Shanghai Results (Opteron 2384, 2.7GHz)

Here are the results, i know i should include a source normally, but i cannot, because i dont want the site to get into any problems, any mod/admin can PM me though, but i dont want to throw such kind of info to the public

Bench: Sungard AA
AMD Opteron 2384: 249.6s
AMD Opteron 2356: 300.2s
Intel Xeon X5470: 180.1s
Intel Xeon L5430: 230.2s

Bench: Caselab Euler 3D
AMD Opteron 2384: 63.1s
AMD Opteron 2356: 84s
Intel Xeon X5470: 64.2s
Intel Xeon L5430: 66.9

Bench: Cinebench R10
AMD Opteron 2384: 19357
AMD Opteron 2356: 15372
Intel Xeon X5470: 24624
Intel Xeon L5430: 19997

Bench: FlamMap FSPRO
AMD Opteron 2384: 427s
AMD Opteron 2356: 512s
Intel Xeon X5470: 309s
Intel Xeon L5430: 376s

Bench: Half-Life2 Build Map
AMD Opteron 2384: 91s
AMD Opteron 2356: 109s
Intel Xeon X5470: 75s
Intel Xeon L5430: 87s

Bench: HWI MySQL - Concurrency 2
AMD Opteron 2384: 1608 trans/s
AMD Opteron 2356: 1498 trans/s
Intel Xeon X5470: 1475 trans/s
Intel Xeon L5430: 1455 trans/s

Bench: HWI MySQL - Concurrency 4
AMD Opteron 2384: 3072 trans/s
AMD Opteron 2356: 2948 trans/s
Intel Xeon X5470: 2753 trans/s
Intel Xeon L5430: 2741 trans/s

Bench: HWI MySQL - Concurrency 8
AMD Opteron 2384: 5646 trans/s
AMD Opteron 2356: 4944 trans/s
Intel Xeon X5470 4633 trans/s
Intel Xeon L5430: 4648 trans/s

Bench: HWI MySQL - Concurrency 16
AMD Opteron 2384: 5961 trans/s
AMD Opteron 2356: 4286 trans/s
Intel Xeon X5470 5846 trans/s
Intel Xeon L5430: 4766 trans/s

Bench: HWI MySQL - Concurrency 32
AMD Opteron 2384: 6123 trans/s
AMD Opteron 2356: 4970 trans/s
Intel Xeon X5470: 6046 trans/s
Intel Xeon L5430: 4838 trans/s

Bench: PovRay 3.7b21 - Chess 1280*1024
AMD Opteron 2384: 58.4s
AMD Opteron 2356: 69.4
Intel Xeon X5470 48.1s
Intel Xeon L5430: 59.1s

Bench: Sissoft Sandra 2009 - Intercore Bandwidth
AMD Opteron 2384: 7.46 Gbit/s
AMD Opteron 2356: 5.79 Gbit/s
Intel Xeon X5470: 25.7 Gbit/s
Intel Xeon L5430: 21.9 Gbit/s

Bench: Sissoft Sandra 2009 - Intercore Latency
AMD Opteron 2384: 107ns
AMD Opteron 2356: 164ns
Intel Xeon X5470 104ns
Intel Xeon L5430: 101ns

Bench: Sissoft Sandra 2009 - L1 latency
AMD Opteron 2384: 3clk
AMD Opteron 2356: 3clk
Intel Xeon X5470 3clk
Intel Xeon L5430: 3clk

Bench: Sissoft Sandra 2009 - L2 Latency
AMD Opteron 2384: 16clk
AMD Opteron 2356: 16clk
Intel Xeon X5470: 18clk
Intel Xeon L5430: 18clk

Bench: Sissoft Sandra 2009 - Memory Bandwidth
AMD Opteron 2384: 20.41 Gbit/s
AMD Opteron 2356: 17.75 Gbit/s
Intel Xeon X5470: 4.05 Gbit/s
Intel Xeon L5430: 4.05 Gbit/s

Bench: Sissoft Sandra 2009 - Memory Latency
AMD Opteron 2384: 105 ns
AMD Opteron 2356: 118 ns
Intel Xeon X5470: 126 ns
Intel Xeon L5430: 127 ns

Idle Power Concumption
AMD Opteron 2384: 100W
AMD Opteron 2356: 109W
Intel Xeon X5470: 170W
Intel Xeon L5430: 169W

Power Consumption (Cinenbench)
AMD Opteron 2384: 197W
AMD Opteron 2356: 226W
Intel Xeon X5470: 277W
Intel Xeon L5430: 227W

Power Consumption (HWI SQL CC 32)
AMD Opteron 2384: 143W
AMD Opteron 2356: 167W
Intel Xeon X5470: 206W
Intel Xeon L5430: 188W

Enjoy ;)

Time will tell I suppose just how accurate this is.

