Archive for the category Kevin Knox
No “I” in TEAM
Posted by Kevin Knox in 11:46 am
Loading ...For those of you following this blog, you have probably put it together that I am in fact a long time Pittsburgh Steelers fan. And just for the record, growing up in NY in the 1970’s as a loud-mouth Pittsburgh Steelers fan was not an easy thing. People ask me why I became a Steelers fan and I usually start rattling off the names – Lambert, Ham, Blount, Greene, Webster, Bleier, but in the end it was really the way the Steelers did their talking on the field they made me put up with the constant badgering by want to be Jets and Giants fans. I idolized guys like Dick Butkus and Ray Nitschke both in the way they played the game but also the way they carried themselves on and off the field. Could you imagine TO, Chad Johnson or Tony Romo pulling their antics with Vince Lombardi or Chuck Noll as their coach? To say the game has changed since warriors like Chuck Bednarick or Jim Taylor played would be the understatement of the year.
I guess the one thing that irks me most is seeing a player celebrate a basic play when his team is down 20 points with 2 minutes to play or if he makes a tackle 35 yards down the field. The saying “there is no I in TEAM” does not apply for the majority of the NFL. Football is a team sport and those that succeed are usually those that work together for the better good of the whole. Likewise in business, companies that deliver value through strong partnership are generally those that are most successful. Nowhere is this truer than in the technology market where value is often generated only through the combination or partnership of hardware, software and services.
With our upcoming launch of "Shanghai," we’re seeing an incredible display of teamwork in delivering stable x86 server platforms. As I mentioned a couple weeks ago, our OEM customers were able to quickly validate their platforms and actually pull their launches in. Feedback I’ve received is that they’ve found Shanghai to be an easy platform to bring to market due to the silicon stability and minimal modifications. We had teams of people dedicated to working with each of our OEMs and thanks to this teamwork you can expect to see Shanghai servers in the market very soon. Similarly, we worked in lock-step with our infrastructure partners. Thanks to this cooperation, our channel partners worldwide will be able to offer customized Shanghai platforms.
Our processors are compatible with a wide range of versions of operating systems, and things will be no different with Shanghai. We continued our close technology partnership with the likes of Microsoft, Oracle, Red Hat, SAP and SuSe. This means that customers can expect to find immediate support for Shanghai in all of the leading enterprise applications.
Now for my picks. While it is tempting to blame my 13 year old son for missing both picks last week, I will take full responsibility myself – although Ryan is running a few extra sprints at practice this week to make it up to all the loyal readers 
This week I am pulling out the big guns and having my 15 year old son Tanner making the picks. Tanner plays halfback and linebacker on his high school team and is truly a throwback to the days of Butkus and Nitschke.
Shoe-In = Cards crushing the Niners
Upset = Eagles over the Giants in a close game
Insight = Closer collaboration between hardware and OS suppliers on RAS feature will accelerate the penetration of x86 into the traditionally RISC/Unix mid-range server space
the contents of such linked sites and no endorsement is implied.
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
The Importance of Consistency
Posted by Kevin Knox in 10:55 am
I was reading today about the Oakland Raiders’ latest fired coach filing a grievance for the salary he still believes he is owed. That got me thinking that just five years ago the Oakland Raiders were in a Super Bowl. I actually had to go back and double check because I couldn’t believe it. Since that time the Raiders have had five coaches in six years. This coach de jour approach makes it near impossible to succeed in the NFL.
I recognize that the NFL is a business and you need to win at all costs. I’m also not going to pretend that I know more about football or how to operate an NFL franchise better than Al Davis. However, I do know a thing or two about the importance of consistency and how churn typically doesn’t have a positive impact on whatever your business may be. My unsolicited advice to Mr. Davis is to find the right guy, help him be successful and stick with him. Oh yeah and don’t forget – JUST WIN BABY. On the flip side, take a team like the Seahawks or Eagles who are consistently at the top of their divisions and look at how long Mike Holmgren and Andy Reid have been at the helm. They say a team takes on the personality of its coach, but this takes consistency, and sometimes patience.
This is what AMD did several years ago and continues to do today with respect to its server roadmap. We sought the advice of our customers, both within our OEM partners and as well as some of our largest end-users, and asked them what some of the key things they needed to see from AMD were beyond the obvious ones like continuous improvements in performance, performance-per-watt and virtualization. The answer we heard over and over again was to keep things as simple and consistent as possible for them as they scale.
