By: Michael S (already5chosen.delete@this.yahoo.com), January 13, 2011 4:22 am
Room: Moderated Discussions
Mark Roulo (nothanks@xxx.com) on 1/12/11 wrote:
---------------------------
>Brett (ggtgp@yahoo.com) on 1/12/11 wrote:
>---------------------------
>>Thank you, I was indeed wrong.
>
>You are welcome :-)
>
>>... This is the reason I expected Intels 4 socket system sales to fall of the edge of the earth.
>>AMD peaked at 56% of the 4 socket market instead.
>>I was wrong.
>
>For what it is worth, it became clear to me at this time that I did *NOT* understand
>the 4 and 8 way socket server market. Given AMD's clear advantage I expected Intel
>to get creamed. Intel did lose market share, but not nearly as much as I expected.
>
>I suspect that something other than performance was the issue. Perceived reliability
>of the Intel systems. Or RAS features on the chip. Or something.
>
>But I don't really know.
>
>-Regards,
>Mark Roulo
>
Bandwidth or not, Intel's Tulsa easily beat contemporary Opterons in TPC-C:
http://www.tpc.org/tpcc/results/tpcc_result_detail.asp?id=106101901
http://www.tpc.org/tpcc/results/tpcc_result_detail.asp?id=106092601
Opteron was king in math, which is pretty much irrelevant for 4S x86.
Opteron was king in server consolidation, that was somewhat relevant back then (now, I think, even consolidation favors 2S). That's, probably, was the main driver behind its market share gains.
Opteron did well in data mining, but that's not that big a market.
But in things like SAP and OLTP Tulsa was faster.
Next generation, i.e. Tigerton vs Barcelona, was more of the same - AMD becoming stronger at their strongholds and Intel increasing lead in theirs.
Then come yet another generation Shanghai and soon thereafter Istanbul vs Dunnington. AMD improvements were much bigger than Intel's and they caught up even at Intel's former strong areas. So AMD had almost as big technical advantage in 4S space as in early Opteron days. But by then, due to Nehalem, 4S itself started too look less and less relevant.
---------------------------
>Brett (ggtgp@yahoo.com) on 1/12/11 wrote:
>---------------------------
>>Thank you, I was indeed wrong.
>
>You are welcome :-)
>
>>... This is the reason I expected Intels 4 socket system sales to fall of the edge of the earth.
>>AMD peaked at 56% of the 4 socket market instead.
>>I was wrong.
>
>For what it is worth, it became clear to me at this time that I did *NOT* understand
>the 4 and 8 way socket server market. Given AMD's clear advantage I expected Intel
>to get creamed. Intel did lose market share, but not nearly as much as I expected.
>
>I suspect that something other than performance was the issue. Perceived reliability
>of the Intel systems. Or RAS features on the chip. Or something.
>
>But I don't really know.
>
>-Regards,
>Mark Roulo
>
Bandwidth or not, Intel's Tulsa easily beat contemporary Opterons in TPC-C:
http://www.tpc.org/tpcc/results/tpcc_result_detail.asp?id=106101901
http://www.tpc.org/tpcc/results/tpcc_result_detail.asp?id=106092601
Opteron was king in math, which is pretty much irrelevant for 4S x86.
Opteron was king in server consolidation, that was somewhat relevant back then (now, I think, even consolidation favors 2S). That's, probably, was the main driver behind its market share gains.
Opteron did well in data mining, but that's not that big a market.
But in things like SAP and OLTP Tulsa was faster.
Next generation, i.e. Tigerton vs Barcelona, was more of the same - AMD becoming stronger at their strongholds and Intel increasing lead in theirs.
Then come yet another generation Shanghai and soon thereafter Istanbul vs Dunnington. AMD improvements were much bigger than Intel's and they caught up even at Intel's former strong areas. So AMD had almost as big technical advantage in 4S space as in early Opteron days. But by then, due to Nehalem, 4S itself started too look less and less relevant.
