Jump to content

Unconfirmed Rumor: No sub $1,000 8-core Haswell-E in 2014


Recommended Posts

Is the waiting over? Perhaps is a bit early to say with confidence, but there is a mixed bag of emotions should the leak that was published in the coolaler.com forum is proven true: no $600 octa-core in 2014, but...its not all bad news I guess!

 

s2011-3_Haswell_E.jpg

 

On the bright side, there is a chance that the 5820K will be pretty aggressively priced, bringing the wish for a more affordable, overclockable hex-core within reach for many more people. I doubt it will be as cheap as the 3820 and 4820K models before it as the 5820K will probably be better than the s1150 mainstream i7s, something the base models for s2011 never were.

 

The 5930K remains a hex-core, which may deter many current s2011 users upgrading, along with the fact that the new s-2011-3 based Haswell-E CPUs will require a new motherboard and of course DDR4 memory that might be overpriced for the first few months after launch.

 

For those wanting the "best" and willing to pay dearly for it tho, Intel did deliver a desktop octa core part under the 5960X code name. With all the Haswell-Es rated at 140W TDP, we know that this baby will probably be harder to push to really high clocks , but it should be easily clockable to better performance than the current Xeon E5-2687W - the fastest single IB-E Octa-Core - for less than half the price.

Link to comment
Share on other sites

Ah nice, thanks for info.

 

Well, to me it is quite underwhelming, you know, there is nothing "high performance/good value" like original 3930k was when it came out. 4930 was underwhelming, almost zero update, but doing that second time in row with 5930k is just...

 

On other hand, feels good how computers keep their value now :- ) Even what, 5 years old i7 2600k is still hell good unit.

Edited by RyderSK
Link to comment
Share on other sites

Thanks for the info. I feared that this will happen... comparable to the introdution of the 980X which was the first amd exclusive desktop hexacore until the 970 came maybe half a year later...

I also thought that the jump from a 1700€ 8-core Xeon to an unlocked 500-600€ 8-core i7 would have been to much for intel. But the clockrate of 3GHz is quite disappointing - even the Xeon has 3.3GHz. Hopefully they will clock well...

And yes, almost no improvement in the 500€ class.

But nevertheless... an unlocked 8-core for $1000 is still nice - like the 980X in 2010.

Link to comment
Share on other sites

But, whats the difference beetween 5930 and 5820?

only 28 instead of 40 PCIe lanes and it is clocked a bit lower - so maybe lower binned chips... but maybe not. Otherwise the 5820K would be the better option for a render workstation / node - if it overclocks equally.

Link to comment
Share on other sites

there is nothing "high performance/good value" like original 3930k was when it came out. 4930 was underwhelming, almost zero update, but doing that second time in row with 5930k is just...

 

On other hand, feels good how computers keep their value now :- ) Even what, 5 years old i7 2600k is still hell good unit.

 

Yeah, I am very grateful I invested in a 3930k (OC'd to 4.3GHz), because it's like chasing your tail trying to stay up-to-date with hardware, and until an 8core at 5GHz is available at a realistic price it would seem a waste of a good chip to upgrade to me. So for now I can breath easy, and concentrate on working.

Link to comment
Share on other sites

Great find, Dimitris.

I wish Intel would produce small number of 12-16 core (+ hyper threading of course) i7 based CPUs with no sacrifice to core clock. Even if that required increased CPU size, TDP, special motherboards and through-the-roof price (I am thinking 2000-4000 USD) I would still get one for the main workstation.

 

I know, I know- then who will buy their dual and quad Xeon based systems...

NVidia is sort of doing it with the GTX Titan Z- they are way too expensive to compete with AMD's best and cost 2 times more.

 

I wonder if one day when most CPU improvements are impossible due to technology limitations Intel will allow us to use multiple non Xeon CPUs in a single system.

Come on Intel- make that niche dream come true.

Link to comment
Share on other sites

My CPU purchases have been the Q6600 to the i7 2600k then to the 3930. Looking back, those were all milestone chips, which makes me happy..

 

CPU upgrade path exactly the same as mine, we made good choices.

My first Q6600 setup last for 6 years, 2006-12.

And thinking about 3930K, I do not see any worth upgrade for few years.

Link to comment
Share on other sites

I wish Intel would produce small number of 12-16 core (+ hyper threading of course) i7 based CPUs with no sacrifice to core clock. Even if that required increased CPU size, TDP, special motherboards and through-the-roof price (I am thinking 2000-4000 USD) I would still get one for the main workstation.

 

HT doubles the threads, but not the performance. In most cases the latter gets a 20-30% increase.

 

55319.png

 

At the same time, doubling the threads gives a big penalty in RAM usage...a 1P 8C/16T i7 with 32GB or RAM - with room to 64 or 128GB even - probably won't have an issue in most cases, but 2P systems with 10C/20T (total of 40 or more threads) and really complex scenes might be pushing RAM requirements towards really expensive RDRAM options and sticks bigger than 8GB (the current ceiling for unregistered DDR3 - don't know about DDR4).