Info is good though, so thank you for posting.

Those aren't bad results at all. It is the MySQL champion in these tests. I think AMD delivered on their performance gain claim over current Opterons with room to spare. Those latencies dropped tremendously.

Just FYI:

AMD Opteron 2384: 2.7 GHz
AMD Opteron 2356: 2.3 GHz
Intel Xeon X5470: 3.33 GHz
Intel Xeon L5430: 2.66 GHz

For clock/clock comparisons: Optie 2384 vs Xeon L5430

all it needs now is a Chipset with HT 3.0 support :yepp:

If this is true then AMD did very well actually :)

The big question is: will it scale to more impressive clocks?

Power consumption is great.

hmmm really nice improvement

Memory Latency = L3 Latency + Memory Latency?
I mean can you somehow see if L3 got better latency on Shanghai vs. Barcelona?

L3 Latency = Intercore Latency?

Quote:

---

Originally Posted by Eson

Memory Latency = L3 Latency + Memory Latency?
I mean can you somehow see if L3 got better latency on Shanghai vs. Barcelona?

L3 Latency = Intercore Latency?

---

sorry, dont know the answer, i got these results throug some logical thinking and playing with URLs

In particular, the first few MySql tests are disappointing.

It would be much better though, if you would underclock the Shanghai to 2.3GHz, so the clock/clock comparison to Barc would be completely clear.

Quote:

---

Originally Posted by terrace215

In particular, the first few MySql tests are disappointing.

It would be much better though, if you would underclock the Shanghai to 2.3GHz, so the clock/clock comparison to Barc would be completely clear.

---

5430 to 5470 doesn't scale well, so i assume clock frequency doesnt matter that much?

Quote:

---

Bench: Sissoft Sandra 2009 - Memory Latency
AMD Opteron 2384: 105 ns
AMD Opteron 2356: 118 ns
Intel Xeon X5470: 126 ns
Intel Xeon L5430: 127 ns

---

The memory latency is just 2ns higher than the one for Dunnington's L3 array. :D

I did some numbers (BTW, power consumption is absolutely nice)

Quote:

---

Originally Posted by justthefax

Just FYI:

AMD Opteron 2384: 2.7 GHz
AMD Opteron 2356: 2.3 GHz
Intel Xeon X5470: 3.33 GHz
Intel Xeon L5430: 2.66 GHz

For clock/clock comparisons: Optie 2384 vs Xeon L5430

---

hmm with that in mind I think AMD has done a tremendous job with shanghai, but they will still definitely have to get their clockspeed up.

Wow.
Amd is back?

Yes indeedy... Shanghai is shaping up nicely.

Do you know if these were single or dual socket results ... I am assuming dual socket based on the numbers.

Jack

Quote:

---

Originally Posted by Eson

Memory Latency = L3 Latency + Memory Latency?
I mean can you somehow see if L3 got better latency on Shanghai vs. Barcelona?

L3 Latency = Intercore Latency?

---

I think the intercore latency would be a better clue, and yes, it looks like L3 improved.

Bench: Sissoft Sandra 2009 - Intercore Latency
AMD Opteron 2384: 107ns
AMD Opteron 2356: 164ns

This would imply improved L3 latency.

Jack

It's nice to see tangible improvements across the block for the transition to 45nm.

Quote:

---

Originally Posted by PetNorth

I did some numbers (BTW, power consumption is absolutely nice)

---

thx, pretty much proofs that the "20-25% more performance" is not clock for clock. :up:

anyway nice performance increasment for deneb.

btw. anyone wana normalize the opteron scores?

Quote:

---

Originally Posted by Hornet331

btw. anyone wana normalize the opteron scores?

---

Here we go:

these opertons are they still on the old HT ? like Barcelona is ?

Quote:

---

Originally Posted by justthefax

I'm doing them. Update in a few mins

---

thx much appreciated. :up:

Quote:

---

Originally Posted by demonkevy666

these opertons are they still on the old HT ? like Barcelona is ?

