Today, as Intel (NASDAQ: INTC) launches the third generation of its Xeon E5 dual-CPU platform, many eyes are on the improvements it brings to the servers in the datacenter. However, the benefits are just as high – if not higher – on the high-end workstation front.

First of all, Haswell core means sped-up AVX floating point, by inclusion of fused multiply-add (FMA) ops for theoretical FP rate doubling in benchmarks like Linpack, for instance. Haswell’s AVX2 also, just as importantly, moves integer processing to the wide parallel AVX engines, essentially offloading anything aside the address calculations to the RISC-like, three-address AVX instruction format and wide register sets. For workstation apps, once re-compiled to take advantage of it, the benefits could be enormous, and also be another gradual move away from the antiquated X86 code base.

Then, the wide choice of a number of cores per SKU – from 8 all the way to 18 – enables you to pick the right balance of per-core speed (i.e. per-thread performance) and core number, depending on the parallelism of your application. Some apps scale less well across many cores, thus preferring high per-core speed, while others like ray tracing make the most out of many cores.

The initial workstation SKU in the Xeon-E5 v3 range, the E5 2687W v3 flavor, is a 3.1 GHz 10-core part that actually uses the 12-core die where 2 cores (and their associated caches) were turned off. Now, its predecessor, the 2687Wv2 on the Ivy Bridge platform, had full L3 caches even if some cores of the die were disabled, a benefit that, I guess, we will only see back in Broadwell-EP (E5 v4) SKUs next year.

Then we come to DDR4 – yes the initial DIMMs aren’t exactly speedy, especially latency-wise, but the lower voltage and other reliability features of DDR4, together with quick improvements in speed and latency expected over the next few quarters, should provide the users the never-before seen capacity on a dual-socket workstation, beyond 1.5 TB RAM, without sacrificing the bandwidth on high load situations like DDR3.

The improvements in the PCIe bandwidth, integrated voltage regulation, and sped-up QPI to 9.6 GT/s also round out the key extra benefits.

Putting it through its paces

Here we look at the initial reference workstation based on this SKU from Intel, packaged by BOXX. The machine itself is compact, using liquid cooling on a SuperMicro X10DAi workstation mainboard with three PCIe x16 v3 slots. This doesn’t max out the platforms theoretical quad-GPU full bandwidth capability, but should be enough for most users. In return, the board has space for 16 DDR4 DIMMs, i.e. a full terabyte of RAM if using 64 GB modules available early next year. The installed RAM was 128 GB, in 8 pcs of Samsung 16 GB ECC DDR4-2133 RDIMMs.

The system came with a Nvidia Quadro K2000, which I changed to AMD FirePro W9100, arguably the most powerful professional OpenGL card available as of today. With 16 GB VRAM and six DisplayPort outputs, the card is able to drive even 8K displays like the one from BOE Technology that we mentioned last week. Intel’s 240 GB + 400 GB (SATA + PCIe) SSD combo completed the picture.

The first benchmark was the brand new SPECwpc all-encompassing workstation productivity benchmark by SPEC, on this system. The suite, which takes a couple hours to run, covers everything from processor to graphics (a.k.a ViewPerf) to overall system performance, and seems to do the job with much less trouble than, for instance, BAPcO SysMark did years ago on the PCs.

Here are the first SPECwpc results, on the dual 3.1 GHz E5-2687W v3 system:

Next, we ran CineBench 15 – note that the system is about twice as fast as an overclocked 4+ GHz Core i7-5960X, the desktop Haswell-E brethren to these Xeons.

In CPU-Z, you can see the data about the CPU.

Then we come to the newest version of SiSoft Sandra. Here is the report on the key performance data.

In our next round, we will be focusing on the changes in performance obtained when changing – and tuning – the main memory, as well as looking at the opportunity for even higher CPU speed. In my own opinion, the workstation market can easily justify higher TDP – and maybe even unlocked – Xeons, especially in both 8 core and 18 core per socket configurations.

The post Haswell-EP Workstation Preview: Xeon E5 v3 Rocks, But Still More To Go appeared first on VR World.

Today, as Intel (NASDAQ: INTC) launches the third generation of its Xeon E5 dual-CPU platform, many eyes are on the improvements it brings to the servers in the datacenter. However, the benefits are just as high – if not higher – on the high-end workstation front.

First of all, Haswell core means sped-up AVX floating point, by inclusion of fused multiply-add (FMA) ops for theoretical FP rate doubling in benchmarks like Linpack, for instance. Haswell’s AVX2 also, just as importantly, moves integer processing to the wide parallel AVX engines, essentially offloading anything aside the address calculations to the RISC-like, three-address AVX instruction format and wide register sets. For workstation apps, once re-compiled to take advantage of it, the benefits could be enormous, and also be another gradual move away from the antiquated X86 code base.

