Friday, March 28th 2014

Intel Finalizes Feature-sets of the First Wave of 9-series Chipsets

Here's what the first wave of Intel 9-series chipsets will look like. It will consist of upper-mainstream chipsets for the LGA1150 platform, and the sole chipset for the LGA2011-3 HEDT platform. LGA1150 will get two new chipsets based on a common silicon, the Z97 Express, and the H97 Express. The Z97 will be the next top-end chipset for the platform, supporting current Core "Haswell" and upcoming "Haswell refresh" and "Devil's Canyon" processors. Motherboards based on the Z97 chipset, like the ones based on the Z87 and Z77, will feature up to three PCI-Express 3.0 slots wired to the CPU, in configurations of x16/NC/NC, x8/x8/NC, and x8/x4/x4. It will support overclocking, something its sibling the H97 Express will lack. The H97 will also lack the PCIe configurations that the Z97 supports.

Both Z97 and H97 will introduce support for PCI-Express M.2 storage, which offers 66.6% higher bandwidth than SATA 6 Gb/s, and the same bandwidth as SATA-Express. Its introduction will spur up development and launches of a new generation of high-performance client SSDs in 2014-15. Apart from M.2, the two chipsets offer six SATA 6 Gb/s ports with AHCI and RAID support. Rapid Storage Technology (RST) and Smart Response Technology (SRT) are common for both chips, however, the Z97 also features Dynamic Storage Accelerator (DST), something the H97 will lack. H97 will exclusively offer Small Business Advantage (SBA), instead. Both chips will feature 8-lane PCI-Express gen 2.0 root complexes to drive onboard devices, and 14 USB ports, including six USB 3.0 SuperSpeed ports.
The X99 Express is a different beast altogether. Designed for the upcoming "Haswell-E" HEDT platform, and the next-generation LGA2011-3 socket (incompatible with current LGA2011), the X99 platform is designed for high-end multi-GPU builds. It will let motherboard designers build boards with up to five PCI-Express 3.0 x16 slots wired to the CPU, in either x16/NC/x16/NC/x8, or x16/NC/x8/x8/x8, or even x8/x8/x8/x8/x8. Letting you toss in a screamingly fast PCIe SSD along with your 4-way multi-GPU setup. Intel appears to have gotten over the SATA+SAS mess it built with the previous X79/C608 common silicon, and gave the X99 a monolithic 10-port SATA 6 Gb/s storage controller, with RST and SRT support. SSD TRIM will be supported on RAID 0 configurations. An 8-lane PCI-Express gen 2.0 root complex drives onboard devices, just like the on the Z97/H97, and a 14-port USB controller, with 6 USB 3.0 ports completes its main feature-set. Intel plans to launch the three around Q2-Q3, 2014.
Source: Hermitage Akihabara
Add your own comment

55 Comments on Intel Finalizes Feature-sets of the First Wave of 9-series Chipsets

#26
lilhasselhoffer
I'm not willing to argue about the X79 PCH again. It's been done to death, and I'm pretty sure everyone has said their piece.

I'm pretty happy with how the SB-e silicon turned out. No matter how you view the PCH, a CPU that could easily overclock to an extra GHz frequency without even breaking a sweat is impressive.

At the same time, SB-Haswell hasn't produced any substantial performance gains in the mainstream. 10-15% increases amount to improvement, but nothing worth buying another system over. X99 is a great proposition if you're coming from something in the X58 or prior generations. If you're already running SB-e there's nothing there to justify that kind of expenditure. As much as I hate to say it, a $200 RAID card is way cheaper than a decent new motherboard.


I'm interested that Intel is actually bringing the mainstream I/O to the HEDT platform. I'm not happy, thrilled, or surprised. Wake me when I can buy a new platform that justifies another $1000 or more expense.