---

Yeah, the new HT is not supposed to be out till something like Q2 of next year. On the plus side, these are supposed to be drop-in upgrades for old Barcs.

I hope I didn't mess up the calculations! :D

But IPC doesn't seem to be improved much. Perf/watt is greatly improved though

I can't think of any way that performance would drop, though, so hopefully these are the minimum amount of improvements.

There's an issue with normalizing the scores though.


For example, see MySQL. It doesn't scale with clockspeed that much even with the Opteron. I suspect that interconnect matters more in this area.


And normalizing power load is also wrong, since other components don't scale down. What we can do though is compare power deltas for respective setups and work out relative power usage instead. :)

Quote:

---

Originally Posted by justthefax

I hope I didn't mess up the calculations! :D

But IPC doesn't seem to be improved much. Perf/watt is greatly improved though

I can't think of any way that performance would drop, though, so hopefully these are the minimum amount of improvements.

---

Just glancing, I did the quick compare with CB10 ... your numbers look sane.

However, look at the simple raw performance -- and if you are interested in DT, CB is very close to a Harpertown clock for clock as is the valve map compile ... these are DT like apps where it looks like AMD has pulled close to Penryn based cores.

This is a good achievement by any measure for just a shrink with tweaks.

Jack

Quote:

---

Originally Posted by justthefax

Here we go:

---

thx, so shanghai is in the up to 10% range, as expected. Means is a good shrink and power figures look also very good to. :yepp:

This should put deneb between kentsfield and yorkfield. If they price them right amd will have a good portofilo up to the midend segment. :up:

Interesting results, Looking pretty good. Be good to see what Deneb brings.

Hard to make IPC assumptions with thsoe results though for sure, since xeon only scales 16% in perf with 25% clock speed thanks to SQL and a few of the others.

anyone got a clue as what the L3 cache speed is set to I know old Barcelona only got 1400mhz-2400Mhz

Anyway, back on topic. VRAD and Cinebench, both of which JJ mentioned, seem to have benefited quite a lot from cache, especially just by raising L3.

VRAD: Yorkfield 6M led Agena by 16%, 12M led by 20. Now it leads by 5%.

CB: YF 6M led Agena by 11%, YF 12M by 13.5%. Now it has 3.3% and of course, 33Mhz handicap. :p:


This might be more optimistic for the AMD chips as they scale better in multi-CPU cases, but I don't think that's the case for 2P mostly. Xeon still does a great job there. And even if it did scale better on 2P for AMD, the desktop chips still have HT3.0 and probably higher L3/NB clocks.

Quote:

---

Originally Posted by Macadamia

No, all you did was astroturf and insult- on several discussions I might add. You don't seem to be interested in any smidgen of civilized discussion, so you might as well do yourself/us (mainly yourself) a favor and stop posting.


Anyway, back on topic. VRAD and Cinebench, both of which JJ mentioned, seem to have benefited quite a lot from cache, especially just by raising L3.

VRAD: Yorkfield 6M led Agena by 16%, 12M led by 20. Now it leads by 5%.

CB: YF 6M led Agena by 11%, YF 12M by 13.5%. Now it has 3.3% and of course, 33Mhz handicap. :p:


This might be more optimistic for the AMD chips as they scale better in multi-CPU cases, but I don't think that's the case for 2P mostly. Xeon still does a great job there. And even if it did scale better on 2P for AMD, the desktop chips still have HT3.0 and probably higher L3/NB clocks.

---

deneb 2.2ghz-2.4ghz. is what I've seen

Ht 3.1 will boost that (3.2ghz) like how HT 3.0 can get to 2.6ghz on some chips and boards.

Go more L3 and core tweaks

Quote:

---

Originally Posted by Hornet331

thx, so shanghai is in the up to 10% range, as expected. Means is a good shrink and power figures look also very good to. :yepp:

This should put deneb between kentsfield and yorkfield. If they price them right amd will have a good portofilo up to the midend segment. :up:

---

Yup. Basically if your AM2+ board is going to support Deneb it looks like the difference in performance with Yorkfield is not enough to be worth switching over given the cost advantage. My guess is for a new system unless you need Nehalem level power it's a tossup between 775 and AM3 now. This is certainly the most competitive AMD has been since late 2006.

disappointing performance but great power benefits.

