Core i7 "Haswell-E" Engineering Sample Pictured

[buildzoid]

The only thing that's really intresting about this chip is that it has 8 cores and that's only because intel didn't want to release a Sandybride-e 8 core with an unlocked multiplier because if you look at IPC scaling and OC headroom scaling on SB IVB HW then you can deduce that these will average 4.4Ghz and have 15% more IPC than the SB-es which do anything from 4.5 to 5Ghz so the hexa core ones will be outclassed by the 4 year old SB-e hexa cores and the 8 cores will be barely faster than SB-e hexa cores.
My favorite manufacturing process is the 32nm SOI that FX processors are done on. It OCs well under LN2 it OCs well under water it OCs well on air and it can survive really high voltage much longer than intel's 32nm or 22nm and is easier to cool on the CPU to heatsink heat transfer level not the heatsink dissipation level.
If AMD made it possible to stick multiple FX chips onto one motherboard my 3960X could find it self saying bye bye because FX 8350s aren't nearly as expensive to replace when you fry them and are funner(more options and better IMC) to OC

 

[radrok]

Sheesh guys you are getting platforms wrong, the one that's supposed to be power consumption friendly is the mainstream one.

People who buy this platform could care less about 140W TDP or 200W TDP to be honest.

And you still need to take a look at performance per watt. That's the relevant number, not the absolute TDP.

 

[Steevo]

Sheesh guys you are getting platforms wrong, the one that's supposed to be power consumption friendly is the mainstream one.

People who buy this platform could care less about 140W TDP or 200W TDP to be honest.

And you still need to take a look at performance per watt. That's the relevant number, not the absolute TDP.

Funny cause when a AMD is overclocked it is very close in performance to much more expensive Intel offerings in all but a few synthetic benchmarks like super pi, but so many people still buy Intel chips and pay more for an extra 1- .5 FPS. So it is perception, nothing more than perception that drives people wanting to purchase these.

 

[radrok]

With AMD you have solid multi threaded performance, with Intel you have both single and multithreaded performance and that's enough for me to choose Intel since I need both.

I don't think any consumer AMD processor can come close to a 5,1 GHz 3930K though.

 

[cadaveca]

P.S. has the availability & price of DDR4 already good?

It should be widely available by the time the platform launches, speeds are low (as is to be expected), and prices will be high(as expected).

With AMD you have solid multi threaded performance, with Intel you have both single and multithreaded performance and that's enough for me to choose Intel since I need both.

I don't think any consumer AMD processor can come close to a 5,1 GHz 3930K though.

I concur.

 

[buildzoid]

With AMD you have solid multi threaded performance, with Intel you have both single and multithreaded performance and that's enough for me to choose Intel since I need both.

I don't think any consumer AMD processor can come close to a 5,1 GHz 3930K though.

Yes at that point AMD cpus lose in everything except for H.264 encoding where they're still equal

 

[TheoneandonlyMrK]

Based on how 140W Intel or 125 W AMD CPUs overclock, I think the reactions are justified. If you run stock, then, yeah, I feel as you do.


But...clockspeed is gonna take a dive here, with 8 cores, I suspect. If not, AWESOME!!!

Hope to have more info soon.

unless that engineering samples your's how would you know how this 140watt 8 core overclock's,,, to be able to compare it to a 125 watt Tdp Amd chip , you cant hence his post was valid ,now all start calling intel on their efficiency already ,oh look 2011 - 3 now thats socket longevity.
and justified??,,,, piledriver(FX) is a year old now so should show poorer efficiency then this anyway, still,,, Go intel eh

 

[cadaveca]

unless that engineering samples your's how would you know how this 140watt 8 core overclock's,,, to be able to compare it to a 125 watt Tdp Amd chip , you cant hence his post was valid ,now all start calling intel on their efficiency already ,oh look 2011 - 3 now thats socket longevity.
and justified??,,,, piledriver(FX) is a year old now so should show poorer efficiency then this anyway, still,,, Go intel eh

Yeah, you're right, the compare is not 100% valid, but for the end consumer, they compare AMD vs Intel, if they even know AMD, so I can make the compare well enough, even if not appropriate.


As to clockspeed, the obvious thing is that is how it will work. If IPC increase enough, it's not a big deal, but you know that those that like to overclock, like bigger numbers, and that's all. I get lower clocks with IVB-E than SB-E, but still, IVB-E is faster.

Of course, I am making an assumption on the 3 GHz stock speed of that ES, for sure. That's the fun of pre-release news..we can guess what will happen, and then find out later. I love speculation.