To those who use the car analogy, it is flawed. Please consider the following replacement. I own an old sports car. I can remove its engine, but to get the extra 200 hp I want stock I need to buy a new car and a new motor. Alternatively, I could install a new engine that fits for a mild performance boost, and couple it with a super charger (expansion device) to get the same functional hp increase. No matter what, my sports car beats out a Honda Civic released at the same time, and even most new Civics. It may not be new, but there's plenty of life still left in it.
Posted on Reply
#27
LAN_deRf_HA
TheGuruStudIt's just another shrink with differences like from IVB to Hasbeen. IPC is maxed and you won't see intel spending any money to change that. They will continue to shrink the dies and use the cheapest silicon possible to leverage profit. New high performance chips don't fit into that equation. You'll have to go Xeon (which is only faster b/c of more cores and cache of course).
It's long been said Skylake will be a bigger jump than most, not sure why, something about "sandybridge team" whatever that means.
Posted on Reply
#28
lilhasselhoffer
LAN_deRf_HAIt's long been said Skylake will be a bigger jump than most, not sure why, something about "sandybridge team" whatever that means.
Intel has several teams working on CPUs at any time. One team is developing SB production, another is finalizing IB silicon designs, and a third is beginning designs on Haswell (etc...). The SB team is coming up again, and the last product they brought to market was much better than its predecessors. It wasn't just a tick-tock step, but an actual jump forward. People are banking on the same type of show with Skylake, given the demonstrated talents of this team in the past.

I'd love to have the same views. Given the less than warm reception of SB-e, the initial issues with IB (SATA connectivity), and the relatively trivial progress that Haswell has made I'm ready for something great. I have high hopes for Skylake, tempered with rather low expectations.
Posted on Reply
#29
Octavean
the54thvoidWell, I suppose it depends on your usage. But I have 2 dedicated x16 pci-e lanes and 4 sata 3 ports. I've been on SB-E since Nov 11 and my system is still stupidly fast. X79 is only outdated in an abstract way. It's like saying a 1990's supercar is outdated. I think the term would be 'classic', even though x79 lacked it's own USB3 controllers and the like, they're all still on board.

With the pace of IPC or frequencies slowing compared to years back and the decent over clocking of the SB chips, there has been no need whatsoever to upgrade (whereas for example, gfx have changed).

X79 has been my best PC choice to date.
Generally speaking I agree.

My Core i7 3930K / Asus P9X79 Deluxe setup has severed me well for years and is still performing admirably. If I were buying a new platform id probably go with X99 / LGA2011-3 but I see no need to upgrade yet and I don't need to add another system at the moment.

I have tried multiple SSD units (up to 4) in a RAID 0 configuration on my Core i7 3930K / Asus P9X79 Deluxe system and on my Core i5 2500K / Asus P8P67 Pro system. This was to increase the single logical capacity more then anything else (up to ~1TB of pure SSD capacity) but it performed well in Both cases. If I recall correctly, Intel promised TRIM support for the X79 platform in RAID 0 but I'm unsure if they ever delivered on that promise. Anyway, TRIM support on X79 and X99 platforms is a reason (however uncommon or special case It might be) to run 2, 3, 4 or more SSD off of the motherboard rather then a RAID card (until Such time as RAID cards start providing TRIM support in RAID 0).

Honestly I'm very curious to see of X99 boards will ever receive Thunderbolt II support. I know a lot of people don’t care about Thunderbolt but that is in no small way due to its cost and availability. However, the X99 platform won't be cheap any more then X79 so it will already command a premium price. I probably wouldn't use Thunderbolt even If I had it but I am curious if Intel will allow Thunderbolt II on another platform without an Intel IGP besides the new Mac Pro,....
Posted on Reply
#30
Popocatepetl
LAN_deRf_HAJust hope Skylake is still coming 2015. Been waiting for a real jump since sandy.
First half of 2016 is more likely, lest you are only interested in preview articles rather than purchasing the real thing.
Posted on Reply
#31
64K
PopocatepetlFirst half of 2016 is more likely, lest you are only interested in preview articles rather than purchasing the real thing.
Where did you read about that? I thought Skylake was on schedule for 2015 unless you were referring to the die shrink of Skylake called Cannonlake in 2016. I would like to see your source because I'm planning on an upgrade with Skylake also next year.
Posted on Reply
#32
Prima.Vera
the next-generation LGA2011-3 socket (incompatible with current LGA2011)
Damn basterds, greed is everywere.
Posted on Reply
#33
Aquinus
Resident Wat-man
Prima.VeraDamn basterds, greed is everywere.
That doesn't mean the pin-out for those 2011 pins are all exactly the same, so it would have been incompatible anyways (DDR3 to DDR4, or did we forget?). Yes, Intel is just trying to make a profit like any other for-profit industry but Greedy are those people who buy EE chips, even more so when all you do with it is gaming.
Posted on Reply
#34
Octavean
Prima.VeraDamn basterds, greed is everywere.
Honestly I think this is more of a concern for people who actually own an older LGA2011 platform and are interested in upgrading to a LGA2011-3 platform. I actually have an LGA2011 platform and I do have "some" interest in buying into the new LGA2011-3 platform but I'm not too terribly bothered by the incompatibility. If I were to buy into the LGA2011-3 platform I would be buying a new processor and motherboard anyway,....

