Intel Announces the Xeon E5-2699A at SC 2016

[Raevenlord]

At the SC16 (Super Computing 2016) conference, Intel introduced its new Intel Xeon E5-2699A v4, which is a newer, faster version of the existing Broadwell E5-2699 v4. The "newer, faster" bit basically amounts to an increased base (from 2.2GHz to 2.4GHz) and boost clock (from 3.4GHz to 3.6GHz). Like its predecessor, the Xeon E5-2699A features 22 cores and 55MB of L3 cache, and the company cited vague improvements to the now-mature 14nm process that which amount to a 4.8% gain in LINPACK performance - which can surely be attributed almost exclusively to its clockspeed increase (not unlike the expected performance differential between Intel's current Skylake and upcoming Kaby Lake architectures). The E5-2699A v4 carries an MSRP of $4,938, which marks an eye-watering 20% increase over the $4,115 of its 200-MHz slower predecessor.



View at TechPowerUp Main Site

 

[RejZoR]

4.8%. WOW!

[SARCASM INTENSIFIES]

 

[Raevenlord]

4.8%. WOW!

[SARCASM INTENSIFIES]

Such performance, very gains.

 

[alucasa]

I'd kill an ant worker for one.

It should be 100% faster than mine.

 

[aldo5]

Intel: "it is been 3 years and 1 generation since 2699 v3 and we finally can present you our results.... price up +20% performance +4.8%"

 

[cdawall]

4.8%. WOW!

[SARCASM INTENSIFIES]

4.8% is a massive increase in the industry that will buy this. Remember yet again these are not consumer products. Price is merely a number to these people, it doesn't affect purchasing.

 

[RejZoR]

It costs 20% more and delivers 4.8% more performance. Much enhanced, very improved.

 

[cdawall]

It costs 20% more and delivers 4.8% more performance. Much enhanced, very improved.

This is something you will see in the oil industry. They don't care about the price. 4.8% faster on a device that can make millions per hour is an acceptable purchase to them. Break that price down it will pay for itself.

 

[Frick]

This is something you will see in the oil industry. They don't care about the price. 4.8% faster on a device that can make millions per hour is an acceptable purchase to them. Break that price down it will pay for itself.

What device using Xeons makes millions per hour? I want to know so I can get me one.

 

[Kursah]

What device using Xeons makes millions per hour? I want to know so I can get me one.

I would assume server farms which specialize in data mining and storage would fit nicely in that market for pretty much any major corporation involved in anything that has user memberships, accounts, tracking, access, etc.

Data metrics on you and I = $$$$

Doesn't matter if its Amazon, Costco, Microsoft, Cisco, Ubuntu, Steam, EA, they'll gladly spend for infrastructure that is powerful enough to handle and process millions of data metrics, store them and be able to maintain databases, which are resource hogs. Not that I care much for this part of the industry, it is there and is only growing more and more and the more data that vendors can acquire and manipulate for reports and storage sooner, the more money they can make faster. Big data = big $$$.

 

[cdawall]

What device using Xeons makes millions per hour? I want to know so I can get me one.

Geographing for oil/gas...not to mention data mining etc

 

[theeldest]

Intel: "it is been 3 years and 1 generation since 2699 v3 and we finally can present you our results.... price up +20% performance +4.8%"

It's actually 4.8% faster than the current generation. The chips being compared: 2699 V4 and 2699A V4. Their yields improved enough to get a couple more bins so they're offering better clocked parts.

 

[theeldest]

This is something you will see in the oil industry. They don't care about the price. 4.8% faster on a device that can make millions per hour is an acceptable purchase to them. Break that price down it will pay for itself.

Also, for many applications the software licensing is the most expensive part. You can pay $20k - $40k per server for the hardware but spend $1m per server for licensing. 5% increase in efficiency is pretty significant.

 

[cdawall]

Also, for many applications the software licensing is the most expensive part. You can pay $20k - $40k per server for the hardware but spend $1m per server for licensing. 5% increase in efficiency is pretty significant.

That's what a lot of people apparently miss. 5% is huge in the market these are targeted at.

 

[Frick]

Geographing for oil/gas...not to mention data mining etc

So it's not as much as device as a farm of devices with hundreds if not thousands of these things.

 

[cdawall]

So it's not as much as device as a farm of devices with hundreds if not thousands of these things.

And what is the increase on 5% across 500?

 

[Aquinus]

And what is the increase on 5% across 500?

5%. Seriously, that's like having 25 more cores. If you're talking about serving content and distributed data processing, that 5% sounds small but, it's being applied to big numbers. From a strictly computational standpoint, 5% faster could mean 5% fewer servers. There can be pretty big cost savings from that perspective but, once again. We're talking about businesses where this matters. This matters for companies like Amazon, Google, and Rackspace... companies that provide cloud services.

 

[cdawall]

5%. Seriously, that's like having 25 more cores. If you're talking about serving content and distributed data processing, that 5% sounds small but, it's being applied to big numbers. From a strictly computational standpoint, 5% faster could mean 5% fewer servers. There can be pretty big cost savings from that perspective but, once again. We're talking about businesses where this matters. This matters for companies like Amazon, Google, and Rackspace... companies that provide cloud services.

And not talking about something simple minded people don't understand.

 

[ZeDestructor]

Intel: "it is been 3 years and 1 generation since 2699 v3 and we finally can present you our results.... price up +20% performance +4.8%"

More like a few months and mild tinker - this is the 2699A v4, that sits on top of the still current 2699 v4

 

[dalekdukesboy]

I know it's got amazing multi-thread potential but for most programs it's a useless cpu. Also, can anyone tell me for sure which xeon's are locked/unlocked? Seems when you throw in the ES versions it gets even more dicey...seems the 1600 series of x79 is unlocked and beyond that and into x99 I've seen people even email Intel and they never get a straight answer.

 

[cdawall]

All 2xxx series xeons are bus and multi locked

 

[INSTG8R]

So many cores!....

 

[dalekdukesboy]

All 2xxx series xeons are bus and multi locked

That was generally what I thought and some 1xxx series were unlocked...however I've heard of some 2xxx series that definitely were unlocked in x79 I believe the ES 2687 is at least by multiple sources I read....like this 2687 ES

http://www.ebay.com/itm/Intel-Xeon-...0x-/162100562943?_trksid=p2141725.m3641.l6368

Me and a couple members researched and it seems that CPU is unlocked according to a few sources and screenies etc but...hate the mystery it appears Intel seems to intentionally keep around their cpus multipliers. I believe when you look on their own spec sheets it doesn't even say unless I missed it on any of their xeons what multiplier range or if it's locked or not, they simply don't tell you.

 

[cdawall]

Maybe the older 2011 stuff. I can tell you my current ES broadwell-e is locked as are the rest of the 2xxx chips on 2011v3

 

[dalekdukesboy]

Blah. Sorry lol, I appreciate your input it's really intel and their intentional lack of transparency on this issue that bugs me. If AMD was competitive at every level including business CPU's I'd bet not only would their pricing go down a notch or two but we've have better innovation since 2011/x79 and more transparency on features like multiplier range etc. In a way I can't blame them at all, why compete with yourself? Why overstretch and risk bad yields or glitches in new architecture when you're competition is way behind while you stroll along comfortably with good profit on safe and relatively small jumps with innovation. That said, shame on you AMD get in the race, but also I think it's fairly criminal to not even list the multiplier range on CPU's...unless I missed it, I went to Intel's site many times checking out the cpu specs and they don't even list the multiplier range, ONLY turbo speed, cache, cpu speed at stock etc.