I hope future amd apu’s will enable the use of ecc memory.

Quad Core, around $100 price point and 45W max TDP? Sounds perfect for a secretary, lawyer workstation.

However, if you compare benchmarks you will find that there are those that are audited, such as SPEC, TPC and SAP. In those cases, our OEM partners (and AMD) have performance labs, we tune and optimize systems, we release benchmarks, and a third party can audit those benchmarks to ensure that they are accurate. In these benchmarks, we easily surpass Intel in performance.

Then there are other benchmarks which are not audited and do not reflect actual performance. I can’t speak to their validity, but if you had to choose (as most customers do) between those that are coming directly from vendors like HP and Dell or those that are coming from some web site, which would you choose?

I think your PR people should check this out:

(link removed)

It was in response to the following blog entry:

http://blogs.amd.com/work/2011/11/13/putting-performance-in-perspective-%E2%80%93-amd-opteron%E2%84%A2-6200-series-processors/

Basically the results are being twisted to make the latest Opterons look very poor. Loads of sites are reproducing the article ad-verbatim.

This is despite many people pointing out to the author in the comments that the AMD system has more storage. Even the author knows that one system is using MLC drive and the other SLC drives if you look at the comments.
These cost much more than the Intel system. On top of this the pricing of the SSDs in the Intel system is not final.

However,he does not mention this in the article.

I actually believe that you are wrong – we have great prodcuts and we do defend them.

Being up to 84% faster than the competition at a fraction of the price puts us in a pretty good position. There will always be those that are predisposed to buy the competitor’s products. They have no intention of changing their minds, so engaging the discussion doesn’t lead anywhere. I’d rather spend my time with customers who are actually open minded and believe there is an honest, open discussion out there.

When you compare the actual, AUDITED benchmarks, you find that we actually have significantly better performance:

HPC: 84% greater LINPACK performance over Intel
Memory Bandwidth (STREAM): 73% greater than Intel
Integer (general applications): AMD scores 543, 29% over Intel’s best 2P at only 421
Floating Point (technical applications): AMD scores 403, 47% over Intel’s best 2P at only 274
Database: AMD scores the highest 2P score ever in TPC-C
Business Applications: AMD beats Intel by 11-12% in SAP

So, if I had to take a side on performance, I would rather look to a benchmark submitted to an audited group (like SPEC, TPC, SAP, etc) where AMD and our competitor can submit their own optimized scores. This, to me, seems to be the best way to truly tell the performance – each vendor gets to post their own results, let the best man win. And, I believe the best has won.

As for the LINPACK and STREAM numbers that I called out above, those are not audited benchmarks, but I can say that both of them a pretty easy to check with your own math. Knowing how many FPUs and how many memory channels (and the speed of the memory) should allow you to correlate the results.

Now, when you combine this with the fact that we are ~35% lower in cost (at roughly the same power level) you see that AMD actually has a great product and customers will have a great opportunity to get better performance AND better price/performance with AMD. As a matter of fact, if you compare our processors to the competition, at the same price level, you see AMD ahead by ~90%. Pretty incredible.

As a senior product marketing manager of AMD, should you write in to defend AMD’s products. What are the reasons that AMD’s product managers won’t stand up and defend AMD’s product? Products not being truly as good or product marketing managers not doing their jobs?

Would you mind to comment about the article below ?
(link edited)

Many thanks.

For example, a 16-core AMD “Interlagos” processor has 8 compute units each with 2 cores. For the purpose of licensing in vSphere 4.1 update 2, this processor is seen as needing only an 8-core license.

More details can be found at:

https://www.vmware.com/support/vsphere4/doc/vsp_esxi41_u2_rel_notes.html

84% faster in linpack
73% faster in memory throughput
39% faster in SPEC FP
25% faster in SPEC Int
24% faster in SPEC JBB
We have the fastest 2P server in the world in TPC-C as well

If you consider this “losing to the competition” then I think that you are the only person in that camp.

As to the comparison, if you read through the blog, you’d notice that I compared two processors at the same price level, which is exactly how customers buy. Customers don’t go out looking for a particular core count, customers have a price in mind. And at the same price level, I am 89% faster than the competition. I’m guessing you would call being 89% faster a loss too, right?

I can’t speak to individual reviewers and their methodologies. If you look at our benchmarks on AMD.com you’ll see that our advantages over the competition are clear:

HPC: up to 84%
Memory throughput: 73%
Integer: 25%
Floating Point: 39%

If you look at HP you’ll see that they posted the fastet 2 socket TPC-C benchmark ever. They also posted leading benchmarks in SAP as well. I know that there are some virtualization benchmarks, but I am sitting in an airline lounge after 20 hours of travel and I am a bit hazy.

Plus, I have a very good feeling that you’ll see some additional benchmarks coming out this week from a major OEM.

So, if our benchmarks look great and our partners’ benchmarks look great, then I would think that the products really do perform great. I can’t speak to anyone else’s results, that is probably more of a question for them.

Have you seen the one below:

What happened? Can you provide more technical details since lots people being confused?

Thanks.

http://www.zdnet.com/blog/datacenter/amd-set-to-bulldoze-the-datacenter/1111

seems like every review i read about bulldozer both server and desktop are totally negative – saying these chips just cant compete with intel…