I still have an old Gigabyte LGA775 board that I installed a Core 2 Quad Q6600 on. The board eventually developed problems but it supported both DDR2 and DDR3. There is no doubt in my mind that Intel could have included DDR4 support and DDR3 support but it would have increased the complexity and likely the price of the hardware. So I guess I am playing devil's advocate here a bit,.....

The way I see it is that people typically want backward compatibility in cases like this because they want a cheap upgrade path but the Intel HEDT platform is usually anything but cheap so its a conflict of interest.

Is It that Intel is being greedy or is it that some upgraders are being cheap?

I'd say its a bit of both.
Posted on Reply
#35
lilhasselhoffer
OctaveanHonestly I think this is more of a concern for people who actually own an older LGA2011 platform and are interested in upgrading to a LGA2011-3 platform. I actually have an LGA2011 platform and I do have "some" interest in buying into the new LGA2011-3 platform but I'm not too terribly bothered by the incompatibility. If I were to buy into the LGA2011-3 platform I would be buying a new processor and motherboard anyway,....

I still have an old Gigabyte LGA775 board that I installed a Core 2 Quad Q6600 on. The board eventually developed problems but it supported both DDR2 and DDR3. There is no doubt in my mind that Intel could have included DDR4 support and DDR3 support but it would have increased the complexity and likely the price of the hardware. So I guess I am playing devil's advocate here a bit,.....

The way I see it is that people typically want backward compatibility in cases like this because they want a cheap upgrade path but the Intel HEDT platform is usually anything but cheap so its a conflict of interest.

Is It that Intel is being greedy or is it that some upgraders are being cheap?

I'd say its a bit of both.
I can see where you've come from, but your logic is flawed. In the days of LGA 775 the memory controller was soldered to the board, not integrated into the CPU. This would allow two controllers to exist, with some kind of switching between them allowing both memory standards to work.

In a modern CPU you'd have to increase die size dramatically to include an extra memory controller. That rather significantly increases silicon size, increases failure points, and would wind up making that $600 CPU closer to $800 (ballpark estimation) Not sure about you, but that sounds insane from where I sit.


I don't like Intel introducing LGA 2011-3, but it isn't unreasonable. We look at the same number of connectors, and see a pretty "obvious" cash grab. What we don't see is the electrical schematics reworking which pins do what, and where they do it. We don't see the reworking of the memory controller for DDR4. Despite not seeing this, people are calling foul. I may not like the new socket, but crying foul is not a reasonable response.
Posted on Reply
#36
HumanSmoke
lilhasselhofferI don't like Intel introducing LGA 2011-3, but it isn't unreasonable. We look at the same number of connectors, and see a pretty "obvious" cash grab. What we don't see is the electrical schematics reworking which pins do what, and where they do it. We don't see the reworking of the memory controller for DDR4. Despite not seeing this, people are calling foul. I may not like the new socket, but crying foul is not a reasonable response.
I'd tend to agree. Whilst some CPU pins are programmable, there are going to instances where a new pin layout is going to be required - especially since more voltage regulation features are moving to the CPU package - including a reworked electrical requirement for DDR4.

