Soldato
- Joined
- 14 Nov 2012
- Posts
- 17,955
- Location
- Close to Swindon, but not Swindon
I would - you may kick yourself for upgrading early!
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Intel demonstrates 65W Broadwell-K at GDC
- Thread starter Dave2150
- Start date
I would - you may kick yourself for upgrading early!
Found something very interesting: http://www.anandtech.com/show/9070/intel-xeon-d-launched-14nm-broadwell-soc-for-enterprise
Intel just launched 'Xeon-d' - featuring 14nm 8 core Broadwell CPU's, clocked at 2.0Ghz (2.2Ghz for the 6 core skew) at just 45W TDP.
The fact that an 8 core Broadwell clocked at 2Ghz has only 45W TDP bodes very well for Broadwell-K and Skylake, in my opinion. We also have to consider that Xeon-D has the Southbridge integrated into the die, that's bound to consume some of the TDP budget aswell, if only a few watts.
The 4 core Broadwell-K at 65w should be able to be clocked quite high, knowing this about the Xeon parts. I understand the Iris Pro with 128MB of l4 cache consumes some of the TDP budget, though I imagine them clocking to 4.0-4.3Ghz at least.
Intel also confirmed that Broadwell gives a 5.5% IPC increase over Haswell (see the slides in the article I linked for the source).
WTB some leaked engineering sample information!
Intel just launched 'Xeon-d' - featuring 14nm 8 core Broadwell CPU's, clocked at 2.0Ghz (2.2Ghz for the 6 core skew) at just 45W TDP.
The fact that an 8 core Broadwell clocked at 2Ghz has only 45W TDP bodes very well for Broadwell-K and Skylake, in my opinion. We also have to consider that Xeon-D has the Southbridge integrated into the die, that's bound to consume some of the TDP budget aswell, if only a few watts.
The 4 core Broadwell-K at 65w should be able to be clocked quite high, knowing this about the Xeon parts. I understand the Iris Pro with 128MB of l4 cache consumes some of the TDP budget, though I imagine them clocking to 4.0-4.3Ghz at least.
Intel also confirmed that Broadwell gives a 5.5% IPC increase over Haswell (see the slides in the article I linked for the source).
WTB some leaked engineering sample information!
Possible low-clocked Skylake engineering sample benchmarked;
http://www.kitguru.net/components/c...chitectural-advantages-of-skylake-processors/
Looks to be no improvement for single core compared with Haswell, but getting on for 20% better under the multi-core bench. All under Geekbench benchmark.
Take with a pinch of salt.
http://www.kitguru.net/components/c...chitectural-advantages-of-skylake-processors/
Looks to be no improvement for single core compared with Haswell, but getting on for 20% better under the multi-core bench. All under Geekbench benchmark.
Take with a pinch of salt.
An Intel-claimed 5.5% fits pretty much exactly with the ~3% figure we know from the lower-end parts.
Don't expect high clocking cpu's if you want to avoid feeling let down. Just because a xeon model fits 8 2GHz cores at a 45W TDP has no indication on the clock ceiling. They simply aren't refining and evolving their fabrication process with high clocks as their main objective.
Soldato
- Joined
- 9 Oct 2009
- Posts
- 9,266
- Location
- United Kingdom
This would be my main concern. I would consider upgrading from my i7 4770K if the single threaded performance increase was significant but I'm not holding my breath.
Intel didn't make a 45w 2Ghz 8 core product on Haswell. I still believe it bodes well for Broadwell, that it's able to do so.
Broadwell might have a clock ceiling that's lower than Haswell, it also might produce golden pixie dust when in operation, we simply don't know.
Excited to find out
Unless things have changed. Then I think it's still supposed to be second quarter 2015. So hopefully not to long now
I would love to see a big performance improvement, but there won't be. As stated a few posts back its all about power consumption and mass production. As it stands the die shrinks are doing a great job decreasing power consumption on the mobile market, silicon is easily made in bulk compared to graphene (which I hear is still in its infancy?)
When the competition put out something worthy of noting then Intel will up their game. As it stands you need a mini power plant to power competitive hardware.
When the competition put out something worthy of noting then Intel will up their game. As it stands you need a mini power plant to power competitive hardware.
Broadwell-K with 128MB of l4 cache should be a good 10% faster than Haswell, if it's clocked similarly to the 4790k. 5.5% IPC gain over Haswell, plus the 128MB l4 cache adds upto double digit performance gains, in CPU tasks, when using a dedicated GPU.
Skylake should be another 15-20% improvement over Broadwell, if rumours are true.
Well I hope you're right to be honest. I just remember when Ivy Bridge rumours first surfaced, and we we're hearing crazy numbers that were completely unrealistic.
I guess the 128MB of L4 cache will make a difference, only time will tell exactly how big a difference.
I guess the 128MB of L4 cache will make a difference, only time will tell exactly how big a difference.
Associate
- Joined
- 31 Oct 2012
- Posts
- 2,243
- Location
- Edinburgh
5.5% in some scenarios, not all - running benches we're less than 10% IPC improvement going from 1st gen i5 to haswell in some cases!
Last edited:
First to 5th generation I5 or I7 would definitely be way more than 10% in every scenario.
Would be interesting to see comprehensive tests from all 5 generations though, to see just how much things have changed.
Soldato
- Joined
- 1 Jun 2010
- Posts
- 7,057
- Location
- Bedford
Looking around the web, it seems generally going from Nehalem to Broadwell is around 30-35% increase in IPC performance. Still quite significant.
Associate
- Joined
- 31 Oct 2012
- Posts
- 2,243
- Location
- Edinburgh
Not sure *every* scenario, though I'd grant you most would be.
As an example, ages ago I pulled some numbers from the firestrike thread and compared some physics scores (CPU scores) between generations - there may be more now but at the time there was only one 4690k result so I used that, along with my own i5-750.
Validated scores:
4690k @ 4.725GHz = 9413
i5-750 @ 3.2GHz = 5946
At first glance we see the newer part is doing much better - but is it higher clock speeds or is it IPC? If we calculate the scores as points per GHz we get:
4690k: 1992
750: 1858
That shows us a 7.2% improvement clock for clock. Not the most convincing step up! However, the overall performance increase is large and many scenarios will be much more flattering for the newer chips
Edit: Of course I could use a higher-clocked i5-750 @ 3.8 result of 7042 of my own to get a higher score for the earlier CPU, basically the same points per GHz
It'd be slightly interesting to find the program that shows the greatest IPC difference - presumably something using new instruction sets? Get a bit of a map of changes generation to generation
Last edited:
Decent article from Anandtech comparing first to fourth generation I7's in productivity:
http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/7
It's much harder to find valid gaming benchmarks, as you need to compare all the generations of i7 using a decent GPU, such as a 290, 970 or 980 etc. The older GPU's won't show so much of a difference between the CPU's. SLI/Crossfire further increases the difference between all i7 generations.