Then, the wide choice of a number of cores per SKU – from 8 all the way to 18 – enables you to pick the right balance of per-core speed (i.e. per-thread performance) and core number, depending on the parallelism of your application. Some apps scale less well across many cores, thus preferring high per-core speed, while others like ray tracing make the most out of many cores.

The initial workstation SKU in the Xeon-E5 v3 range, the E5 2687W v3 flavor, is a 3.1 GHz 10-core part that actually uses the 12-core die where 2 cores (and their associated caches) were turned off. Now, its predecessor, the 2687Wv2 on the Ivy Bridge platform, had full L3 caches even if some cores of the die were disabled, a benefit that, I guess, we will only see back in Broadwell-EP (E5 v4) SKUs next year.

Then we come to DDR4 – yes the initial DIMMs aren’t exactly speedy, especially latency-wise, but the lower voltage and other reliability features of DDR4, together with quick improvements in speed and latency expected over the next few quarters, should provide the users the never-before seen capacity on a dual-socket workstation, beyond 1.5 TB RAM, without sacrificing the bandwidth on high load situations like DDR3.

The improvements in the PCIe bandwidth, integrated voltage regulation, and sped-up QPI to 9.6 GT/s also round out the key extra benefits.

Putting it through its paces

Here we look at the initial reference workstation based on this SKU from Intel, packaged by BOXX. The machine itself is compact, using liquid cooling on a SuperMicro X10DAi workstation mainboard with three PCIe x16 v3 slots. This doesn’t max out the platforms theoretical quad-GPU full bandwidth capability, but should be enough for most users. In return, the board has space for 16 DDR4 DIMMs, i.e. a full terabyte of RAM if using 64 GB modules available early next year. The installed RAM was 128 GB, in 8 pcs of Samsung 16 GB ECC DDR4-2133 RDIMMs.

The system came with a Nvidia Quadro K2000, which I changed to AMD FirePro W9100, arguably the most powerful professional OpenGL card available as of today. With 16 GB VRAM and six DisplayPort outputs, the card is able to drive even 8K displays like the one from BOE Technology that we mentioned last week. Intel’s 240 GB + 400 GB (SATA + PCIe) SSD combo completed the picture.

The first benchmark was the brand new SPECwpc all-encompassing workstation productivity benchmark by SPEC, on this system. The suite, which takes a couple hours to run, covers everything from processor to graphics (a.k.a ViewPerf) to overall system performance, and seems to do the job with much less trouble than, for instance, BAPcO SysMark did years ago on the PCs.

Here are the first SPECwpc results, on the dual 3.1 GHz E5-2687W v3 system:

Next, we ran CineBench 15 – note that the system is about twice as fast as an overclocked 4+ GHz Core i7-5960X, the desktop Haswell-E brethren to these Xeons.

In CPU-Z, you can see the data about the CPU.

Then we come to the newest version of SiSoft Sandra. Here is the report on the key performance data.

In our next round, we will be focusing on the changes in performance obtained when changing – and tuning – the main memory, as well as looking at the opportunity for even higher CPU speed. In my own opinion, the workstation market can easily justify higher TDP – and maybe even unlocked – Xeons, especially in both 8 core and 18 core per socket configurations.

This post originally appeared on VR World.

The post Haswell-EP Workstation Preview: Xeon E5 v3 Rocks, But Still More To Go appeared first on VR World.

A day before Intel’s Developer Forum kicked off in San Francisco, Intel (NASDAQ: INTC) officially unveiled the Xeon E5-2600 v3 known previously by its code name of Grantley.

On stage, Intel’s Diane Bryant, the company’s general manager of its Data Center Group, said that the “big data” economy was driving demand for a new series of Xeon Chips.

“The digital services economy imposes new requirements on the data center, requirements for automated, dynamic and scalable service delivery,” Bryant said on stage. “Our new Intel processors deliver unmatched performance, energy efficiency and security, as well as provide visibility into the hardware resources required to enable software defined infrastructure.

“We are continuing to be the most energy efficient processor on the planet,” she continued,” promising 27 world records in performance across the product line.”

“[Intel is] excited about the role this platform is going to play as we all collectively re-architect the datacentre from static to dynamic and from proprietary to open standards,” she later said on stage.

