News - Intel Haswell-E prices and launch slides leaked

[HEXUS]

Pricing ranges from $389 for the Intel Core i7 5820K up to $999 for the Core i7 5960X.

Read more.

[kalniel]

Memory speed support seems to be very low - at only 2133mhz there's only limited benefit to DDR4. I'll be waiting for ~3200 to become mainstream, and hope the CPUs will have some support for it.

[Tabbykatze]

That memory speed is probably only the recommended max but could probably push it far further.

A lot of lanes and a lot of gucci figures there!

[Corky34]

Memory speed support seems to be very low - at only 2133mhz there's only limited benefit to DDR4. I'll be waiting for ~3200 to become mainstream, and hope the CPUs will have some support for it.

So does the non Haswell-E if you go by what Intel claims is supported, they claim DDR3-1600.

[kalniel]

So does the non Haswell-E if you go by what Intel claims is supported, they claim DDR3-1600.

Which is fine and equivalent to the DDR4-3200 I'm looking for.. so I'm not hugely impressed by DDR4-2133.

[directhex]

That's a big price gap between the three models

[ecat]

Looks like another entry in the very long line of yawn inducing CPU releases from Intel, and no, the other bunch are not doing any better.

More cores/threads: May benefit some but it is still a struggle to find SW that can fully use more than 4 cores, add another core if you must to cover OS overheads.

More PCI lanes: Great if you need more GPU throughput, at what point does the CPU become the bottle neck these days? Also great for the new 'Go Even Faster' FlashDrive, these are perhaps the next truly worthwhile upgrade, you should buy one right now to use as a boot partition - that dig was uncalled for, sorry, PCIe flash drives are looking quite good.

DDR4: Okay, I'll have to buy a new motherboard anyway, may as well buy some new ram. What effect will this have on overall cpu speed?

More USB and more SATA: Great but nothing I cannot add to my existing PC for considerably less money than a full upgrade.

In fact let's do this...
Cost of a 4 channel SATA 3 expansion card?
Cost of a USB3 expansion card?
Cost of a nice PCIe flash drive?
Cost of a nice GPU to upgrade you existing system?
Cost of some extra RAM?

Now let's compare the above with the final costs of a new MB, new RAM and new CPU.

----
Finally, two of my favourite quotes from the slides

"Up to 79% more multi-thread performance over 4 core platform"

How do they achieve this? By adding another 4 cores + hyper threading of course. Doh!


"Over a Decade of Extreme Edition Innovation" "Over 40x Compute Performance Since 2003!"

In 12 years they have moved from the quite ****ty (ie NetBurst) single core Pentium 4 series which topped out at a clock speed around 3.7GHz to the new 16 thread i7 5xxxx series which tops out at a clock speed around 3.7GHz.

What?
3.7GHz twelve years ago vs 3.7GHz now, in 2014?
The maximum official CPU clock speeds have remained pretty much unchanged for over a decade. Even if you allow for an overclock to 4.7GHz, the maximum clock speed generally achievable on a desktop PC has increased by about 25% over the past 12 years.

But what about power consumption?
Okay, I'll grant you the point. The P4 was a true egg fryer, many improvements have been made wrt all aspects of power consumption.

And what about Cache, memory bandwidth and other IPC improving technologies?
Using Intel's own x40 as the overall increase, we can attribute x16 to threading which leaves a whopping 40/16 = 2.5x. In 12 years the true effective speed of your CPU has increased by a factor of 2.5 approximately.

People comment on the slow downs in the PC market, some apparently wonder why this should be happening...

The last 12 years, most certainly the last 6, have been the most mind numbingly boring years x86 CPU development has ever seen...

Could these two statements be connected?

[nacasatu]

Looks like it'll be another 2 years then by the time I upgrade, going on these prices :s <sigh>

[shaithis]

The main reason for people to use these over standard boards/chips was the PCI-Express lanes.......which has been pretty much made moot with the move to PCI-E 3.0....or I guess if you need 12 or 16 threads!

[kalniel]

The main reason for people to use these over standard boards/chips was the PCI-Express lanes.......which has been pretty much made moot with the move to PCI-E 3.0....or I guess if you need 12 or 16 threads!

Or SSDs really - taking up 4 pci-e 3.0 lanes doesn't leave enough for high end CF/SLI on desktop boards.

Add in quad channel, DDR4, and oodles of cores/cache etc. and they're quite tempting for workstation loads.

[Tunnah]

Finally, two of my favourite quotes from the slides

"Up to 79% more multi-thread performance over 4 core platform"

How do they achieve this? By adding another 12 cores of course. Doh!


"Over a Decade of Extreme Edition Innovation" "Over 40x Compute Performance Since 2003!"

In 12 years they have moved from the quite ****ty (ie NetBurst) single core Pentium 4 series which topped out at a clock speed around 3.7GHz to the new 16 thread i7 5xxxx series which tops out at a clock speed around 3.7GHz.

What?
3.7GHz twelve years ago vs 3.7GHz now, in 2014?
The maximum official CPU clock speeds have remained pretty much unchanged for over a decade. Even if you allow for an overclock to 4.7GHz, the maximum clock speed generally achievable on a desktop PC has increased by about 25% over the past 12 years.