 

[TheoneandonlyMrK]

Yeah, you're right, the compare is not 100% valid, but for the end consumer, they compare AMD vs Intel, if they even know AMD, so I can make the compare well enough, even if not appropriate.


As to clockspeed, the obvious thing is that is how it will work. If IPC increase enough, it's not a big deal, but you know that those that like to overclock, like bigger numbers, and that's all. I get lower clocks with IVB-E than SB-E, but still, IVB-E is faster.

Of course, I am making an assumption on the 3 GHz stock speed of that ES, for sure. That's the fun of pre-release news..we can guess what will happen, and then find out later. I love speculation.

Fair enough,, as is 140 watts for 8 real cores and 16 logical ones, and as is 125 watt for 8 logical AMd cores fair enough imho.

I also thought socket 2011 was still rev1 though so rev3 is crazy and no backwards compatibility make's calling it socket 2011 rev3 pretty damn daft imho,, madness since some are fully aware what an interposer exactly is ,it makes all this socket swapping nonesense the money grab it is(i do except ddr4 as a reasonable excuse but not within the same socket type/lineage unless back compat)

 

[cadaveca]

Well, as the cores get smaller, and smaller, maintaining clocks is even harder and a more exact "science" is needed. I think Intel and AMD are both doing good jobs, and their products are appropriately priced (given that I am pretty sure that sales have increased since the beginning of the year), so if a socket change is needed, that doesn't bother me much, but at the same time, I would like the change to also bring more than just some USB and SATA speed.... but what they'd add I have no idea.

Moving to a new DRAM controller makes the socket change make sense, really, more pins to the DIMMs, I think...

 

[TheoneandonlyMrK]

Well, as the cores get smaller, and smaller, maintaining clocks is even harder and a more exact "science" is needed. I think Intel and AMD are both doing good jobs, and their products are appropriately priced (given that I am pretty sure that sales have increased since the beginning of the year), so if a socket change is needed, that doesn't bother me much, but at the same time, I would like the change to also bring more than just some USB and SATA speed.... but what they'd add I have no idea.

Moving to a new DRAM controller makes the socket change make sense, really, more pins to the DIMMs, I think...

yeh I agree but they should have given it a clearly different socket name, as it's more then a revision if no older socket 2011 cpu fits it ,,,still ,be nice in a folder/cruncher wouldn't it

 

[radrok]

I guess they just name their socket based on the pinout, it still has 2011 pins so 2011-3 makes sense

 

[librin.so.1]

Fair enough,, as is 140 watts for 8 real cores and 16 logical ones, and as is 125 watt for 8 logical AMd cores fair enough imho.

comparing AMD's "logical" cores to Intel's logical ones is like comparing apples to screwdrivers. Please don't do it.

 

[buildzoid]

As to clockspeed, the obvious thing is that is how it will work. If IPC increase enough, it's not a big deal, but you know that those that like to overclock, like bigger numbers, and that's all. I get lower clocks with IVB-E than SB-E, but still, IVB-E is faster.

Tell that to my WATER COOLED 3960X's Cinebench R11.5 score because the closest WATER COOLED i7 4xxx chip score on HWbot is a 4930K and it's more than 0.4 points behind

Fair enough,, as is 140 watts for 8 real cores and 16 logical ones, and as is 125 watt for 8 logical AMd cores fair enough imho.

The 8 cores = 16 only works when there are different logical operations to do simultaneously so if you have loads and loads of multi threaded floating point only calculations you only have 8 cores worth of compute but when you have something like Interger addition running parallel with floating point division you get 16cores but with slowed down and shrunken cache and extra load on IMC resulting in a maximum of +30% multi threaded performance(intel's number). Which explains why an FX8350 is on par with an intel hexa core when encoding video(a whole load of the same simultaneous operations).

 

[HumanSmoke]

Funny cause when a AMD is overclocked it is very close in performance to much more expensive Intel offerings in all but a few synthetic benchmarks like super pi, but so many people still buy Intel chips and pay more for an extra 1- .5 FPS. So it is perception, nothing more than perception that drives people wanting to purchase these.

