First "real" Haswell CPU preview, 4670K

[BallaTheFeared]

I agree, but I will say I only care about mins and when the cpu is limiting.

My i3-540 gets decent "Avg" where my 7950 is limiting, however it's the "mins" when my cpu is limiting that concern me, and that's what I want to see... How much faster is this new cpu when it matters is all I care about.

I understand it's hard for reviewers to hit all these things, or specifically seek them in games, which is why I personally value low res cpu tests as it takes a lot less effort to get much closer results to those that I want to see in the first place.

Hopefully when the new review comes out they use low res and at least a Titan or one of the aftermarket GTX 780s.

 

[toyota]

The Tom's article looked hopeful, while that chinese article seemed to show no gain. Could be they were gpu limited. Clearly we need more reviews with a wide variety of games and GPUs to really know what kind of improvements are coming.

what Tom's article?

 

[Lepton87]

Spent $2K on a 3930K/Titan box a few days back. Still not interested. Haswell is much more a tock for mobile than performance.

Why Titan and not GTX780? Titan seems like such a huge waste of money now that 780 is out, that is unless you need FP64 compute performance.

 

[Lepton87]

I really want to see what the TDP is when the CPU is at 100% compared to SB and IB. This is not including the iGPU, because I don't really care about that. The overall TDP is up, but I really want to see what the cores in HW can do vs. SB and IB. That will hopefully tell us more about what performance to expect/watt.

You are confusing some terms, the TDP will always be the same, it is an arbitrary number given by the chip-maker. You probably meant power consumption numbers. I don't see how HW cores can draw less power then IB cores at 100% load, it's gonna provide around 2x the FLOPS of IB due to AVX2 and FMA3. I would rather see how they stack up in some real world loads. It's gonna be a while till we will start seeing apps using both AVX2 and FMA3.

The Tom's article looked hopeful, while that chinese article seemed to show no gain. Could be they were gpu limited. Clearly we need more reviews with a wide variety of games and GPUs to really know what kind of improvements are coming.

I agree about the wide variety of games but not GPUs, Titan,7970GHz,GTX690,7990 should be more then enough. The fastest single GPU and multi GPUs cards from both camps should be all that everybody could ask for. In reality low resolution results (720p or less)should be basically the same with any of those cards. I see no point in testing 7950 or GTX680 etc.

 

[frozentundra123456]

what Tom's article?

Sorry, wrong reference. Meant the article quoted by the op.

There was an earlier Toms article but that was not the one I should have referred to.

 

[Lonbjerg]

At 720p ???

Show me an IPG that can do maxed out gaming at +1080P...in something else besides PONG...

 

[Lonbjerg]

Why Titan and not GTX780? Titan seems like such a huge waste of money now that 780 is out, that is unless you need FP64 compute performance.

Why is it a waste...it's faster than the GTX780...why settle for second place?

Show me on the doll were the Titan touched you.

 

[AtenRa]

Show me an IPG that can do maxed out gaming at +1080P...in something else besides PONG...

Completely irrelevant

 

[Fjodor2001]

I think the real interesting benchmarks will be on power consumption. That's where Haswell is supposed to excel. Intel announced a few days ago that they are claiming 50% longer battery life compared to IB. If that is true it's quite impressive. But I'd like to see reviews from reputable sites confirming that before getting too excited.

 

[Lonbjerg]

Completely irrelevant

Good thing you consider IPG's' irrelevant at 1080p ^^
I take it you consider 30 FPS to be to low too right?

 

[colonelciller]

Conclusion: When you're cpu limited expect up to 20% more performance.

at what settings?

you CANNOT extrapolate data from one setting on a new CPU to other settings where data was not shown. That's not science, that's not engineering... that's MARKETING.

 

[BallaTheFeared]

Presumably at the same, but with lower res to lessen the gpu load thereby increasing the cpu load.

We don't want to get rid of things like reflections and shadows, since they're often cpu intensive.

 

[boxleitnerb]

at what settings?

you CANNOT extrapolate data from one setting on a new CPU to other settings where data was not shown. That's not science, that's not engineering... that's MARKETING.

Yes you can - under the assumption that you're in a CPU bottlenecked scene at this other setting, too. That's what it's all about. And it absolutely can happen, because absolute CPU load is independent of resolution:
If my CPU can do 40fps at 720p, it still can only do 40fps at 1080p or 1440p - no matter what GPU solution I have.

The only thing that happens at higher resolutions is, that you overshadow the demands on the CPU. But that doesn't mean you have to or that it is always like that.

 

[AtenRa]

Good thing you consider IPG's' irrelevant at 1080p ^^
I take it you consider 30 FPS to be to low too right?

I find 8-12fps at 1080p completely irrelevant as both your last two posts.