While the previous generation of Intel Xeon chips had a core count that went as high as 12, the third generation of Xeons offers up to 18 cores with clock speeds of up to 3.7 GHz . The Xeon E5-2600 v3 will also be the first in the series to support DDR4 RAM, which offers substantial clock speeds above DDR3’s JEDEC speeds as well as power savings due to lower operating voltages.

Another new feature of the Xeon E5-2600 v3 is the update of Intel’s Quickpath Interconnect (QPI) that’s capable of transferring data at 9.6 GTps, much faster than existing front side bus (FSB) technology.

To aid in the inquiry of fault detection, Intel is including a new range of telemetry sensors to give administrators more information and metrics on CPU usage, memory and I/O.

The new Xeon is available in 22 different versions that vary wildly depending on required clock speed and cores. Pricing begins at $213 and tops out at $2,072. For the workstation, the chip is available as the Xeon E5-1600 with pricing ranging from $295 to $1,723.

The post Intel Officially Unveils Haswell-Powered Xeon at IDF 2014 appeared first on VR World.

As more and more X99 motherboards leak, it only made sense to release a portion of the trove of boards that haven’t already been announced or leaked.

The Asus X99-E WS, a Bright Side of News* exclusive, is a particularly interesting one because it is Asus’ (TPE:2357) latest high performance motherboard for workstation users. Because X99 is an entirely new platform there are very likely going to be a lot of people searching for new workstation boards like the X99-E WS to upgrade their workstations.

In terms of features, you get all the expected things like Haswell-E support, a plethora of PCIe 3.0 slots, which includes support for 4-way CrossFireX and SLI with X16 links. This board is specifically designed to support both Core i7 and Xeon E5 class chips (including the E5-1600, thanks to retailer leaks) thanks to being only a single socket workstation board. Because this is a workstation board, ASUS also talks about the boards overall power efficiency focusing on their Dr. MOS, 12k hour capacitors, ProCool power connector and beat thermal choke.

The Asus X99-E WS also has a Q-code logger and Dr. Power, even though it also has a plethora of connectivity which make it all the more attractive. That includes dual Gigabit NICs, a whopping 10 USB 3.0 connectors on the rear IO alone, not to mention the standard audio connectors, eSATA and the expected IO. You will get the full 8-slot DDR4 configuration, which is to be expected even though you very likely won’t get much memory overclocking support like the Rampage series of boards.

Asus X99-E WS IO

In addition to all of this, you will be getting backed by an Asus 3 year warranty which brings the MSRP price of this board to a whopping $519. Yes, this is a fairly high price for a single socketed board that isn’t an overclocking board, but X99 is a performance platform and this is a workstation board which should mean you are paying for higher quality components and reliability.

In fact, some retailers already have it available for pre-order but don’t really have any details or pictures of the board. So, you can already get it for $499 via some retailers, already lower than the $519 MSRP.

The post Asus X99-E WS Motherboard Leaked appeared first on VR World.

The Portland suburb of Hillsboro, where all Intel’s high end product operations – and its main cash cow – are located, was unusually hot for this time of the year, with temperatures almost touching 30 Celsius (90 Fahrenheit) some days.

So was Intel inside (pun intended), overheated in preparation for the imminent launch of new workstation, server and, yes, high-end desktop Haswell flavours that will have a public debut before the September IDF opens its doors. These were already well written about by many in the media community, so this time it’s pointless repeating what’s already widely known.

What is interesting is where Intel would go from here. Will the company focus on the Xeon and related enterprise and high end client products which do bring in the high-margins?  Or get embroiled deeper in the fight for the current fad of the day, the “all-popular but hard to make money” ultra-mobile gadgets?

The situation in the two markets can’t be more opposite: in the first, Intel’s Datacenter Group is an absolute industry leader, with the estimates of its market dominance hovering around, or above, 90% — of the highest-profit market in the general IT hardware space. After a bit of lull few years ago, the product launches are again on a yearly basis, keeping the tick-tock regular. Aside of increasingly hungry – perhaps vengeful – IBM with its global promotion of POWER8, there are no real global competitors in this space at the moment, performance-wise or presence-wise.

Intel as the underdog

On the other side, in the highest volume, but questionable margin, ultramobile space, with its plethora of smartphone and tablet offerings, Intel was, and still is, an underdog. Maybe it is in a worse position than AMD was versus Intel in the x86 space a decade ago, or that Alpha and MIPS were versus x86 fifteen years ago.

At least, during those respective times while trying hard to enter the main arena both of those competitors had their protected niche markets where they ruled — while it all lasted. In both cases, it was based on the combination of performance and feature advantages and customer base loyalty, at least for specific apps where Intel couldn’t match those competitors then.