 

As far as base clocks on multi-core CPUs, remember that intel and AMD alike need to cater with more or less one design, both workstation and datacenter/server markets, which have concluded that 130-150W is the "upper comfort" zone for aircooled CPUs in 1U/2U racks. Massive aircoolers, CLC or open loop water coolers etc, can push this envelope up quite a bit, but experimenting is not what an enterprise environment needs: they need 100% of the promised specs as easily upgradeable and as compact as possible + intel wants to monetize on their R&D for as long as possible thus most Xeons don't even offer Overclocking options.

 

We should be happy that we can at least squeeze so much performance these days with the current i7 lines.

Link to comment
Share on other sites

  • 5 weeks later...
  • 3 weeks later...
  • 2 weeks later...

currently using 3930k@4.29Ghz, thinking is it upgrade whole system platform (CPU/Mainboard/RAM) to 5860x(8 core 3.0Ghz) or just upgrade to E5- 2690v2(10 core 3.0Ghz) better? since x79 mainboard can support e5 processor, 2690v2 around 2000usd, 5960x + x99 board + ddr4(let say 32gb) should be more than 2000usd. any idea?

Link to comment
Share on other sites

  • 3 weeks later...
That's some crazy temperatures. Don't see it worthy to run 5960X at more than 3.5-3.6 GHz. 100C with H100i at 4.5Ghz :- D ? lol.

 

Can't wait. Now 880.

 

You just cannot do it with a H100...you cannot really run many current s2011 6-core CPUs @ 4.5 without breaking 85oC torture testing, so pumping a 8-core with it won't really do better.

 

Same 3930K that does 85oC @ 4.6GHz under a H220, and similarly @ 4.5 for a SilverArrow. H100 and pretty much any 240mm CLC is not that better than a decent air cooler (only if it gets much much louder).

 

 

If its a proper overclocker, the 5820K will be easily the best value out of the 3...don't see any reason to get the 5930K instead. The 40 PCIe lanes over the 28 PCIe lanes is a joke of an upgrade: the 4770K often beats the 4930K with 16 PCIe lanes vs. 40, even in 3-SLI configs.

Almost no single GPU can really benefit having access to more than 8 PCIe 3.0 lanes - CPU+GPU(s) simply cannot saturate this kind of bandwidth long enough to make a difference.

28 PCIe 3.0 lanes leave a decent headroom for PCIe based SSDs or w/e.

Link to comment
Share on other sites

What's up with everyone being "shocked" (article words) by the 28 lanes limitation on 5820K ? It's not like someone will buy the 300 euro CPU and stick more tripple 780Ti there.

Otherwise nice line-up, and many interesting motherboards. Can't wait to start building.

 

The prosumer market is the important one for the consumer grade silicon. For small businesses that rely on CPU limited throughput, such as video editing, video production, scientific computation and virtualization, having the high performance in a single, low-cost product can produce a significant upgrade in throughput, allowing projects to be completed quicker or with more accuracy. While these prosumer would love the higher powered Xeons, the cost is overly prohibitive, particularly in the long term, or the lack of memory overclock capability has a negative effect.

 

Nice article though, I am glad this small tidbit got there. Sometimes reading these websites is like watching Idiocracy movie, with 99perc. of users convinced the full technology cycle is driven by computer games now.

Xeon=But do you run Battlefield faster on 20 cores ? Tesla=Can't run Crysis, sucks, why would you pay 3000 ? Top NEC monitor=Huh ?No nVidia G-Sync ? Shit.

Edited by RyderSK
Link to comment
Share on other sites

What's up with everyone being "shocked" (article words) by the 28 lanes limitation on 5820K ?

I don't know... but the 5820K is really a nice render node cpu.

 

Can't wait to start building.

Me too... but i have to wait til october - before i disassemble my workstation i have to finish some projects ;)

Link to comment
Share on other sites

Think I will stick with my 3930K for as long as DDR4 holds that kind of pricing :-/

 

But...tempted.

 

As for 5820K, yes, its best bang / value. Ppl complaining about lanes are ignorant.

s1155/1150 I7s with even less lanes were doing better with GPUs in most scenarios already - even with triple / quad SLI (irrelevant to ArchVIZ and GPGPU rendering altogether), given a PLX chipped motherboard the quads would do better or at least as good in gaming.

5820 has MORE lanes than those i7s.

 

GPGPU is also not limited by PCIe bandwidth whatsoever.

Those that have a vague idea of coin mining with GPUs - the most widely used GPGPU application in the consumer market so far - should be happy to know that even PCIe 3.0 1x can feed the GPUs with enough data to work @ 100% load/efficiency without issues.

 

So are ppl going for super-speed DDR4 btw, as in 1st benchmarks the faster clocked 4C s1150 + "slow" Dual channel DDR3 still beats stock or lower OCed (still amazingly fast) 5960X + quad channel DDR4 in single threaded, proving that Haswell was never "better" than what DDR3 could do, and so is Haswell-E.

 

RAM speed is not the bottleneck, like, EVER for quite some time.

 

Save money for other components, or just save money. Don't go for super expensive DDR4, dominator fubar etc.

Edited by dtolios
Link to comment
Share on other sites

  • 2 weeks later...

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share

×
×
  • Create New...