Quote:

---

Originally Posted by justthefax

http://www.xtremesystems.org/forums/...1&d=1226458213

---

I bet it's better than that when at same clocks as power consumption never scales linearly.

Can you cut the whining and offtopic ?

Quote:

---

Originally Posted by AliG

hmm with that in mind I think AMD has done a tremendous job with shanghai, but they will still definitely have to get their clockspeed up.

---

this is the normal shanghai launch, q1 will give additional HE and SE cpu's

Quote:

---

Originally Posted by Eson

Memory Latency = L3 Latency + Memory Latency?
I mean can you somehow see if L3 got better latency on Shanghai vs. Barcelona?

L3 Latency = Intercore Latency?

---

l3 latency goes with NB speed, barcelone NB = 1.8, shangai has 2.0 and 2.2 depending on the model. The one tested has 2.2

Quote:

---

Originally Posted by justthefax

Here we go:

---

your calculations on power consumption are wrong, for now the 2,3ghz shangai is rated in the 75W ACP platform, so it will be less then 65nm but not that much.

also the MYSQL is not scaling linear, just check the differences between Xeon

Quote:

---

Originally Posted by gallag

disappointing performance but great power benefits.

---

lol nice crap post, check the benches again against 2,66ghz xeon and 2,7 shangai, i would say nice performance, yeah the 3,33ghz is better but so is it price and power and it has minimal volume against the other parts.

btw 45nm prices are also way lower then current barcelona. you'll get 2,5 for the same price as 2,3 so that also brings them back in competition.

Quote:

---

Originally Posted by LOE

will someone bother to figure how much is improvement in the most important aspect of servers - the power/performance ratio

---

About 30% ... 25% improvement in power and ~ 5% per clock for various items (but I am weary that the clock for clock compare is accurate -- not intentionally, justthefax simply scaled according to clock, however we know that it is not always the case that 10% clock bump translates to exactly 10% performance).

Quote:

---

Originally Posted by LOE

will someone bother to figure how much is improvement in the most important aspect of servers - the power/performance ratio

---

do you want the perf/watt benches? ;)


Performance/Watt Cinebench
AMD Opteron 2384: 98.3
AMD Opteron 2356: 68
Intel Xeon X5470: 88.9
Intel Xeon L5430: 88.1

Performance/Watt HWI MySQL
AMD Opteron 2384: 42.8 trans/W
AMD Opteron 2356: 29.8 trans/W
Intel Xeon X5470: 29.3 trans/W
Intel Xeon L5430: 25.7 trans/W

For servers the most important attribute is mostly power efficiency, that's is my experience from companies anyway. Maybe they have clocked (clocked parts on the processor) this shanghai in order to be as power efficient as possible?

Quote:

---

Originally Posted by duploxxx

lol nice crap post, check the benches again against 2,66ghz xeon and 2,7 shangai, i would say nice performance, yeah the 3,33ghz is better but so is it price and power and it has minimal volume against the other parts.

---

No, he is right. In the face of hype the results are disappointing. Some people took AMD's word (marketing!) at face value: 15-20% IPC gains, and even insulted the sceptics, I believe some people need to apologise. :p:

Expected IPC gains and great power consumption if those benches are to be trusted.

In this test-suite (workstation/sever workloads mostly) IPC is about equal to Penryn with better power consumption. :clap: For desktop, though, Penryn is still untouchable - no fb-dimms and no need for high bandwidth.

Quote:

---

Originally Posted by Jacky

In this test-suite (workstation/sever workloads mostly) IPC is about equal to Penryn with better power consumption. :clap: For desktop, though, Penryn is still untouchable - no fb-dimms and no need for high bandwidth.

---

dont forget the server parts of amd always have downclocked NB/L3 and lame memory

too bad there arent any desktop benches available :(

Quote:

---

Originally Posted by Bellisimo

dont forget the server parts of amd always have downclocked NB/L3 and lame memory

too bad there arent any desktop benches available :(

---

You have posted most of their results, maybe you could post their testbeds and configuration?
Is it really the norm that Intel's low power xeon draws 60W more than the 2356, I know there's a clock difference, but the difference in power draw is huge nonetheless.
I guess it's because K10's power saving features are superior + fb-dimm power draw?

Promising data, I don't mean to sound :banana::banana::banana::banana:y but I'd like to see the very benchmarks before I place a proper judgement. If it's anything like the op describes, it's a good step ahead for AMD.