 

[bunnyfubbles]

Why is it a waste...it's faster than the GTX780...why settle for second place?

Show me on the doll were the Titan touched you.

well, he did pass up the 3970X and settle for the 3930K...

 

[Magic Carpet]

I agree, but I will say I only care about mins and when the cpu is limiting.

My i3-540 gets decent "Avg" where my 7950 is limiting, however it's the "mins" when my cpu is limiting that concern me, and that's what I want to see... How much faster is this new cpu when it matters is all

I seem to share your thinking. Minimum fps is what really spoils gaming experience and more often its the cpu to blame. Cant wait to see benches.

 

[Sable]

And thats exactly what i would like to know, if the CPU will grand me any gains at the resolution I and most of the people play and thats at 1080p. Bench the games at 720p but also show us if we will get any gains at 1080p as well(obviously not with low settings).

Would you consider upgrading a CPU+Motherboard if you would know that you will only gain a 2% at 1080p ??

There aren't any official reviews yet, just wait for techreport and anandtech to post theirs and you'll no doubt see some attual real world benchies.

 

[galego]

I think the real interesting benchmarks will be on power consumption. That's where Haswell is supposed to excel. Intel announced a few days ago that they are claiming 50% longer battery life compared to IB. If that is true it's quite impressive. But I'd like to see reviews from reputable sites confirming that before getting too excited.

With everything else disabled or shut down and idle status? Maybe

 

[Mallibu]

With everything else disabled or shut down and idle status? Maybe

That's why it says "up to". Pretty much how the supposed FX scheduler hotfix was "up to 10%" but proved to be 0.5% all around (). Except in this case the all around benefits will be around 25%

What are your machine specs again?

 

[Fjodor2001]

With everything else disabled or shut down and idle status? Maybe

Well what's most interesting is real usage scenarios. Web browsing, watching movies, etc. I wonder what battery life improvements we can expect then?

I wonder if Haswell will be able to beat the perf/watt improvements we saw going from AMD Zacate->Kabini.

 

[SiliconWars]

I wonder if Haswell will be able to beat the perf/watt improvements we saw going from AMD Zacate->Kabini.

There's no chance of that I'm afraid. Zacate to Kabini is one of the greatest improvements in perf/Watt we have ever seen and Haswell doesn't have the smaller process to help it.

 

[exar333]

You are confusing some terms, the TDP will always be the same, it is an arbitrary number given by the chip-maker. You probably meant power consumption numbers. I don't see how HW cores can draw less power then IB cores at 100% load, it's gonna provide around 2x the FLOPS of IB due to AVX2 and FMA3. I would rather see how they stack up in some real world loads. It's gonna be a while till we will start seeing apps using both AVX2 and FMA3.



I agree about the wide variety of games but not GPUs, Titan,7970GHz,GTX690,7990 should be more then enough. The fastest single GPU and multi GPUs cards from both camps should be all that everybody could ask for. In reality low resolution results (720p or less)should be basically the same with any of those cards. I see no point in testing 7950 or GTX680 etc.

TDP definitely will not be the same. Loading at 100% just the CPU vs. CPU + GPU is completely different. It provides more of an apples to apples comparison between CPUs when you remove the GPU power consumption. Hopefully we see 20% less power at most speed levels. I would LOVE to see better power consumption especially at pr around 4.5-5ghz, but that definitely remains to be seen....

 

[escrow4]

well, he did pass up the 3970X and settle for the 3930K...

I don't mind spending an extra $300 on a GPU. Spending an extra $400 for 3MB cache and maybe, just maybe, a better binned CPU, nope. Plus Sandy Bridge E isn't exactly new, I'm not going to sink that much into this platform. Broadwell E and DDR4, well sure, lets go all out.

Back on topic though, Haswell so far seems meh for the part enthusiasts care about - more grunt. Even the wider execution units and new extensions - so-so.

 

[Khato]

Well what's most interesting is real usage scenarios. Web browsing, watching movies, etc. I wonder what battery life improvements we can expect then?

I wonder if Haswell will be able to beat the perf/watt improvements we saw going from AMD Zacate->Kabini.

In actual processor constrained (medium to heavy load) scenarios? No, it'd be difficult to match that for a number of reasons. I find it somewhat amusing to note that quite a few expressed skepticism that Intel could deliver 4-5x better performance per watt with Silvermont than current generation Atom... despite the fact that AMD has just delivered 3x better performance per watt with Jaguar in some benchmarks and had a far better starting point.

Anyway, the more interesting metric with Haswell will be if the low load scenarios give the kind of increased performance per watt that Intel has been promising. There's always the possibility that the Haswell ULT SKUs will be more efficient than Jaguar in that case.

 

[Idontcare]

What are your machine specs again?

I see this becoming a meme in a matter of weeks