But what about power consumption?
Okay, I'll grant you the point. The P4 was a true egg fryer, many improvements have been made wrt all aspects of power consumption.

And what about Cache, memory bandwidth and other IPC improving technologies?
Using Intel's own x40 as the overall increase, we can attribute x16 to threading which leaves a whopping 40/16 = 2.5x. In 12 years the true effective speed of your CPU has increased by a factor of 2.5 approximately.

People comment on the slow downs in the PC market, some apparently wonder why this should be happening...

The last 12 years, most certainly the last 6, have been the most mind numbingly boring years x86 CPU development has ever seen...

Could these two statements be connected?

You talk about clock speed as if that's what defines a CPU's performance. If the 3.7ghz hadn't improved, then why do single thread benchmarks consistently show otherwise ? Or here's a test for you, break out a Pentium 4, and disable all but 1 core on your latest chip and see how they perform day to day side by side.

Sure CPUs haven't improved much for the consumer who likes to do regular upgrades and feel the difference, but there also hasn't really been a need to. Back in the day I would upgrade every year, 2 tops, and every time it was amazing..that's when clock speed increases did most the work as everything was single core. But then multi-core came along and the standard work load just became a dream; in the 4 years since I bought my i5 I have not once thought "hmmm I could really do with a new CPU" because it's fast enough for everything I do.

One thing about this article though:
"Those seeking to create a CrossFire or SLI graphics PC could benefit from the Intel Core i7 5930K or above with the dual x16 PCI-e lanes."
Doesn't PCIe 3.0 at 16x/8x have way more than enough bandwidth for SLI ? I'd only recommend the more expensive chips if you wanted to go Tri-SLI.

[kalniel]

One thing about this article though:
"Those seeking to create a CrossFire or SLI graphics PC could benefit from the Intel Core i7 5930K or above with the dual x16 PCI-e lanes."
Doesn't PCIe 3.0 at 16x/8x have way more than enough bandwidth for SLI ? I'd only recommend the more expensive chips if you wanted to go Tri-SLI.

Yes, I'd have thought x8/x8 would be more than enough at v3 PCI-E. But maybe Hexus know something about some future cards or something.

[shaithis]

One thing about this article though:
"Those seeking to create a CrossFire or SLI graphics PC could benefit from the Intel Core i7 5930K or above with the dual x16 PCI-e lanes."
Doesn't PCIe 3.0 at 16x/8x have way more than enough bandwidth for SLI ? I'd only recommend the more expensive chips if you wanted to go Tri-SLI.

I thought that was odd also. I'd have thought that the biggest benefit would be from less CPU bottlenecking with 2 or 3 cards. Even 4x PCI-E shows minimal fall-off in performance still, so the new cards would have to be a gigantic step-up in performance to require 16x PCI-E 3.0 to operate at optimum.

[ecat]

You talk about clock speed as if that's what defines a CPU's performance. If the 3.7ghz hadn't improved, then why do single thread benchmarks consistently show otherwise ? Or here's a test for you, break out a Pentium 4, and disable all but 1 core on your latest chip and see how they perform day to day side by side.

Err.....

And what about Cache, memory bandwidth and other IPC improving technologies?

I think I have that covered. We can look further into cache levels and sizes, memory bus width + data transferred per clock + latency and pipelines + branch prediction + out of order execution and other goodness? I reckon your effective IPC is up around x2.5 over 12 years and a large part of that came from the move to Core2 iirc so maybe we should move our start point to 2006 and see how the numbers shake out.

Sure CPUs haven't improved much for the consumer who likes to do regular upgrades and feel the difference, but there also hasn't really been a need to. Back in the day I would upgrade every year, 2 tops, and every time it was amazing..that's when clock speed increases did most the work as everything was single core. But then multi-core came along and the standard work load just became a dream; in the 4 years since I bought my i5 I have not once thought "hmmm I could really do with a new CPU" because it's fast enough for everything I do.

... and "The PC Market is in Decline" as a result.
Quite seriously, you cannot come up to me and justify a spend on 10 new PCs based on the claim that they are 30% faster than the ones you currently use. The productivity gains in applications that do require solid single core (or even less than 6 core because 'threading is difficult') performance are simply not there.

I hope Intel are simple dragging their heals as a result of the lack of competition. It is indeed sad if 4.5GHz is truly the best speed modern technology can offer.

Edit:
As I said, this looks to be another Yawn Feast(TM) launch.

If a sufficiently large segment of the user base, enthusiasts, media can hammer this point home then maybe, just maybe we can goad Intel into making "Something Worth the Upgrade Cost".

[shaithis]

As I said, this looks to be another Yawn Feast(TM) launch.

If a sufficiently large segment of the user base, enthusiasts, media can hammer this point home then maybe, just maybe we can goad Intel into making "Something Worth the Upgrade Cost".

Indeed....what makes it worse is that they keep changing the socket. By all means release a CPU that is merely 5% faster.....but when you kill off compatible motherboards doing it? Very sly.
At the current rate, my 2 Ivy chips will last until their motherboards die.

[CAT-THE-FIFTH]

The Core i7 5820K is limited to 28 PCI-E lanes unlike higher end Haswell R SKUs it seems.