Broadwell-K i7 5775C/i5 5675C

Dave2150 · 23 Mar 2015 at 19:15

So, finally the clock speeds and cache size have been leaked:

Looks very disappointing for anyone who was waiting for these, or Z97 owners who assumed the 5th generation of I7's would have better performance than the 4th generation.

Compared to the 4790k, the top Broadwell CPU (I7 5775) has 2MB less of L3 cache, 700Mhz lower base clock and 700Mhz lower turbo clock.

The only saving grace for these CPU's will be the 128MB of L4 cache (Iris Pro) which can be utilized to aid CPU performance when a discrete GPU is installed. Anandtech calculated this to offer rough 10% extra performance in most CPU tasks.

Intel quote Broadwell to be a 5.5% IPC increase over Haswell. When coupled with the 10% from the 128MB l4 cache will provide an estimated 15.5% IPC increase, though I highly doubt actual benchmarks will actually demonstrate this, probably more like a 7% increase combined.

Source: http://chinese.vr-zone.com/146637/intel-lga-1150-broadwell-for-desktop-only-have-i7-5775c-and-i5-5675c-03232015/

Also doesn't bode well for Skylake that their 14nm process has regressed clock speed so severely, though maybe they'll have a better silicon process for those chips.

Martini1991 · 23 Mar 2015 at 19:19

What's the C denote?

And I'm not sure why anyone would expect anything more than ~10% improvement. It's actually better for everyone that they don't go crazy and violate AMD.
The next logical step for Intel basically makes AMD all but redundant. (I3's becoming what we know as i5's, the i5's becoming what we know as the i7 X7XX and then the i7 becoming a hexcore)

I don't know why people get hyped for Intel launches at the moment....

Although, pretty sure your figures are all wrong

Dave2150 · 23 Mar 2015 at 19:30

Martini1991 said:
What's the C denote?

And I'm not sure why anyone would expect anything more than ~10% improvement. It's actually better for everyone that they don't go crazy and violate AMD.
The next logical step for Intel basically makes AMD all but redundant. (I3's becoming what we know as i5's, the i5's becoming what we know as the i7 X7XX and then the i7 becoming a hexcore)

I don't know why people get hyped for Intel launches at the moment....

Although, pretty sure your figures are all wrong

No idea what the 'C' means. These CPU's are unlocked parts, however. Unknown why the dropped the 'K'.

Which figures do you believe to be wrong?

Intel stated that Broadwell is on average a 5.5% IPC increase from Haswell.

With 700Mhz less clock speed than the 4790k, 2MB less l3 cache, I very much doubt this CPU will equal it in performance.

As I said above, the only unknown at the moment is the 128MB l4 cache, Anandtech estimated this to give on average a 10% performance increase.

Though I still don't think this l4 cache will be enough to let it equal Devil's Canyon in performance.

Skiddley · 23 Mar 2015 at 19:43

C indicates unlocked.

skaif · 23 Mar 2015 at 19:46

glad I got my 4790k when i did then! Going to have no urges to upgrade for a while

Ronald · 23 Mar 2015 at 20:27

So improve IPC by 10-15%, loose over 15% on clock speed, both base and turbo...

The only saving grace will be power consumption as TDP is down significantly... so even without soldered heatspreader they should run cooler.

wazza300 · 23 Mar 2015 at 20:40

the naming is confusing

clock for clock it should be quicker so might even out the clockspeed drop

Dave2150 · 23 Mar 2015 at 21:12

Seems that LGA1150 Broadwell chips are the direct successors to the 4770R (http://ark.intel.com/products/76642/Intel-Core-i7-4770R-Processor-6M-Cache-up-to-3_90-GHz?q=4770R).

Exact same CPU on 22nm Haswell, though the Haswell part had higher turbo clocks.

Interestingly, the 4770R has the same TDP, 65W, suggesting the die shrink to 14nm netted them absolutely nothing - clock speeds are lower than on Haswell, and TDP is the same.

I guess they didn't do a 4790k successor due to the 14nm process clocking worse than Haswell did, so it would be impossible for them to make a faster CPU than the 4790k.

Andy7 · 23 Mar 2015 at 21:23

Disappointing as usual from Intel since the Sandy Bridge chips.

Read my thread here: http://forums.overclockers.co.uk/showthread.php?t=18623237

malachi · 23 Mar 2015 at 21:29

Chances are these will work on z87 chipset and not only z97 if the changes are that small.

Dave2150 · 23 Mar 2015 at 21:40

malachi said:
Chances are these will work on z87 chipset and not only z97 if the changes are that small.

Chances are that no-one on Z87 or Z97 will even care, as this CPU is very unlikely to outperform the 4770k or 4790K, imo.

It's clocked slower than both parts, has 2MB less L3 cache than both parts.

The only thing it has going for it is the 128MB of l4 cache.

Hopefully these clock speeds are not final, maybe there are engineering sample clock speeds.

Sylver123 · 23 Mar 2015 at 21:46

This doesn't look good for the mainstream parts compared to the 4790k. Hopefully the high end Broadwell parts that are getting released on the x99 socket 2011-3 platform are much better.

nails666 · 23 Mar 2015 at 22:00

Well if those stats are true then I'm glad I just went for the 4790K rather than waiting for these.
I can't quite believe that they wont release a 'proper' top end chip that betters the 4770K/4790Ks though.

malachi · 23 Mar 2015 at 22:48

nails666 said:
Well if those stats are true then I'm glad I just went for the 4790K rather than waiting for these.
I can't quite believe that they wont release a 'proper' top end chip that betters the 4770K/4790Ks though.

Maybe because they have no reason to. Not as if AMD are keeping them on their toes.

rossiboy001 · 24 Mar 2015 at 09:46

All the more reason to stay on sandybridge for another year, i believe a 2700k will still be enough to handle a couple of top end cards like 390x, maybe not 3.

David Bisset · 24 Mar 2015 at 10:56

Of course the claimed 5.5% could already be including any gains from the extra cache which would be unfortunate.

r22snapper · 24 Mar 2015 at 11:25

I can't help thinking about Skylake. For all the much-trumpeted 15-20% IPC increase that we are allegedly going to see, it is no gain if the chips are 10-15% lower clockspeed than Haswell.

DragonQ · 24 Mar 2015 at 11:26

David Bisset said:
Of course the claimed 5.5% could already be including any gains from the extra cache which would be unfortunate.

Also, you can't just add percentage increase from two sets of gains as in the OP. For example, if IPC improves by 10% and clock speed improves by 10%, the total increase would be 21%, not 20% (assuming linear performance increase per clock cycle, which isn't quite true usually).

Dave2150 · 24 Mar 2015 at 11:42

DragonQ said:
Also, you can't just add percentage increase from two sets of gains as in the OP. For example, if IPC improves by 10% and clock speed improves by 10%, the total increase would be 21%, not 20% (assuming linear performance increase per clock cycle, which isn't quite true usually).

My estimates in the OP are just rough estimates. The only concrete figure we got from Intel was the 5.5% IPC increase over Haswell. This was without the 128MB l4 cache.

Dave2150 · 24 Mar 2015 at 11:43

r22snapper said:
I can't help thinking about Skylake. For all the much-trumpeted 15-20% IPC increase that we are allegedly going to see, it is no gain if the chips are 10-15% lower clockspeed than Haswell.

Wouldn't surprise me if Skylake is launched only for mobile/tablet etc this year, and Desktop will come next year, when the 14nm process is capable of higher clocking parts.

My god we've been stuck on 22nm Haswell's for what seems like infinity, almost as bad as the 28nm node GPU's are stuck on.