sure Broadwell will bring even better performance.

 

If not, there's always Skylake and Skymont.

 

Ivy Bridge was 5-10% faster than Sandy Bridge, falling right in the middle for FSX performance when paired with fast RAM, as FSX Mark 11 testing shows.

 

Since Intel switched to a "tick-tock" release cycle, what I have described already is exactly what we have seen. Intel releases a new architecture, then 12-18 months later they'll bring out a shrunken version of that chip to get ready for the new manufacturing process. That shrunken chip brings with it typically 5-10% performance increases whereas new architectures can add anywhere from 10-20%. You can see this for yourself by Googling it if you don't believe me.

You can be pretty sure that we'll not see Broadwell in 2013, it will be a 2014 part (at least according to my sources). Broadwell will benefit FSX more because of higher clock frequency because of the die shrink.

 

I don't know what most people want out of fps improvements, for me 3 fps gain (30-33 fps) isn't worth the cost. But for those on older hardware they may see, 25-50% gain pending their CPU, chipset, RAM, etc.

 

The biggest improvements to my FSX framerates and overall "fluidity" have been:

 

1. Overclocking CPU (in my case 5Ghz 3960X)

2. Quad channel ram (X79 chipset)

3. SATA3/6 with Intel SSDs (oddly some SSDs are actually not good with FSX)

 

It's not just about framerates, from most of my flying I don't really need to exceed 30 fps and the "fluidity" aspect is much more important to me ... the fluidity/smoothness was improved the most by moving to Quad Channel RAM (X79) and SATA3/6 Intel SSDs.

 

Video cards have done nothing to improve performance, my GTX 570 2GB, GTX 690 4GB, and ATI 5870 2GB produce almost exactly the same framerates on average ... for FSX that is. For X-plane 10, well that's an entirely different story, the GTX 690 had a very clear performance advantage over my ATI and GTX 570.

Going 30 ---> 33 FPS? Meh

 

And even if it is released in the spring, it probably won't be ready for primetime until several months later, after the early adopters feel the pain of having the initial kinks worked out.

 

Interesting stuff, but if I were looking for a new CPU now, I wouldn't be waiting for Haswell.

You can be pretty sure that we'll not see Broadwell in 2013, it will be a 2014 part (at least according to my sources). Broadwell will benefit FSX more because of higher clock frequency because of the die shrink.

 

I never said Broadwell was a 2013 part, I assumed you were referring to Broadwell coming in 2014, hence your statements about "expect more performance in 2014". I then stated that Broadwell is a refresh, and made the case that it is basically impossible for it to bring more performance to the table than Haswell will bring us. As for higher performance due to higher clock speed, that's basically a pipe dream, as we've seen with the transition from 32nm Sandy Bridge to 22nm Ivy Bridge, clock speeds are actually down at the high end because of Intel's ridiculous decision to use thermal paste instead of flux-less solder. If you care to ask your engineer friends why they hamstrung overclockers like that we'd all be delighted to hear the answer!

So if you're getting 30 fps now in FSX, moving to Haswell-DT may get you 33 fps.

 

More like you will be getting 36 FPS from SB.

Intel have A LOT of headroom in their CPUs ... I look at a CPU by how well it will overclock and remain stable. If you compare Intel clocks (the "released" cycles are well below what my Intel engineer friends have done with some of their samples) then yes, there isn't much of a difference.

 

But Broadwell will have much improved headroom thanks to the die shrink and reduced power consumption. I can assure you the released clocks has nothing to do with thermal paste vs. flux-less solder and everything to do with 150 Watt units and the problems that introduces to motherboard designs including line noise and other FCC requirements.

Intel have A LOT of headroom in their CPUs ... I look at a CPU by how well it will overclock and remain stable. If you compare Intel clocks (the "released" cycles are well below what my Intel engineer friends have done with some of their samples) then yes, there isn't much of a difference.

 

Yes, they have a lot of headroom, and if you read my previous post re: overclocking you would see that the shrunken CPUs actually have *LESS* headroom. If anything, Broadwell may overclock LESS than Ivy Bridge and Haswell (if the trend continues).

 

But Broadwell will have much improved headroom thanks to the die shrink and reduced power consumption. I can assure you the released clocks has nothing to do with thermal paste vs. flux-less solder and everything to do with 150 Watt units and the problems that introduces to motherboard designs including line noise and other FCC requirements.

 

:lol: This is funny. First of all, the FCC has absolutely nothing to do with this artificial power constraint you mention, otherwise motherboard manufacturers wouldn't be allowed to expose overclocking options to their end-users. Secondly, flux-less solder vs. thermal paste has been demonstrated to be responsible for the loss of several hundred MHz at the high-end for overclockers when comparing Sandy Bridge (32nm) vs. Ivy Bridge (22nm) due to the absurd temperatures produced when over-volting to achieve stability at the desired clockspeed. I know, I've done it on 4 different Ivy Bridge chips and have posted the results in this very forum.

Yes, they have a lot of headroom, and if you read my previous post re: overclocking you would see that the shrunken CPUs actually have *LESS* headroom. If anything, Broadwell may overclock LESS than Ivy Bridge and Haswell (if the trend continues).

 

I guess we disagree, less power consumption (which is what Broadwell brings) will provide MORE headroom (along with other efficiencies).

 

First of all, the FCC has absolutely nothing to do with this artificial power constraint you mention, otherwise motherboard manufacturers wouldn't be allowed to expose overclocking options to their end-users.

 

I guess we disagree again ... FCC influences motherboard designs please read: http://transition.fcc.gov/Bureaus/Engineering_Technology/Documents/bulletins/oet62/oet62rev.pdf Sections 15.107, 15.109, 15.101a and more.

 

Secondly, flux-less solder vs. thermal paste has been demonstrated to be responsible for the loss of several hundred MHz at the high-end for overclockers when comparing Sandy Bridge (32nm) vs. Ivy Bridge (22nm) due to the absurd temperatures produced when over-volting to achieve stability at the desired clockspeed. I know, I've done it on 4 different Ivy Bridge chips and have posted the results in this very forum.

 

Unless you performed these "tests" in a controlled environment and went over the results with a multitude of EEs, I wouldn't draw too much of a conclusion. In my own experience, reducing voltage along with other EFI settings has resulted in higher achievable clock rates upto 5Ghz and a very stable platform. But to each his or her own, glad you believe in your methods and they work for you.