For those interested, here are what the three LGA2011's (LGA 2011-0 "Socket R", LGA 2011-1 "Socket R1", and LGA 2011-3 "Socket R3") look like
...and if you're into the nuts and bolts (land assignments start on page 212)
Posted on Reply
#37
Octavean
lilhasselhofferI can see where you've come from, but your logic is flawed. In the days of LGA 775 the memory controller was soldered to the board, not integrated into the CPU. This would allow two controllers to exist, with some kind of switching between them allowing both memory standards to work.

In a modern CPU you'd have to increase die size dramatically to include an extra memory controller. That rather significantly increases silicon size, increases failure points, and would wind up making that $600 CPU closer to $800 (ballpark estimation) Not sure about you, but that sounds insane from where I sit.


I don't like Intel introducing LGA 2011-3, but it isn't unreasonable. We look at the same number of connectors, and see a pretty "obvious" cash grab. What we don't see is the electrical schematics reworking which pins do what, and where they do it. We don't see the reworking of the memory controller for DDR4. Despite not seeing this, people are calling foul. I may not like the new socket, but crying foul is not a reasonable response.
I hear what you are saying but basically it just sounds like you are agreeing with what I just said to begin with and for what its worth I am well aware that memory controllers have now moved to the CPU. Keep in mind I already said it would increase the cost and the complexity to support both memory types. You basically cited the same things but only added possible additional points of failure (which goes hand in hand with increased complexity and cost) and die size which again is a cost issue . HEDT processors often have a larger die so its something to expect going in.

Still I'm not the one complaining about backward compatibility and I don't mind buying a new CPU+Motherboard (and in this case DDR4 RAM). Nor am I the one complaining about the likely price (probably similar to X79 / LGA2011 pricing on CPUs and board pricing).

I was only pointing out to those that were complaining about the lack of backward compatibility that what they were asking for could easily increase the price of the hardware in question significantly. In addition, they should also keep in mind that an Intel HEDT platform would be expensive to begin with without the unnecessary backward compatibility so requesting it is just adding more expense to an already expensive platform.

And lastly (and this is the real kicker) some people may be asking for backward compatibility for a previous platform (LGA2011) that they may not even own for a future platform (LGA2011-3) that they may never own. I'm not saying anyone here did it but I have seen it happen before in other threads.

As a Core i7 3930K / Asus P9X79 Deluxe LGA2011 platform owner I'm OK, with having to buy a new CPU and motherboard if I buy into the new upcoming Haswell-E LGA2011-3 platform. I think ill probably just stick with my current Sandy Bridge-E platform but when Haswell-E is released I just know I'll feel the itch.
Posted on Reply
#38
Aquinus
Resident Wat-man
OctaveanI hear what you are saying but basically it just sounds like you are agreeing with what I just said to begin with and for what its worth I am well aware that memory controllers have now moved to the CPU. Keep in mind I already said it would increase the cost and the complexity to support both memory types. You basically cited the same things but only added possible additional points of failure (which goes hand in hand with increased complexity and cost) and die size which again is a cost issue . HEDT processors often have a larger die so its something to expect going in.

Still I'm not the one complaining about backward compatibility and I don't mind buying a new CPU+Motherboard (and in this case DDR4 RAM). Nor am I the one complaining about the likely price (probably similar to X79 / LGA2011 pricing on CPUs and board pricing).

I was only pointing out to those that were complaining about the lack of backward compatibility that what they were asking for could easily increase the price of the hardware in question significantly. In addition, they should also keep in mind that an Intel HEDT platform would be expensive to begin with without the unnecessary backward compatibility so requesting it is just adding more expense to an already expensive platform.

And lastly (and this is the real kicker) some people may be asking for backward compatibility for a previous platform (LGA2011) that they may not even own for a future platform (LGA2011-3) that they may never own. I'm not saying anyone here did it but I have seen it happen before in other threads.

As a Core i7 3930K / Asus P9X79 Deluxe LGA2011 platform owner I'm OK, with having to buy a new CPU and motherboard if I buy into the new upcoming Haswell-E LGA2011-3 platform. I think ill probably just stick with my current Sandy Bridge-E platform but when Haswell-E is released I just know I'll feel the itch.
I get what you're saying here, considering I'm in the same boat, just with a 3820 instead. With DDR4 being so new and the only perk for 2011-3 being a new an improved PCH, I'm reluctant to think it's a worth while upgrade. Consider the improvements in performance over the last 5 years. In retrospect, comparing it to the past, it's been pretty slow with respect to IPC.

