I know that, but the idea is that Intel sees two markets for these high clocked hexcores:
One with 40 PCIe lanes and one with 28 PCIe lanes.
My guess is that the next generation of lower PCIe lane hexcores comes from a Xeon-D type die.
Those 2 hexcores are also clocked higher than any octocore. The elephant in the room.
So..
This time it isn't a disabled part but a just a Server Xeon with Server stuff disabled. Doesn't wipe away what they did in the past.
http://ark.intel.com/products/82765/Intel-Xeon-Processor-E5-1650-v3-15M-Cache-3_50-GHz
Matches this
http://ark.intel.com/products/81900/Intel-Xeon-Processor-E5-2643-v3-20M-Cache-3_40-GHz
Don't get me wrong, I don't have some anti-Intel agenda here. Just stating that HEDT is still a profitable market for them.
I know that, but the idea is that Intel sees two markets for these high clocked hexcores:
One with 40 PCIe lanes and one with 28 PCIe lanes.
My guess is that the next generation of lower PCIe lane hexcores comes from a Xeon-D type die.
You link 2 hexcores. Not sure what purpose.
While the cache and memory system is the main benefit of the LGA2011 chips. They quickly come short when its not being utilized to its full potential.
This is one of the things people that just demand more cores fail to understand.
Those hexa cores were clocked equal or higher than the 5930k which was your elephant in the room.
As for those Benchs, if a software isn't capable of taking advantage of the extra cores, than at worst a higher core CPU would perform equal to a lower core one, all things being equal.
But all things are not equal as
1) Intel's higher core CPUs are clocked lower than the mainstream part in this case against the 4790K
2) Haswell-e is DDR4 based which has significantly worse memory latency right now than mainstream Haswell. Hence the lower bench score.
Finally, Haswell-E might be based on non defective dies. But they did need to provide compelling reason for users to upgrade from their old CPUs & buy new stuff as the IPC hasn't moved greatly since SandyBridge times.
The compelling reason being offered here, moar cores....
Might be gimmicky, but as long as people keep buying new products, its a success for them.
I have seen most 5960Xs do around 4.3-4.5Ghz. That's where majority of original Haswell users were before DC. Higher cores doesn't always mean lower clocks.I talk about hex vs octocores. The fact is more cores=lower clocks. And then you sit in problematic cases where a lower core but higher clocked outperform the higher core chip. Thats the elephant in the room people act like doesnt exist. Same reason why chips like the 8891v3 exist. Quadcore with 45MB cache and quadchannel because it gives the best result. Not 6, 10, 14 or 18 cores. No, 4!
And in case of IPC, again, here is history:
I have seen most 5960Xs do around 4.3-4.5Ghz. That's where majority of original Haswell users were before DC. Higher cores doesn't always mean lower clocks.
I know Haswell is approximate ~10% faster than Sandy. But that won't get all gamers to buy new chips, however moar cores does.
Well that's a problem that's hard to fix & some companies would like to have.If more cores does, then they are just ignorant gamers.
Gamers should actually ask for more cache instead.
Not sure why you felt so much in love with the Xeon-D.
For a hexcore user I'll bet the TDP ends up lower compared to hexcore on HEDT. This assuming clocks are the same and cores have the same density on both platforms.
The fact it is a SoC also makes it more efficient for small motherboards.
To reduce TDP you also have to reduce the chip
Well that's a problem that's hard to fix & some companies would like to have.
Like how AMD branded FX 8350 as 8 Cores, so better than Intel. Intel isn't except either, but you get the picture.
Edit:- Also better coders would be a welcome addition. Just saying..
And a hexcore that you gonna try and sell to who again?
Xeon-D octocore is already reduced compared to HEDT octocore. (ie, less memory channels, less PCIe lanes, less cache, etc).
People in this forum want hexcores ---> http://forums.anandtech.com/showthread.php?t=2424749
Well its obvious that people aint that stupid. Because chips like the FX is just a complete flop.
The coding issue isnt going away anytime soon. And it may never do.
I do wonder however why gamers dont ask themselves what they really need and then put a pressure on Intel in that regard. It would be interesting to take a regular LGA1150 quad, 5960X and a Xeon with 45MB cache and then isolate the differences with same core amount and clocks. Then I think gamers would wake up and realize they want more cache on their CPUs. Back in the Core 2 days people was smart enough to understand and test this.
Jee.DDR4 or stacked memory isnt solving the cache issue. Stacked memory got the same slow latency btw as regular DDRx.
IMHO all you need is a high enough turbo on these cpus, regardless of core count. If you have 100+W TDPs and aren't loading more than 4 cores why it shouldn't be possible?
Also remember their 220W TDP FX that works at 4.7GHz base, we know that Zen will be >40% better for clock (it's over excavator the statement) so at 3.36GHz maximum you already have its single thread performance.
Implement turbo up to 4GHz and it will be within 20% of Skylake quad core but with 4 more cores and better single than a 5.5GHz Piledriver.
Jee.
You are correct.
HBM is about providing Higher Bandwidth, latency is still an issue.
More Cache it is then. Lol..
Happy medium between Cache & # of Cores needs to be found.
And what is the result again? A 45W 2Ghz base, 2.6Ghz turbo octocore at 14nm with 12MB cache on 14nm vs a 55W 1.8/2.9Ghz octocore at 22nm with 20MB cache and quadchannel memory and QPI links.
I brought that example up in the other thread:
http://forums.anandtech.com/showpost.php?p=37373887&postcount=942
I'm thinking Xeon-D's turbo is probably artificially limited for product segmentation reasons.
Well 6770K Skylake is still rumored to have 8MB Cache.I think we need to reach/pass 20MB cache before we even talk about more cores in a meaningful way for (average/gaming) endusers. The good part is that cache is the easist, most power efficient and cheapest thing to do of the 2.
Well 6770K Skylake is still rumored to have 8MB Cache.
Maybe Intel is waiting till AMD releases Zen or something competitive enough & then in response they just raise the Cache size on their next chip in addition to other things & yay more performance.
Thing is, they are smart enough to know that more Cache would help, so they are holding back on purpose.
Yee, no. Its TDP limited, simple and easy.