Our customers are recognizing the benefit of this today, particularly as we prepare to launch our latest AMD Opteron processor code-named "Shanghai." Shanghai drops into our existing infrastructure which has been in place since 2006. Why does this matter? Well frankly, in the current economic and IT spending climate, we are hearing from our customers that it has never mattered more.
People who have never been involved in managing large amounts of servers tend to overlook the importance of having a consistent processor architecture and infrastructure. A constant churn in processor architecture can result in complexity that can create more costs associated with software and labor.
As technology leaders, we as an industry should be doing everything we can to help our customers win. The best way we can help them is to make things as simple as possible for them when we bring new technologies to market.
Virtualization is a good example of this. As virtualization continues to grow and mature, migration has become a significant buying factor for customers. Having a consistent architecture definitely helps enable a smooth migration process. Our customers constantly highlight this as one of the reasons why they continue to scale with AMD. However, they also want to be able to migrate across heterogeneous processor environments as well. Unfortunately, some solution providers in the virtualization hardware and software ecosystem act as if giving customers the ability to do this (and not get locked in to one architecture) will slow the pace of innovation. More like slow the pace of helping customers win.
OK, I made my shoe in and missed my upset last week and in the spirit of keeping consistency, I asked Mac for the picks this week and he told me to ask his brother because he was gearing up for Halloween. So based on input from my 13 year old son Ryan (#51 OL, DE and kicker) here are the picks for the week:
Shoe-In = Cincinnati staying winless against Jacksonville
Upset = The Boys stealing one from the Giants in NY
Insight = Companies which fail to change their IT management processes and procedures will fail to reap the full benefits of server virtualization
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied. |
Signs of Life
Posted by Kevin Knox in 11:47 am
Two weeks away from the keyboard and it seems like an eternity. My last blog left off with the tie donning Mike Nolan still walking the Niners sideline, the Raiders looking more like a CFL team than an NFL team and AMD in questionable financial condition. Well things have certainly changed on all fronts. And in the end change can be a good thing and can lead to a renewed focus and sustainability.
First, the 49ers announced earlier this week that they were letting go of Mike Nolan and promoting Mike Singletary to head coach. While you can’t help but to love Nolan and his “traditional” approach to football I have to admit that I am a bit surprised it took this long. The Niners just are not winning. Handing over the reigns to a gritty, no-nonsense, hard-nosed ex-linebacker like Mike Singletary will certainly be a contrast in personalities. It remains to be seen whether the Niners can win with the team they have, but I for one am sure excited to see change in San Francisco, especially if the Niners take on the personality of their new head coach.
As for the Raiders, they replaced their head coach three weeks ago, putting Tom Cable in charge, and actually won a game this week, beating the Jets in overtime on an amazing 52 yard field goal by Sebastian Janikowski. Now granted the Jets are not lighting things up themselves and Brett Favre would have to be considered the disappointment of the year thus far (another blog for another day), but a win is a win. The team seems to have a new found spirit and approach to the game. Granted the offense still looks more like the Edmonton Eskimos than the Oakland Raiders, but the defense and special teams has shown a new sign of life.
Speaking of a sign of life, AMD announced last week that we’re shipping our new series of Quad-Core AMD Opteron™ processors – codenamed “Shanghai.” We’ve found our footing and are in the midst of preparing for not only the defense of our server business but also going on the offensive with the gains we expect in terms of performance and power-consumption. “Shanghai” was designed with energy-efficiency and performance-per-watt in mind and based on early feedback, our industry partners are fully recognizing this. Industry analysts are already hearing positive things from their clients and how OEMs are working to build systems.
I can tell you from my own experience, once OEM partners had samples in their hands – the entire game began to change. Because of our commitment to platform stability, OEM partners have been able to quickly validate their new platforms and as a result of this, we expect that you should begin to see the first systems with “Shanghai” available before the end of the year.
I am really proud of our “Shanghai” team, as they have led an amazing turnaround in the way AMD designs and brings to market new processors. Not only has the team achieved the recognition we were seeking, the lessons they have learned, and new processes they have established, are already being applied to future processor programs.
And as my colleague Pat Patla, GM of Server Marketing, said, “We’re firing on all cylinders.”