Remember when DDR3 first came out? It was hardly faster than DDR2 initially and I don't think we'll see amazing numbers out of it until they figure out how to improve their first version of DDR4 IMCs and in my own estimation, the PCH is not enough to make me think jumping ship is worth while, even if I had the money to spend and were seeking an upgrade.
Posted on Reply
#39
Assimilator
I still can't get my head around why Intel went with "2011-3" instead of going with a pinout/name that wouldn't confuse people (say, 2012?). Because I guarantee you that there will be forum posts involving n00bs who tried to force a 2011-3 CPU into a 2011 board or vice versa... maybe it's Intel's way of making money from stupid people? :p
Posted on Reply
#40
Octavean
Some time ago I had my old HP MediaSmart EX490 Windows Home Server go down. When that happened I quickly repurposed my Core i5 2500K / Asus P8P67 Pro based system into a new server running Windows Server 2012 Essentials. In effect losing a system that had purpose to me. I was able to restore my old HP server since then,.....twice. First it was a power supply failure which I replaced (had to modify the new PSU to HP's proprietary spec) and the second time it was the server's OS drive that failed (didn't lose any data thankfully). I prefer and trust the new Windows 2012 Essentials server though so I kept both servers running.

So I have been down a system for some time now. I've been able to deal without it though. Still I had tentative plans to build a new system about the time Haswell-E will likely launch. That wouldn't be a replacement to my Core i7 3930K / Asus P9X79 Deluxe based system but rather an addition.

Still, since I've gone this long without it,....I could go even longer. I guess that's why I've been on the fence a bit and have gone back and forth,......
Posted on Reply
#41
Aquinus
Resident Wat-man
AssimilatorI still can't get my head around why Intel went with "2011-3" instead of going with a pinout/name that wouldn't confuse people (say, 2012?). Because I guarantee you that there will be forum posts involving n00bs who tried to force a 2011-3 CPU into a 2011 board or vice versa... maybe it's Intel's way of making money from stupid people? :p
Stupid people shouldn't be using (or need for that matter,) a HEDT platform anyways. :p
I remember the first infraction I got was for calling someone an idiot for running a skt2011 CPU without a cooler because he thought it didn't need one because it didn't come with one. You get what you deserve I suppose. I wonder if Intel's overclocking plan covers stupid as well. :laugh:
Posted on Reply
#42
Octavean
AssimilatorI still can't get my head around why Intel went with "2011-3" instead of going with a pinout/name that wouldn't confuse people (say, 2012?). Because I guarantee you that there will be forum posts involving n00bs who tried to force a 2011-3 CPU into a 2011 board or vice versa... maybe it's Intel's way of making money from stupid people? :p
I'm guessing you are kidding right :)

If not I think that outcome is unlikely,..... for the most part.

The X79 /C600 / LGA2011 platform was far less popular and far less pervasive in no small way due to its added cost and the same will likely be true of Haswell-E / X99 / LGA2011-3

If I were guessing, I would expect Haswell-E /X99 / LGA2011-3 motherboards to start at about ~$300 to ~$350 USD when they are released if not more. Haswell-E processors will likely be released in the same performance / price trifecta of ~$1000 (EE / Extreme), ~$550 to ~$600 (K suffix) and ~$300 (entry level). History suggest the entry level processor may be delayed due to various reasons and likely will not have the same feature set as its more expensive alternatives. So for example, the entry level may be months delayed and have only 4 or 6 cores with Less comprehensive overclocking features. and so on,......making them less desirable in almost all respects other then price.

Add to that the extra cost of DDR4,....

Bottom-line is that most people will be unwilling or unable to buy into the more expensive X99 LGA2011-3 platform and instead opt for the cheaper yet still very capable Z97 or Z87 LGA1150 platform.

Still, "you cant fix stupid" so if someone is going jam the wrong CPU into the wrong socket then that will likely happen on any platform they get a hold of.

