Istanbul VS Nehalem VS Harpertown (big pics)

[akhilles]

AMD Opteron 2435 CPU Review

Author: James Gorbold
Published: 7th July 2009

Opteron 2435 CPU details

* Frequency: 2.6GHz
* Core: Istanbul
* Packaging: Socket F
* L1 cache: 128KB per core
* L2 cache: 512KB per core
* L3 cache: 6MB accessible by all cores
* Memory controller: 2.2GHz
* Memory: Dual-channel ECC registered DDR2-800MHz
* TDP: 75W













Final Thoughts

Clearly, AMD has done a great job engineering the Opteron 2435, as these six-core CPUs have a very similar thermal profile and power consumption to that of a pair of identical frequency quad-core Opterons. Given the extra performance, the six-core models provide in certain applications, this is no mean feat.

However, while the comparatively low power draw of the Opteron 2435s may make them attractive to those upgrading a power-starved data centre, the way by which AMD has artificially limited performance by keeping the clock frequency down means they aren't an attractive purchase for a workstation or HPC server.

In these scenarios, the most important aspect is how much processing power you can squeeze into a single box, and for that, the Xeon W5580 reigns supreme, especially now that their price has tumbled by nearly £300 to £1,198.57 inc VAT.

bit-tech.net | Review - AMD Opteron 2435 CPU Review

High Performance Linpack on Xeon 5500 v. Opteron 2400

Written by Shane Corder - Cluster Engineer
Tuesday, 16 June 2009 16:23

Results:

CPU Model	Problem Size (N)	Theoretical Peak 	Actual Peak 	Efficiency 	Node Cost 	$ per GFLOP
Nehalem X5550 2.66GHz 	35840 	85.12 GFLOPS 	74.03 GFLOPS 	86.97% 	$3,800.00 	$51.33
Istanbul 2435 2.6GHz 	41216 	124.8 GFLOPS 	99.38 GFLOPS 	79.63% 	$3,500.00 	$35.21

Conclusion:

When viewing HPL results there are two interesting figures to look at: the Actual Peak which is what is measured by the benchmark and comparing this number to what the theoretical best performance the processor can provide (Theoretical Peak). This is referred to as the efficiency. We've also included the rough prices of the systems and a GFLOP per dollar rating. As you can see, AMD beats Intel on GFLOPS per dollar and peak performance, but loses on overall efficiency. This shows us that while the 6 cores per CPU that AMD Istanbul is offering provides better raw horsepower, the overall system architecture is not as balanced as Intel's Nehalem. The lower efficiency rating is most likely caused by the lack of memory bandwidth, and increased cache snoops in the Istanbul system. The CPU's sit idle for a longer period of time while waiting for data from main memory and while checking for cache hits in all of the system's 12 cores . Memory bandwidth can have a huge impact on overall system performance, and is beyond the scope of this document--it will be covered in a latter post.

So, while the Nehalem may have the best performance per core and higher efficiency, the Istanbul does a good job of making up for its deficiencies by adding additional cores. When choosing a system architecture for your next cluster, HPL should be only one of the benchmarks you use in your evaluation, We will be updating this blog with more performance results and benchmarks over the next couple of months. If you have any ideas or code that you'd like to see tested, please let us know -- send an email to scorder@advancedclustering.com This e-mail address is being protected from spambots. You need JavaScript enabled to view it.

Advanced Clustering | High Performance Linpack on Xeon 5500 v. Opteron 2400 | Company Blog

[arisythila]

I'd really like to see clock for clock comparisons. I know that processor is AMD's top dog, and this one is Intels top dog. But it would be interesting to see how they did for clock for clock. More to see where they run together...

[ryman554]

I'd really like to see clock for clock comparisons. I know that processor is AMD's top dog, and this one is Intels top dog. But it would be interesting to see how they did for clock for clock. More to see where they run together...

Didn't that (small) HPC example do just that?
2.6 vs. 2.66?

[Apt403]

I'd really like to see clock for clock comparisons. I know that processor is AMD's top dog, and this one is Intels top dog. But it would be interesting to see how they did for clock for clock. More to see where they run together...

That, and the 2435 is poised to take on the X5500, they're in the same price point, around a thousand.

Even though they're comparing the fastest CPU each manufacturer is producing ATM, it's still not a fair comparison, considering the W5580 is a $1600 part, it's in a completely different price range than the 2435.

Plus, the Istanbul chips aren't running on the Hypertransport 3.0 bus yet, no mobos will support it until AMD's 800 (IIRC) chipset lands sometime in the next couple months, so the available bandwidth during that test wasn't really optimal.

And Crysis seems to like appreciate more clockspeed than anything, so it's not exactly the most fair test to be run in this comparison.