I have to admit that Mac came through for me last time helping me hit the shoe-in and the upset. This week I am a bit hesitant taking his advice after seeing him skip back to the huddle at his game on Sunday (yes, I told him there was no skipping in football). In any case, here we go:
Shoe-In = Philadelphia at home over Atlanta
Upset = The Colts hand the Titans their first loss of the year
Insight = Consolidation among hardware OEMs will be accelerated by the troubled economy
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied. |
Goal-Line Stand
Posted by Kevin Knox in 3:10 am
OK, I am ready for the criticism, so bring it on. How did I miss the shoe-in pick and nail the upset; well, as they say in the NFL, “On Any Given Sunday”. And I know some of you are saying that I was too hard on Brett Favre last week and how he threw for 6 TD passes this week, but I would just remind you to take a look at who he did it against. The Steelers vs. Ravens game was more like a street brawl than a football game and left both teams bloodied and bruised….the way football is suppose to be played. There was another clash with the Texans (yes I said the Texans) taking the Jags to OT but lost on a FG, and then of course there was the battle of Ohio won by the Brownies over the Bungles (I use the word battle very lightly given how both of these teams are playing). But my favorite game of the week was without a doubt the Bears and Eagles match-up.
Down 24-20 in the fourth quarter, Correll Buckhalter took a first down run down to the Bears 3, leaving the Eagles in position for an almost sure touchdown. With shades of Dick Butkus, George Halas and the 46 Defense, the Bears defense reached deep down and stuffed the Eagles on 3 consecutive plays to keep their lead and go onto win the game 24-20. Yes we could debate why the Eagles ran 3 times from the 1, but you have to give credit where credit is due, the Bears defense stepped up when they had to and stonewalled the big bad Eagles. The momentum swung to the Bears and you knew it was over at that point. So just when things looked to be at their worst, the Bears stepped up, reached deep down and did what they had to do….hats off to the Bears.
Now I’ll share my picks for the week, in an attempt to redeem myself after missing my first “shoe-in”. I actually confirmed my picks this week with my 7-year old son, Mac, who is the anchor of the O and D lines of his pee-wee flag team so we are in great shape:
Shoe-In = New England BIG over the Niners
Upset = Miami at home over the Chargers
Insight = Cloud Computing may not materially change enterprise architectures for 2 years
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied.
|
What’s in a Name?
Posted by Kevin Knox in 10:01 am
Before I get into this week’s thought, let me emphasize the point I made last week regarding the importance of the O-Line as the foundation of the offense. I painfully sat through the entire Steelers vs. Eagles game this week and I can honestly say, I have never seen a worse offensive line performance as I did by the Steelers. Not only was Big Ben under constant attack by the Eagles defense, but the Eagles also held Fast Willie to a handful of meaningless yards. I have never seen an offensive line so overpowered and so confused as the Steelers looked. Of course it is not just the O-Line’s fault, but as the foundation, as they go so goes the Steelers offense. And based on last week’s post, I would be remiss if I didn’t acknowledge a huge improvement this week by the Bengals O-Line in a tough OT loss to the Giants.
OK, onto my big Week 3 take away. If you would have told me that in Week 3, Kerry Collins, Gus Ferrotte and Brian Griese would have led the Titans, Vikings and Buccaneers to victory I would have told you that you were nuts. Now in an attempt to get the 0-3 Rams ship turned around, they are turning to 15-year veteran Trent Green.. These are guys who could just as easily be retired playing golf as playing football, never mind winning games in the NFL. Meanwhile some of the staples of the NFL quarterback ranks are having somewhat of a tough start. Peyton Manning looks like a rookie rather than a field general, Brett Favre has not even come close to living up to the NY hype and Derek Anderson’s poor play may in fact give Cleveland a reason to let Brady Quinn try and earn some of the millions they pay him.
Point here is, don’t too easily write off players who may have gone through a rough patch here and there, as they may be just what a certain situation requires. At the same time, teams need to be careful not to get so enamored with big name players that they lose perspective on what’s really important…WINNING. For example, I know it’s still early in the season, but it’s apparent to me already that Brett Favre is never going to live up to the expectations that the Jets, the NY fans or the media have set for him. The more interesting question is at what point this year (if ever) do the Jets swallow pride and pull Brett in lieu of a better option at QB (knowing the Jets I would say the answer is never).
Speaking of tough starts and rough patches, AMD had some challenges in bringing our Quad-Core AMD Opteron™ processor, also known as “Barcelona,” to market. And if you were to believe some of the lurid press reports during the past few months, you’d think AMD was on the verge of an early retirement. Yet, within the past year, Barcelona has become a leading performance and performance-per-watt solution for x86 servers across many of the workloads that are critical for enterprise customers today. Just as Collins, Ferrotte and Griese can still contribute and win at the highest level, the AMD Opteron processor remains a leader in the x86 server market. Also, let’s not lose sight of the fact AMD has only been playing in the server market for five years and over that short period of time, we have been able to deliver impressive improvements in performance and performance-per-watt.[1] We intend to continue this trend later this year with the planned launch of our 45nm “Shanghai” processor.
Substance and stability are the backbone of a good football team – not name and flash. The Titans could attest to that while admitting that paying nearly $60 million for a player doesn’t always equal success. Sometimes a good look under that surface will reveal that things may not be as great as they seem.
In the server world, I would suggest that hype is the exact thing you should avoid. Last time I checked a jingle or a guy with his face painted blue is not helping solving datacenter problems. Based on my conversations with customers, improved performance and functionality is critical for IT decision-makers, but equally important is stability and predictability. And Shanghai – without the hype – is being designed to deliver – in a big way – the performance and functionality enterprises need to win in their industries.
Please keep the comments coming – the good the bad and the ugly. One request I have gotten from several folks was to make some predictions on the upcoming week. Rather than predicting the obvious I thought I would give you a “Shoe-in”, an “Upset” and an “Industry” insight:
Shoe-In = Dallas at home easily over the Skins
Upset = Chiefs upset the Broncos for their first win
Insight = Technology and datacenters are a major purchase consideration amongst Wall St. firms looking to acquire one another
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied. |
[1] In 2003, AMD introduced AMD Opteron™ with one-core (130nm) at 1.8GHz (89w) with no L3 cache.
In 2007, we introduced Quad-Core AMD Opteron™ with four-cores (65nm) at 2.30GHz (95w) with 2MB L3 cache and IPC improvements.
Importance of a Solid Infrastructure
Posted by Kevin Knox in 1:32 pm
It was quite a Week 2 in the National Football League. Blown call. Risky 2-point conversion. Brett Favre’s first loss as a Jet. The Patriots being the Patriots. And of course another Steelers hard-nose win vs. the Browns.
One team who continues to surprise many – in a bad way – is the Cincinnati Bengals. They lost their home opener 24-7 to a team without their starting quarterback. And after two games, here’s how they stack-up against the other teams: ranked 30th in points per game, 31st in total offense and 32nd in Passing Yards per game. (Just a reminder; there are only 32 teams in the NFL.) And to add further insult to their performance, the Bengals have the distinction of having more “12 men on the field” penalties then they do touchdowns.
But what is most surprising is the cadre of Pro-Bowl caliber offensive weapons the Bengals posses; Houshmandzadeh, Chad Johnson (I refuse to refer to him as Ocho Cinco – a post for another time) and Carson Palmer.
While no one can disagree that these three aren’t playing at the top of their game, there seems to be a much larger problem – the foundation of the offense line. A lackluster rushing game generally has more to do with poor O-Line performance than poor back performance. And judging by the pressure on Palmer all day, there’s definitely a weakness with their offensive line.
Personally, I love offensive linemen and would offer that they are the toughest, most dedicated unsung heroes of any winning team. They are the foundation of the infrastructure that makes players like Palmer and Johnson so great. When the O-Line clicks so do the Bengals and right now they are not clicking.
Thinking about infrastructure and foundation, Intel this week officially announced their “Dunnington” server processor with all the glitz you would expect from a marketing company like Intel (more cores!, more cache!, more power!). However the core infrastructure of the chip is still fundamentally flawed with an off-die memory controller and bottlenecks galore in the overall micro-architecture.
Just as you’ll likely see some team (see Oakland Raiders) fire their head coach as a stopgap measure, “Dunnington” is another stopgap solution our competition has targeted for multi-processor based servers. It continues to leverage Intel’s current approach to multi-core computing for multi-processor servers, and its use of a traditional Front-Side Bus (FSB) architecture.
So in many respects the Bengals and Dunnington are suffering from the same problem (you won’t see that comparison anywhere else for sure), weak core infrastructure and foundation. While I don’t have a lot of confidence in the Bengals beating the Giants this week, I do believe they have a better chance than “Dunnington” does in adding significant value for many IT shops.
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied. |
On the NFL Pre-season and Processor Hype…
Posted by Kevin Knox in 8:47 am
Every week during football season, I will post some of my observations from the previous weekend’s NFL games on this blog.
While the majority of the headlines from week one focused on Tom Brady and his season-ending knee injury, one of the things that I think was overlooked was the absolute throttling the Pittsburgh Steelers put on the Houston Texans.
Given the record of the two teams over the past few years, this normally would just be business as usual in the NFL. However, for some reason, because the Houston Texans happened to finish 8-8 last season, they were suddenly the favorite sleeper pick to make the playoffs as a wildcard team by many media outlets in the pre-season.
As I sat down to watch the Texans and Steelers play, I was looking forward to watching what many were expecting to be a close game. However, those hopes ended quickly as I watched the Steelers jump out to a 21-3 halftime lead. The lead was extended to 35-3 by the end of the third quarter as Willie Parker completely ran over the Houston’s defensive line, one of their supposed strengths coming into this season.
The game left me wondering what all the hype was about with respect to the Texans. This was a team that if you look at the personnel, really hadn’t changed much from the previous season. It’s just that some of their younger players, of which they have several, were probably a bit more mature. I know it’s only one game but right now it appears the Texans are the current poster child for why you shouldn’t buy into the hype until you see for yourself.
This naturally had me thinking about the current landscape for microprocessors (because when I’m not thinking about football, I think about processors) and where some of the hype exists today. One of the things I have been hearing a lot about in the past few weeks is how our competitor plans to bring a six-core processor to market designed for multi-processor servers.
The fact that their new processor is the first x86 processor to have six-cores is a nice achievement but my question is what else is there? How much have they really improved the bottlenecks in their architecture that have resulted in inefficiencies in their previous platforms when it came to memory-intensive workloads? Does adding two more cores in fact make their front-side bus work that much harder and thereby exacerbating their current bottleneck?
The other question customers should be asking is do you have to pay a power penalty with the addition of two more cores, a larger cache when you combine that with the already power-hungry fully-buffered DIMM memory their platforms leverage?
As AMD gets ready to ship its “Shanghai” processor in the fourth quarter, customers can expect to get even more performance, more memory bandwidth and even better power efficiency than we offer today with our current Quad-Core AMD Opteron processors. Also, you can expect to see platforms shipping from most OEMs that offer the “Shanghai” processor before the end of the year.
So much like the NFL pre-season, I would encourage you to not buy into the hype and ask the right questions when it comes to new server processor technologies being brought to market later this year by both AMD and our competitor. Better yet, talk to your system solution provider and let them help you test the platform that best suits your needs, there may be a sleeper pick out there which is more than just hype.
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied. |
Virtualization: Something Old, Someone New and Some Really Big Cost Savings
Posted by Kevin Knox in 12:28 pm
I always find it interesting to hear people singing the praises of x86 virtualization and talking about how this recently introduced technology is already revolutionizing the industry. Fact of the matter is that virtualization technology was originally introduced for the purpose of time sharing on mainframes in the early 1970’s. One could probably trace the roots of x86 virtualization to the early 90’s when IT managers finally realized that people’s desks and wiring closets were probably no place for servers and started to relocate them into the datacenter. The next step in the evolution of x86 virtualization was a few years later when IT managers realized they could safely run multiple applications on a single server simultaneously. And if you are reading this blog, you can probably figure out what came next…..running multiple versions of an OS on a single piece of hardware, or what has affectionately become known as Virtualization. While certainly an interesting history, there are two major happenings on the near horizon that I believe will permanently change the face of virtualization.
First is the notion of virtualization becoming “intelligent.” Think about the possibilities when an enterprise infrastructure becomes smart enough to essentially manage itself based on a predefined set of rules and current application requirements. Application A runs for 1 hour and when it is finished, it shuts down, creates a new virtual session with ½ its resources to start a nightly backup and it reallocates the other ½ of the resources to another session getting ready to kick off a major batch run. While just an example, hopefully it shows the possibilities of what some are calling Dynamic Provisioning. In the end what this will enable is the elimination of human capital, which for most IT shops is still the largest budget item they have. No longer will masses of IT people be needed for starting, stopping, configuring and managing virtual sessions. On top of that, by optimizing the software environment in such a way, less, better utilized hardware will result.
The second and probably more important happening is Microsoft’s entry into the virtualization space. While Microsoft has been on the fringes for a while and dabbled with partnerships, acquisition and new technologies, they finally seem to have a clear strategy and roadmap for virtualization technology. This is not to suggest they will be better or worse than anyone else, but rather having Microsoft in the game will help drive innovation, push the competition and allow for a more competitive marketplace, which ends up being a good thing for end users. Microsoft’s years of experience in operating systems and infrastructure software will also help enable new features and function and enable emerging technologies such as Dynamic Provisioning. One thing I can assure you from my years at AMD is that competition drives innovation and value for the end-users, for the ecosystem and the industry as a whole.
To illustrate this, consider the following advances that AMD has driven into the x86 processor over the past 5 years together with greater choices in virtualization software that are starting to open the doors for IT departments, enabling them to run a larger range of workloads on virtualized servers. 64-bit and multi-core processor technologies now provide the robust platform needed for memory-intensive virtualization. More advanced power management capabilities of the processor are helping to reduce power and cooling costs. AMD-V technology, hardware-enabled virtualization inherent in our processors, is enabling software like Hyper-V to handle the most demanding applications.
So what does the future hold? Well I can’t tell you that, but I can assure you we are in the midst of the most interesting time in the history of x86 virtualization from both a competitive as well an innovation perspective. As Bob Dylan once sang “The Times They Are A-Changin”.
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for
the contents of such linked sites and no endorsement is implied. |
You Can’t Make a Cake with Just Flour… Why Performance Benchmarks Alone Don’t Work
Posted by Kevin Knox in 3:54 am
Seems like everywhere I turn people are talking about IT best practices. But what about worst practices? After all, most of us were taught that you need to learn from your mistakes. So let me take a second to share some real-life worst practices with respect to server selection and inappropriate use of performance benchmarks to get you thinking:
One HPC end customer once told me that he knew his application would not perform as well on one particular configuration as another, but achieving a specific benchmark score and having a higher ranking on the Top 500 list than their chief competition was what was driving their decision
A large server cluster was purchased based on performance benchmarks and the theoretical performance it could achieve. After spending months in product and vendor selection, procurement, acquisition, assembly, the user found that the datacenter didn’t have enough power to run the servers and the system had to be replaced by lower power servers with an alternate chip architecture (can you guess what the alternate was?)
A Fortune 500 company reported that they were seeing almost a 10x performance difference between 2 systems they were testing using home grown benchmarks. Sensing something didn’t seem right; we took a look at the benchmark code and discovered that the dataset being used fit entirely in the L2 cache of one of the systems. After minimally increasing the dataset size, the results changed very quickly and very dramatically in favor of the other system.
Pretty ugly stuff if you ask me. Now this is not to suggest that benchmarks are unimportant to consider when selecting hardware. They are, but as a sole selection criteria, they may very well lead you astray. The terms benchmarks” and “performance” are not synonyms and are not interchangeable. With the advent of things like server virtualization, the growing importance of power and energy, and the quest for ever better application scalability, the “one size fits all” benchmark approach to server selection no longer works when looking for optimal performance. Consider the following perspective on performance:
- Virtualization = the amount of performance I can get out of a single physical piece of hardware via virtualization software. Better utilization. Let me elaborate a bit more here as I know this is the hot topic for everyone today. End-users must not only look at how fast their virtual sessions can run, but also how many virtual sessions can run efficiently on a single server.
- Power and Energy = The amount of performance I can get out of a piece of hardware per a given amount of power. Performance Per Watt.
- Scalability = The amount of performance I can get out of an application by adding additional processors or cores to the available compute resources.
- Benchmarks = The amount of performance generated for a specific set of routines and/or application modules. Benchmarking your own applications is always best.
- Price = The amount of performance for a fixed amount of money. Or sometimes people use a fixed amount of performance for the best price. Price/Performance.
So what does all of this mean? Simply put, performance is not what it used to be.
By looking differently at performance across a set of relevant vectors, IT decision-makers can significantly improve server selection and better map to specific application and business requirements. My hope is that this comes across as somewhat of a wakeup call and light bulbs are going off as you read this. In the coming weeks, AMD plans to release a number of documents on this topic and to provide metrics and frameworks for helping create server selection models which look beyond just performance as a benchmark result.
I know some of you might suggest that I’m just trying to detract attention from our benchmark scores. Fact of the matter is AMD wins more benchmarks today than we have in years, so this is far from a self-serving, but hopefully it’s self motivating for you and keeps you from being a worst practices example.
|
Kevin Knox is Vice President of Worldwide Commercial Business at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied. |
(4.93 out of 5)