Not necessarily. Intel's HEDT platform is often chosen over the (Intel) mainstream chipset boards because of the native tri/quad GPU support. There is more than enough evidence available (even without trawling the HWBot and 3DMark leaderboards) to suggest AMD's platform suffers in comparison with Intel in multi-GPU and mGPU+high res....somewhat ameliorated if you're comparing an overclocked FX with a stock Intel proc I guess.
http://www.anandtech.com/show/6934/choosing-a-gaming-cpu-single-multigpu-at-1440p/9
http://www.tomshardware.com/reviews/fx-8350-core-i7-3770k-gaming-bottleneck,3407.html
http://vr-zone.com/articles/amd-fx-...-8ghz-multi-gpu-gaming-performance/17494.html
http://www.extremetech.com/computin...te-gaming-showdown-5ghz-fx-9590-vs-i7-4960x/2 (1080p w/HD 7990 and $1K CPU :smh: )

 

[btarunr]

For as long as I can remember, this is the longest time frame I have ever seen Intel stay on the same socket for any particular platform.

Same pin count, but not same socket. You can't use IvyB-E on an X99 motherboard, or Haswell-E on your X79.

 

[TRWOV]

Yeah, I would have posted "OMFD 'DAT TDP", but then I saw Your post. Really, wasn't Intel supposed to be the low power consumption "team" while AMD was pushing "outrageously high" TDPs like 125w?
I suppose even the Low TDP Saints Intel have problems dealing with the increasing leakage currents of their ever-shrinking process. kekeke...

With Haswell, Intel moved part of the VRM section onto the die so it has a handicap in that regard. Even so, 140w is a pretty good number considering the 3970X (Ivy Bridge-E hexacore) has a 150w TDP.

 

[buildzoid]

With Haswell, Intel moved part of the VRM section onto the die so it has a handicap in that regard. Even so, 140w is a pretty good number considering the 3970X (Sandy Bridge-E hexacore) has a 150w TDP.

The Ivybridge-E hexaxcores are all 130W

 

[BorisDG]

3970X (Ivy Bridge-E hexacore) has a 150w

3970X is Sandy-E. Ivy-E (4930K/60X) as buildzoid said is 130W.

 

[TRWOV]

oh, yeah, I forgot to take Intel's screwed naming conventions into account

Still my point stands, going from a 130w hexacore to a 140w octacore is no small feat.

 

[Hayder_Master]

"and a PCI-Express gen 2.0 root complex for third-party onboard controllers" ???


and why still only 40 lines ? when they going to increase it? i want to see 4X pci-e 16x

 

[Aquinus]

"and a PCI-Express gen 2.0 root complex for third-party onboard controllers" ???


and why still only 40 lines ? when they going to increase it? i want to see 4X pci-e 16x

That's a crap ton of extra pins that they would have to shove into the CPU socket. Keep in mind that the bulk of those pins are DRAM and PCI-E and 2011 pins is already a pretty big socket. I think that 40 lanes for a single CPU is pretty awesome to begin with. As for 16/16/16, it's really not a big deal. PCI-E 3.0 has enough bandwidth to drive a modern GPU on 8 lanes and even then, 2011 is gear more towards servers, so the idea is that those slots aren't always going to be used for GPUs, but instead a RAID controller, fibre card, or some other high bandwidth I/O device.

Also, Intel does have a 150-watt TDP 12c IVB-EP Xeon, but you're paying out the nose for it. 4k USD or something like that.

 

[meSaad]

Does Haswell-E release before Broadwell ? Does release date has been unveiled? When?

 

[Steevo]

Not necessarily. Intel's HEDT platform is often chosen over the (Intel) mainstream chipset boards because of the native tri/quad GPU support. There is more than enough evidence available (even without trawling the HWBot and 3DMark leaderboards) to suggest AMD's platform suffers in comparison with Intel in multi-GPU and mGPU+high res....somewhat ameliorated if you're comparing an overclocked FX with a stock Intel proc I guess.
http://www.anandtech.com/show/6934/choosing-a-gaming-cpu-single-multigpu-at-1440p/9
http://www.tomshardware.com/reviews/fx-8350-core-i7-3770k-gaming-bottleneck,3407.html
http://vr-zone.com/articles/amd-fx-...-8ghz-multi-gpu-gaming-performance/17494.html
http://www.extremetech.com/computin...te-gaming-showdown-5ghz-fx-9590-vs-i7-4960x/2 (1080p w/HD 7990 and $1K CPU :smh: )

This helps to show that if building a new PC Intel is still the way to go unless you need heavily threaded computations or are on an moderate budget. Some or most of these are horrible hypothetical scenarios (it shows the weakness well though), who runs crossfire but couldn't afford higher resolution than 1080?

AMD still has an issue with memory latency in the CPU's and it shows.

 

[Deadlyraver]

Glad I'm waiting