I mean, I'm not saying that AMD would win hands down in every benchmark, Intel still DEFINITELY has the advantage, but this review didn't really show Istanbul's true potential.

[Viperabyss]

I actually want to see them using benchmarks that actually matters, not this Crysis, Cinebench, and H.264 BS.

[Z3r0c00l]

That, and the 2435 is poised to take on the X5500, they're in the same price point, around a thousand.

Even though they're comparing the fastest CPU each manufacturer is producing ATM, it's still not a fair comparison, considering the W5580 is a $1600 part, it's in a completely different price range than the 2435.

Plus, the Istanbul chips aren't running on the Hypertransport 3.0 bus yet, no mobos will support it until AMD's 800 (IIRC) chipset lands sometime in the next couple months, so the available bandwidth during that test wasn't really optimal.

And Crysis seems to like appreciate more clockspeed than anything, so it's not exactly the most fair test to be run in this comparison.

I mean, I'm not saying that AMD would win hands down in every benchmark, Intel still DEFINITELY has the advantage, but this review didn't really show Istanbul's true potential.

Because there is no true potential for Istanbul. Once Istanbul get's those upgrades Intel will have it's EX Xeons out the door (8 core with HT).

[Apt403]

Which will probably be in a different price range than the Istanbul parts at that point. The only Nehalem part ATM that's in the same price range as the Shanghai chips, for example, is the dual core (no HT) 5502.

Nehalem EX and AMD's Magny-Cours procs should be hitting around the same time, early 2010, late 2009. Both should be pretty competive with each other, depending on the price point, but I doubt AMD will have the performance advantage outright.

Too bad AMD didn't have it's own version of hyperthreading... Intel's got Quickpath to compete with hypertransport, AMD should come back with a hyperthreading clone.. Or hire some of the SPARC engineers and just get an x86 version of the concept behind the UltraSPARC T2. 8 cores, 8 threads per core, 64 simutaneous threads.

I dunno, I've been awake for the better part of 50 hours now, so I'm not thinking too clearly.

When it comes down to it, I'm buying from whoever gives me a better price/performance ratio, so I'm rooting for both sides.

[arisythila]

Were talking the server market. Not the desktop market.

I run a couple of data centers. From what i've seen the amount of power you draw is CRUCIAL. I defiantly want to see the 6 core against similar price point to the Intel at the similar price. So if its a 1000 dollar processor, you shouldn't compare it to a 1600 dollar processor. I would also like to see real world server tests. SQL performance, memory bandwidth. We are currently running a couple of AMD servers in my dallas facility. I was looking at the nehalem setup. Were currently running shanghai processors in the Dallas servers.

~Michael

[Apt403]

What're the specs on the Shanghai servers and what's running on them?

[arisythila]

one is a linux server,

processor       : 7
vendor_id       : AuthenticAMD
cpu family      : 16
model           : 4
model name      : Quad-Core AMD Opteron(tm) Processor 2376
stepping        : 2
cpu MHz         : 2311.863
cache size      : 512 KB
physical id     : 1
siblings        : 4
core id         : 3
cpu cores       : 4
apicid          : 7
fpu             : yes
fpu_exception   : yes
cpuid level     : 5
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm 3dnowext 3dnow constant_tsc pni cx16 popcnt lahf_lm cmp_legacy svm extapic cr8_legacy altmovcr8 abm sse4a misalignsse 3dnowprefetch osvw
bogomips        : 4623.17
TLB size        : 1024 4K pages
clflush size    : 64
cache_alignment : 64
address sizes   : 48 bits physical, 48 bits virtual
power management: ts ttp tm stc 100mhzsteps hwpstate [8]

[root@vps1 ~]# free
             total       used       free     shared    buffers     cached
Mem:      16410088   13537028    2873060          0    1994200    6224164
-/+ buffers/cache:    5318664   11091424
Swap:      8388600          0    8388600

[root@vps1 ~]# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/mapper/VolGroup00-LogVol00
                      1.8T   58G  1.7T   4% /
/dev/sda1             494M   75M  394M  16% /boot
tmpfs                 7.9G     0  7.9G   0% /dev/shm

The other is the same but windows.

~Michael

[beerandcandy]

the intel one has DDR3

[randomizer]

Aren't these still using HT2.0?

[Mandrake]

Aren't these still using HT2.0?

AFAIK they are now using HT3.0 for the links that connect the processors together, but the links to the chipset are still HT1.0 or 2.0 (as a limitation of the aging Nvidia chipset).

[arisythila]

yeah, the interconnects between the processors and memory are 3.0, but the connections between the north bridge and the CPU are still @ 2.0 unless they have changed something.