Haswell to Broadwell IPC

[Hulk]

If you are interested I normalized the recent Broadwell 5775C (3700MHz) and Haswell 4790K (4400MHz) scores for clockspeed.

Broadwell IPC advantage as follows (in percent):
Cinebench Single Thread: 3.2
Cinebench Multi Thread: 4.0
Handbrake v0.9.9 LQ: 1.6
Handbrake v0.9.9 2x4K: 13.3
Hybrid x265, 4K: -0.1
Agisoft Photoscan Benchmark: 17.7
Dolphin Emulation Benchmark: 2.0
WinRAR 5.01: 44.6
3D Particle Movement Single Threaded: 2.6
3D Particle Movement Multi Threaded: 3.3
FastStone Image Viewer 4.9: 1.5
Kraken 1.1: 23.1
WebXPRT: 17.4
Google Octane: 27.1


Average Haswell to Broadwell: 11.5%

Average Haswell to Broadwell, not including WinRAR: 9.0%

Average Haswell to Broadwell, not including WinRAR, Kraken, WebXPRT, Google Octane: 4.9%

EDIT - Some posters are stating that the 4790K actually operates at 4200MHz with 4 cores loaded. If the 4790K is actually running 4200MHz in all benches except the single threaded one then the averages are as follows. This is kind of hard to believe because it would mean that in many of the benches not using Crystalwell or special functions that Broadwell has worse IPC than Haswell.

Average Haswell to Broadwell: 7.1%

Average Haswell to Broadwell, not including WinRAR: 4.7%

Average Haswell to Broadwell, not including WinRAR, Kraken, WebXPRT, Google Octane: 1.1%

 

[Arachnotronic]

That's pretty darn close to Intel's claim of 5.5% uplift from HSW to BDW.

 

[crashtech]

Within contemporary expectations, I would think. Thanks for breaking it down!

 

[Enigmoid]

This is largely pointless other than a ballpark estimate unless you actually take note of the turbo speed during the benchmark.

 

[CHADBOGA]

If you are interested I normalized the recent Broadwell 5775C (3700MHz) and Haswell 4790K (4400MHz) scores for clockspeed.

Broadwell IPC advantage as follows (in percent):
Cinebench Single Thread: 3.2
Cinebench Multi Thread: 4.0
Handbrake v0.9.9 LQ: 1.6
Handbrake v0.9.9 2x4K: 13.3
Hybrid x265, 4K: -0.1
Agisoft Photoscan Benchmark: 17.7
Dolphin Emulation Benchmark: 2.0
WinRAR 5.01: 44.6
3D Particle Movement Single Threaded: 2.6
3D Particle Movement Multi Threaded: 3.3
FastStone Image Viewer 4.9: 1.5
Kraken 1.1: 23.1
WebXPRT: 17.4
Google Octane: 27.1


Average Haswell to Broadwell: 11.5%

Average Haswell to Broadwell, not including WinRAR: 9.0%

Average Haswell to Broadwell, not including WinRAR, Kraken, WebXPRT, Google Octane: 4.9%

Does the i7 4790K always hit its turbo speed of 4400Ghz?

 

[SteveGrabowski]

Does the i7 4790K always hit its turbo speed of 4400Ghz?

At stock clocks it runs 4.2 GHz when all cores are being used.

 

[positivedoppler]

Given Haswell's higher clock speed, it's a pointless upgrade unless there's some incredible oc
I wonder if DX12 will support Intel igpu with either amd or nvidia gpu.

 

[sefsefsefsef]

Unfortunately, it's not an apples to apples comparison, because i7-4790k doesn't have the 128MB L4 cache. It would be better to compare it to i7-4770R, and then normalize for clockspeeds.

 

[iSkylaker]

Isn't a clock-per-clock comparison a better idea?

 

[crashtech]

Isn't a clock-per-clock comparison a better idea?

Yeah, but oddly enough no one is doing that yet. I'm wondering if it's some part of an agreement with Intel to only allow review sites to release the data from stock clocks right now.

 

[ClockHound]

Yeah, but oddly enough no one is doing that yet. I'm wondering if it's some part of an agreement with Intel to only allow review sites to release the data from stock clocks right now.

Not a single enthusiast site has some Broadwell OC info. A coincidence? You decide. Sure seems like a information management campaign to me. But why?

 

[MrTeal]

Not a single enthusiast site has some Broadwell OC info. A coincidence? You decide. Sure seems like a information management campaign to me. But why?

With all these chips in reviewers hands, it's hard to imagine OC numbers not leaking.

Anandtech mentioned some pretty severe BIOS issues which is why they only have part 1 now. Perhaps overclocking is having some issues across the board on Z97 right now.

 

[jj109]

http://weekly.ascii.jp/elem/000/000/341/341607/

4.2 GHz at +0.05 V, no boot at 43x.

There's more luck involved since Broadwell C is binned for 65W TDP at 3.3-3.7 GHz, not for >4.0 GHz overclocking like Haswell DC.

 

[VirtualLarry]

There's more luck involved since Broadwell C is binned for 65W TDP at 3.3-3.7 GHz, not for >4.0 GHz overclocking like Haswell DC.

Do you think that Broadwell's FIVR components are limited physically in such a way that end-users might be limited in their ability to overclock Broadwell CPUs over 4Ghz, even if the core silicon could take it?

 

[coercitiv]

Do you think that Broadwell's FIVR components are limited physically in such a way that end-users might be limited in their ability to overclock Broadwell CPUs over 4Ghz, even if the core silicon could take it?

The alleged 5Ghz overclock would contradict such limitations.

 

[.vodka]

Do you think that Broadwell's FIVR components are limited physically in such a way that end-users might be limited in their ability to overclock Broadwell CPUs over 4Ghz, even if the core silicon could take it?

If the rumors are to be believed, Skylake won't have the FIVR and we'll be able to see which one overclocks better over 4GHz since both are 14nm. Yeah, it'll then be different architectures but some comparison could still be made.

Average Haswell to Broadwell: 11.5%

Average Haswell to Broadwell, not including WinRAR: 9.0%

Average Haswell to Broadwell, not including WinRAR, Kraken, WebXPRT, Google Octane: 4.9%

Nice increase from that L4 cache at this point, over the official 5.5% figure. I hope Intel starts including it going forward in more CPUs.

 

[Sweepr]

Edit: wrong thread.

 

[LTC8K6]

Impressive given the clock speeds we have seen these chips get to.

 

[sm625]

I'd like to see more data before I am willing to believe that kraken performance really did increase by 23%.

 

[JoeRambo]

I'd like to see more data before I am willing to believe that kraken performance really did increase by 23%.

128MB of L4 does wonders with cache miss happy code and all that JS "processing" is like that. Actually AT own benches have results with decent jump when adjusted for clock.

 

[Headfoot]

Lol, I love people who still put "processing" in quotes when talking about javascript. Programming language elitists unite

 

[escrow4]

Going off those scores it will be a long time before I upgrade this 4770 non K in this box.

 

[tenks]

Sorry for my ignorance, but when I first read the OP I understood normalize as he made the clocks the same. And then a few replys later I learned that's not what he did and others are asking for a clock for clock comparison. Thats not what he did? If not, then what did he do and why, I'm really confused?

 

[StinkyPinky]

At stock clocks it runs 4.2 GHz when all cores are being used.

This always bugged me. If all four cores run at 4.2....then why is it marketed as just 4.0?

 

[ctsoth]

This always bugged me. If all four cores run at 4.2....then why is it marketed as just 4.0?

Because different motherboards and OEMs are more or less agressive when it comes to Turbo Frequencies. My mobo will turbo all 4 cores to 4.4