Обсуждение: Core 2 or Opteron
Hello, We're planning new server or two for PostgreSQL and I'm wondering Intel Core 2 (Woodcrest for servers?) or Opteron is faster for PostgreSQL now? When I look through hardware sites Core 2 wins. But I believe those tests mostly are being done in 32 bits. Does the picture change in 64 bits? And I also remember that in PostgreSQL Opteron earlier had huge advantage over older Xeons. But did Intel manage to change picture now? Thanks, Mindaugas
These benchmarks are all done using 64 bit linux: http://tweakers.net/reviews/646 Best regards, Arjen On 7-12-2006 11:18 Mindaugas wrote: > > Hello, > > We're planning new server or two for PostgreSQL and I'm wondering Intel > Core 2 (Woodcrest for servers?) or Opteron is faster for PostgreSQL now? > > When I look through hardware sites Core 2 wins. But I believe those tests > mostly are being done in 32 bits. Does the picture change in 64 bits? > > And I also remember that in PostgreSQL Opteron earlier had huge advantage > over older Xeons. But did Intel manage to change picture now? > > Thanks, > > Mindaugas > > > ---------------------------(end of broadcast)--------------------------- > TIP 9: In versions below 8.0, the planner will ignore your desire to > choose an index scan if your joining column's datatypes do not > match >
> We're planning new server or two for PostgreSQL and I'm wondering Intel > Core 2 (Woodcrest for servers?) or Opteron is faster for PostgreSQL now? > > When I look through hardware sites Core 2 wins. But I believe those tests > mostly are being done in 32 bits. Does the picture change in 64 bits? We just migrated from a 4-way opteron @ 2 GHz with 8 GB ram to a DL380 G5 with a 4-way woodcrest @ 3 GHz and 16 GB ram. It was like night and day, system load dropped, not just quite a bit, but almost by a factor of 100 in worst case scenarios. Going from a 64 MB diskcontroller to a 256 MB ditto probably helped some and so did a speedup from 2 -> 3 GHz, but overall it seems the new woodcrest cpu's feel at home doing db-stuff. This is on FreeBSD 6.2 RC1 and postgresql 7.4.14. > And I also remember that in PostgreSQL Opteron earlier had huge advantage > over older Xeons. But did Intel manage to change picture now? That was pre-woodcrest, aka. nocona and before. Horrible and the reason I went for opteron to begin with. But AMD probably wont sit idle. The link posted in another reply illustrates the current situation quite well. regards Claus
> These benchmarks are all done using 64 bit linux: > http://tweakers.net/reviews/646 I see. Thanks. Now about 2 core vs 4 core Woodcrest. For HP DL360 I see similarly priced dual core 5160@3GHz and four core E5320@1.86Ghz. According to article's scaling data PostgreSQL performance should be similar (1.86GHz * 2 * 80% = ~3GHz). And quad core has slightly slower FSB (1066 vs 1333). So it looks like more likely dual core 5160 Woodrest is the way to go if I want "ultimate" performance on two sockets? Besides that I think it should consume a bit less power!? Mindaugas
On 7-12-2006 12:05 Mindaugas wrote: > Now about 2 core vs 4 core Woodcrest. For HP DL360 I see similarly > priced dual core 5160@3GHz and four core E5320@1.86Ghz. According to > article's scaling data PostgreSQL performance should be similar (1.86GHz > * 2 * 80% = ~3GHz). And quad core has slightly slower FSB (1066 vs 1333). > > So it looks like more likely dual core 5160 Woodrest is the way to go > if I want "ultimate" performance on two sockets? > Besides that I think it should consume a bit less power!? I think that's the better choice yes. I've seen the X5355 (quad core 2.66Ghz) in work and that one is faster than the 5160 we tested. But its not as much faster as the extra ghz' could imply, so the 5320 would very likely not outperform the 5160. At least not in our postgresql benchmark. Besides that you end up with a slower FSB for more cores (1333 / 2 = 666 per core, 1066 / 4 = 266 per core!) while there will be more traffic since the seperate "dual cores" on the quad core communicate via the bus and there are more cores so there is also in an absolute sence more cache coherency traffic... So I'd definitely go with the 5160 or perhaps just the 5150 if the savings can allow for better I/O or more memory. Best regards, Arjen