Compare it to today’s ultramobile battlefield. Intel has sunk enormous resources, both financial and manhours, in getting into that almost totally ARM-dominated market. Over the past few years this seriously affected its balance sheet in the process. But Intel, like others, had its protected market: the high-end server side fund their low-end ultramobile peers. Yet, despite fairly good performance of its Atom-based mobile offering – in quite a few cases these measurably outperform their ARM competition – and huge investment in Android apps porting, the results are still only trickling in.

Lessons learned

Let’s go back in time to a period when Alpha and MIPS had even greater comparative performance advantage over the x86 in their respective heyday.

At the high end, that extra performance mattered much more than in a smartphone, whose primary functions should, after all, be calling and texting. But the companies behind them, while not small by any means, still couldn’t handle Intel marketing competition and lack of will by other partner vendors to fully support them. So, at least outside China, they failed.

Now, Intel faces a “central committee” of all-powerful global vendors like Samsung, Huawei, Nvidia, Apple, LG and, of course, Qualcomm, all working with the little ARM Plc, to push ARM forward.

Now ARM is hardly the best architecture around. In fact, if you really wanted to find something worse in performance, architecture and scaling than the x86, ARM and SPARC are the only real candidates, aside of the “good ship Itanic.” An architecture originally designed for a low-end desktop PC (see: BBC Micro) and embedded apps, never for high-performance computing, can in reality only stay within the ultramobile space unless major, major changes are made – which impact the now “golden” compatibility with the past apps.

After all, it took ARM nearly 30 years – from 1985 “Acorn RISC Machine” to 2014 Cortex-A57 – to have a proper 64-bit processor, while MIPS and Alpha were fully 64-bit in 1990 and 1991, respectively. Even the x86 has now over a decade of 64-bit existence.

And yes, those ARM alliance vendors fight each other like nobody’s business every day – they are each other’s worst enemy. However, Intel’s entry would unite them all against a “common enemy” who should not be allowed a chance at the dominance, at least not the way it has in the PC world.

Does Intel need an exit strategy?

Even with shareholder pressure of the “my daughter’s iPad doesn’t have Intel Inside: fix it or you’re fired!” sort, the question is how deep Intel should go into the smartphone and tablet quagmire?

Something like FullHD to UHD 2-in-1 running on Broadwell ULV does make sense, as it is essentially a PC Ultrabook with a Tablet mode or vice-versa. Windows is still more of a productivity platform than Android, so there would be a definite differentiation.

However, the mainstream ultramobile battlefield, with cut-throat prices for both SoC chips and the end products, may not be the best thing for Intel to enter. Perhaps a reasonable goal of creating and maintaining a 10% market presence in the smartphone and tablet field, not unlike that of Apple in the desktop and laptop space, would fit the best. It would be big enough to create a nice unique-value

niche and have most of the apps running native, but it would not be seen as a major threat to the ARM side, and other things would basically continue as usual.

However, on the high end, where those same ARM vendors are drooling after Intel’s high margin, four digit priced chippery, Intel has to stay resolute and, by accelerating the product launches and keeping the huge performance delta, show to those vendors that it will take forever and a day for them to catch up. Broadwell EP should not be delayed from the yearly refresh cycle, and neither should its Skylake follow-on. The profitable enterprise SSD, networking and interconnect programs are there as well, and they should move forward at the same rapid pace.

If there’s a way to justify even higher per-socket chip prices for even more powerful CPUs for even denser datacenters – where power and space are a constraint – then maybe there is a fresh way forward.

How about looking back at those previous non-x86 RISC architectures that still leave ARM in the dust as a way forward for Intel, while using the existing socket and chip infrastructure? After all, x86 being x86, there seems to be some sort of practical ceiling – somewhere around $5,000 per socket in Xeon E7 series – that the market is willing to accept.

This is still only about one-third of what IBM can get away with its top end POWER8 offerings, not to mention its ultrafast, hugely pricey MCM flavours. What if we had a much faster complementary RISC, yet Xeon E7 socket-compatible solution that provides enough extra performance, footprint and feature benefit that the users are willing to pay $10,000 per socket for it?

Especially if much higher instructions per cycle per core could be achieved even in usual apps compared to the x86? The Chinese “Shenwei” Alpha program, leading to a fairly compact 100 PFlop machine in about a year’s time, could – maybe – be the right hint. And yes, it already leaves ARM in the dust.

The post Intel Navigating the New Landscape: Focus on the Golden Goose, or Fight for the Peanuts With the ARM Crowd? appeared first on VR World.