Also, it stands to reason that LGA2011-3 CPUs / Sockets will be keyed differently then LGA2011 CPUs / Sockets,...
Posted on Reply
#43
Ferrum Master
They should have named it socket E(expensive) :D

But seriously, basing on all experience starting from ancient edoram times... first chipsets with new generation of RAM always were drownable puppies... it will be a stillborn... large scale beta test.

On the other hand... I cannot still see any reason to upgrade from X79, if one(moneybag) needs bunch of sata3 drives there are always entry level pci-e cards from LSI or 3ware and etc for 200-300$. And it will perform like hell good.
Posted on Reply
#44
Octavean
Ferrum MasterBut seriously, basing on all experience starting from ancient edoram times... first chipsets with new generation of RAM always were drownable puppies... it will be a stillborn... large scale beta test.
I don't think I can agree with that,.....

X58 / LGA1366 and DDR3 was and still is IMO a fairly formidable combo. It was a great platform when it was first launched and still is today.
Posted on Reply
#45
Ferrum Master
OctaveanI don't think I can agree with that,.....

X58 / LGA1366 and DDR3 was and still is IMO a fairly formidable combo. It was a great platform when it was first launched and still is today.
It wasn't the first chipset utilizing DDR3... the case was... actually it put voltage restrictions thus rendering the first DDR3 generation useless.
Posted on Reply
#46
Hilux SSRG
OctaveanX58 / LGA1366 and DDR3 was and still is IMO a fairly formidable combo. It was a great platform when it was first launched and still is today.
I still use this configuration as my gaming rig, talk about a chip/chipset that has had a great shelf life.
Ferrum MasterIt wasn't the first chipset utilizing DDR3... the case was... actually it put voltage restrictions thus rendering the first DDR3 generation useless.
Agree to the voltage restrictions, but for the desktop/enthusiast market X58 was first with DDR3. The server/xeon market had DDR3-800/1066 a year prior if I remember correctly.
Posted on Reply
#47
Ferrum Master
Hilux SSRGI still use this configuration as my gaming rig, talk about a chip/chipset that has had a great shelf life.



Agree to the voltage restrictions, but for the desktop/enthusiast market X58 was first with DDR3. The server/xeon market had DDR3-800/1066 a year prior if I remember correctly.
I also had X58...

First one really was the BAD AXE 975X, but in terms of DDR3 it was the X38...
Posted on Reply
#48
Hilux SSRG
Ferrum MasterI also had X58...

First one really was the BAD AXE 975X, but in terms of DDR3 it was the X38...
I stand corrected, the X38 did have it - although I skipped that chipset myself.

But I'm really looking forward to Skylake.
Posted on Reply
#49
Octavean
Hilux SSRGI stand corrected, the X38 did have it - although I skipped that chipset myself.

But I'm really looking forward to Skylake.
Indeed I stand corrected as well but it doesnt really matter because The X38 chipset wasn't IMO deserving of the term " drownable puppies or stillborn" either. If I still had my old X38 board I probably will still be using it in some capacity or another.
Posted on Reply
#50
Octavean
Haswell-E will effectively replace the recently released Ivy Bridge-E CPUs. This will be the first time that Intel will be providing an 8-core CPU in their desktop lineup. Intel will be offering 6-8 core CPUs with their Haswell-E lineup that has a massive 20 MB of L3 smart cache and the same integrated voltage regulator as Haswell. This means the flagship Haswell-E chip will ship with a TDP of around 130-140W which is about 10-20W under the i7-3970X which only has 6 cores. Intel is shooting for a 55% IPC improvement over quad cores with Haswell-E.
www.thinkcomputers.org/intel-demos-haswell-e-cpu-with-ddr4-memory-on-x99-chipset-based-motherboard/

6 to 8 cores suggests to me that the entry level 5000 series Haswell-E processor may have 6 cores. That's better then Sandy Bridge-E and Ivy Bridge-E entry level processors at 4 cores. Also if the pricing for entry level is the same as in the past then ~$300 would be a great price for a 6 core i7 processor.

55% IPC improvement over quad cores with Haswell-E,.......that sounds lofty.
Posted on Reply
Add your own comment
Apr 25th, 2024 22:14 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts