Tuesday 27 December 2011

Some Thoughts on AMD versus Intel

The following is a direct copy and paste of a post to an e-mail list that I belong to. It discusses some of the reasons behind our not dealing with AMD product as a rule.

AMD versus Intel

We were burned by a bad run of Asus boards a while back.

Our first venture back into AMD with an Asus based board ended in failure. That was the last time we looked at Asus for boards and AMD for processors.

Asus has some great products. We run with their corporate EeeBox PCs both factory built and using their bare bones units for certain applications. They are extremely fast, light weight, and easy to mount almost anywhere. With an 80GB 320 Series Intel SSD installed the EeeBox PC EB1012P is an awesome entry level (word processing, some picture editing, and light weight computing needs) unit that can outperform any entry level PC the way we configure them.

The only recent beef I have with Asus is that their support and RMA process is a bear to work with. A real PITA x100.

On the Intel side of things we have had one or two manufacturer’s situations that required RMAs (bad caps). We have had a number of RMAs that were due to user “interaction” with the systems (wool + polyester + Canadian dry winters = dead on touch). Many case manufacturers have the grounding shields around the front USB ports _not grounded_ in any way.

As far as the AMD versus Intel situation, the biggest cost to us when it comes to supporting non-Tier 1 servers, desktops, workstations, and laptops is in the troubleshooting and RMA process. We are on the hook for a good chunk of that if the problem turns out to be hardware related after all. The exception to that is for on-site troubleshooting which is billable in most cases.

AMD has for years been trying to build up an ecosystem similar to Intel for their partners. The problem with AMD on all platforms is that it does not _own_ all aspects of the ecosystem. Intel does. As a result, when we have a problem we have 1 number to call, an advanced warranty program that sends out a replacement to our door within 24 hours, and an amazing support structure both on the Web and via phone.

So, not only do we pay a premium for Supercom, Tyan, LSI, and other components for an AMD based server we have to figure out which way to go when we have a hardware problem. There are no finger pointing matches when it comes to Intel product. There are when it comes to trying to deal with several different manufacturers that try and blame the other (certain Tier 1s are bad for this at their entry level support/warranties via phone too) on the AMD platform. BTDT and burned that t-shirt but good man.

When it comes to server product we have the Intel Server Configurator. This online tool gives us the ability to tailor a server system to our client’s needs. It also contains a wealth of components that have been tested and approved for each server component combination possible with Intel server product. At last look by us there was virtually no such database for AMD based server products. So, for the system builder they may need to spend hundreds if not thousands of dollars for components and then a lot of time testing those components to see if they play nice together.

We have 2 Intel Server Adapters (Quad Port ET2s) that were mistakenly ordered for a 2 node cluster (we assumed (yeah yeah) that they would show up on the approved list) sitting on a shelf waiting for a new Intel Server System that they are approved for. That error on my part cost us almost a grand.

We spent over $15K-$25K for product R&D this last year and who knows how much in time. We did end up with a viable product at the end of it and it is selling really well.

http://blog.mpecsinc.ca/search/label/Intel%20SR1695GPRX (Post 2 outlines everything – 2 Node for under $20K).

We are working with IBM on a similar configuration using their products (gives us access to 4 hour response times for hardware issues). Dell and HP have failed to follow-up on any inquiries or nowhere near capable of a similar setup. IBM has been really good about facilitating our configuration questions.


We have put a lot of time and effort into getting to know what products and system configurations are best suited to our client’s needs. This effort includes both physical and Cloud based products.

Unfortunately AMD has not come up with a viable partner support system that makes their products attractive to us at this time. Plus, any time we have ventured back into AMD territory, as mentioned above, we have hit a wall or product failure that lead to RMA pain.

So, we will stick with Intel for both server and desktop/workstation platforms for now. Their products have been very reliable with any support or RMAs being facilitated with an excellent partner ecosystem.

Philip Elder
Microsoft Small Business Specialists
Co-Author: SBS 2008 Blueprint Book

*Our original iMac was stolen (previous blog post). We now have a new MacBook Pro courtesy of Vlad Mazek, owner of OWN.

Windows Live Writer


Josh Gay said...

On the NIC's beware, Intel (NIC) on Intel (Board) is generally supported, but they DON'T much of the time get that info onto the tested list, and support (Server aka EPSD at least at 1st tier will happily take the cop out of yeah not on the list, try one on the list). Had this happen, NIC wasn't even detecting, so I bought one on the list, with overnight shipping.... no detect, tried a dual port that was handy, no detect.... they replaced the riser, still no detect, so had to do the mainboard. Had a chat with a nice man from the networking division at the last event about this..... yes the "newer" card is "supported" just hasn't run through the "validation" process @ EPSD.... but like other times in the past (if the memory isn't on the list, sometimes it better to lie so they'll actually walk you far enough to find the bad component.... and yeah, at least once it was a DOA stick, that worked on the bench for preinstall and failed onsite.

Philip Elder Cluster MVP said...


We have had our fair share of issues with older Intel NICs though not so much with current technology.

With regards to RAM we stopped using non-Intel certified RAM a long time ago due to that very situation where we deployed a server with non-certified RAM. It ran fine for about 2 weeks on our system builder's bench.

Just prior to us picking it up the box started spontaneouosly rebooting on them. Turned out to be the non-certified Intel RAM. We changed that out and all was okay.

That situation came close to putting us in a real bad spot with our client.

So, we always use Intel certified even if it means that we use one generation back for situations where we need 32GB of RAM (SR1695GPRX or S3420GPLX or RX).

It's too risky to go otherwise in our experience.

Thanks for the comments. :)