I know this is not right but a fun compair only...

Opteron 2356/2384 2.3 to 2.7 = 0.4ghz more clock speed(gives about 19,54% more with some core improve ment to)
Xeon 2.66 to 3.33 = 0.67ghz more clock speed(gives about 15,95% more with some core improve ment to)

I wanted to compair how much 0.4ghz more clock speed on the Shanghai give vs Xeon in increase performence.

15.95/0.67 = 23,80
23,80*0.4= 9.52%

So 0.4ghz more clocks gives about 9.52% more performence in the tests on the Xeon
And 0.4ghz more clocks gives about 19.54% more performence in the test on the 2356/2384

Vs the old 65nm cpus..

But this is only stupid numbers.....they may not scale same also.

Sry for my bad english.

Quote:

---

Originally Posted by Jacky

You have posted most of their results, maybe you could post their testbeds and configuration?
Is it really the norm that Intel's low power xeon draws 60W more than the 2356, I know there's a clock difference, but the difference in power draw is huge nonetheless.
I guess it's because K10's power saving features are superior + fb-dimm power draw?

---

thats not accessible through URL tricks, sorry

Quote:

---

Originally Posted by Jacky

No, he is right. In the face of hype the results are disappointing. Some people took AMD's word (marketing!) at face value: 15-20% IPC gains, and even insulted the sceptics, I believe some people need to apologise. :p:

Expected IPC gains and great power consumption if those benches are to be trusted.

In this test-suite (workstation/sever workloads mostly) IPC is about equal to Penryn with better power consumption. :clap: For desktop, though, Penryn is still untouchable - no fb-dimms and no need for high bandwidth.

---

if you use hyping=truth yes then you are right, you will see in many released benchmarks (mainly spec since those are the preferred supplier benchmarks for both intel and amd) that the gains are fully within there marketing hype.

it is like it is already been for a long time, you buy certain servers for specific application types and group them, if you don't do that you have no idea about IT. Although Intel was able to regain some market share on that ground because of the much higher clock speed against barcelona, this is now gone again.

desktop is another discussion, at least they will be able to regain some market share and competition.

for desktop they

Quote:

---

Originally Posted by duploxxx

for desktop they

---

they WHAT? :p:

Anyway, @ Jacky, my stance on Deneb was that it would be win/lose partial WRT Yorkfield, and I'm probably going to maintain it. Those who asked for a complete win without a revamp of at least the caches would be delusional.

If what Johan of Anandtech referred to in his VM testing was indeed the Shanghai CPU, AMD was spot on in their promises there. As far as I remember, the 25+% promise was always focused on serverside, desktop never mentioned.

Quote:

---

Originally Posted by Slay0r

Promising data, I don't mean to sound :banana::banana::banana::banana:y but I'd like to see the very benchmarks before I place a proper judgement. If it's anything like the op describes, it's a good step ahead for AMD.

---

actually these are very benchmarks, i can still access the pictures through url :)

Quote:

---

Originally Posted by JumpingJack

About 30% ... 25% improvement in power and ~ 5% per clock for various items (but I am weary that the clock for clock compare is accurate -- not intentionally, justthefax simply scaled according to clock, however we know that it is not always the case that 10% clock bump translates to exactly 10% performance).

---

JJ is right.

Even though it's very sweet of justhefax to "normalize" the results, his calculations say exactly nothing.

A calculation that would come closer, is to first find out the scaling co-efficient for those benchmarks with a "normal" Agena (compare 2.3 to 2.7 performance to see how well it scales) and then use those coefficients to track back how the shanghai would do at 2.3 Ghz.

All in all this looks to be very promising. We allready knew amd would not get performance crown any time soon, but it seems that at this point, they could be taking the perf/watt crown back from intel.

Ran a few of those shanghai benchmarks here at 2.3GHz with mem at 667MHz (1) and 2.7GHz at 800MHz(2). I repeated each benchmark 5 times and the derivation stays sub 2% for all results.
http://www.abload.de/img/shanghai-projectionp1s0.jpg
Based on that I estimated how that 2356 barcelona would operate at 2.7GHz with mem at 800MHz.
Last row shows improvement of the 2384 over the estimated 2356 at the same clocks. I assumed both opterons ran with 2GHz NB speed.
Don't know where to get those other benchmarks, expecialy mysql would be interesting.

this should be the test config

Intel: Intel S5000PSL, 8 GB DDR2-800 FBDIMM (8 modules). AMD: ASUS KFSN5-D, 8 GB DDR2-800 (8 modules). both: Samsung Spinpoint F1 1TB, Intel X25M 80GB SSD (database benchmark only), Cooler Master iGreen 850W

Quote:

---

Originally Posted by Jacky

No, he is right. In the face of hype the results are disappointing. Some people took AMD's word (marketing!) at face value: 15-20% IPC gains, and even insulted the sceptics, I believe some people need to apologise. :p:

Expected IPC gains and great power consumption if those benches are to be trusted.

In this test-suite (workstation/sever workloads mostly) IPC is about equal to Penryn with better power consumption. :clap: For desktop, though, Penryn is still untouchable - no fb-dimms and no need for high bandwidth.

---

No he is not right. There is a reason why I included here http://www.xtremesystems.org/forums/...7&postcount=14 Xeon 3.33 Vs. Xeon 2.66. Notice that there is only 16% improvement when clock difference is 25%. So, you can´t do an accurate estimation between 2.3 Barcelona and a simulated 2.3 Shanghai as justthefax did -thanks anyways- (17% less clock, 17% less perf).

So if Barcelona perf scale is similar to what we see with Xeon 2.66 - 3.33, then I guess there is an ~8-10% average improvement clock for clock between Barcelona and Shanghai in these benchs (and except MySQL virtualization or perhaps Sunguard, they aren't server workloads at all). And that perfomance improvement is just fine, keeping in mine we are only talking about a new process with a few tweaks, and not about a new arch.

Of course, in the other hand, power consumption improvement is superb.

Quote:

---

Originally Posted by Miss Banana

JJ is right.

Even though it's very sweet of justhefax to "normalize" the results, his calculations say exactly nothing.

---

JJ stated they are not accurate, I concur. Saying they tell us nothing is exaggeration.
It's true that extrapolation from those numbers underestimates K10.5 performance slightly, but the numbers do not lie, it's just a shrink with very good power consumption.

Quote:

---

A calculation that would come closer, is to first find out the scaling co-efficient for those benchmarks with a "normal" Agena (compare 2.3 to 2.7 performance to see how well it scales) and then use those coefficients to track back how the shanghai would do at 2.3 Ghz.

---

I have an idea. Let's compare intel scaling numbers, which are available from the same source. Intel and AMD have the same clock-scaling, Intel scales less if limited by bandwidth in some tests, but on the whole they're similiar - I've seen the clock-scaling tests done.
Unsurprisingly cinebench and povray scale almost perfectly with a 25% clock-change on Intel's crippled FSB, FlamMap FSPRO and Sungard AA scale nicely plus/minus a couple percent.
So we can use those benches to asses K10 vs K10.5, because the clock difference is only ~17% and it should show perfect scaling.
In those 4 tests K10.5 is on average 3-4% faster, it's a die shrink QED.

EDIT:

Quote:

---

Originally Posted by PetNorth

So if Barcelona perf scale is similar to what we see with Xeon 2.66 - 3.33, then I guess there is an ~8-10% average improvement clock for clock between Barcelona and Shanghai in these benchs

---

I'm fine with that, I actually expect something like that. 8-10% seems healthy for a die shrink (penryn is in that range too, probably 8% or slightly lower).

EDIT2:

Quote:

---

Originally Posted by Bellisimo

Intel: Intel S5000PSL, 8 GB DDR2-800 FBDIMM (8 modules).

---

I'm no expert, correct me if I'm wrong but isn't this intentional crippling of the Intel system? Doesn't every single fb-dimm controller consume power? So if they used 2*4GB they could have reduced the fb-dimm consumption by up to 75%?

EDIT3:

Quote:

---

Originally Posted by duploxxx

it is like it is already been for a long time, you buy certain servers for specific application types and group them, if you don't do that you have no idea about IT.

---

Yes, you do, but the overall success of an arch can be predicted more easily when you compare average IPC change at the same power draw, clock scaling, die size and yields.

Quote:

---

Originally Posted by duploxxx

Although Intel was able to regain some market share on that ground because of the much higher clock speed against barcelona, this is now gone again.

---

Personally I believe it's going to be a tough time for AMD because of i7, it will eat their only high ASP market (Server, HPC) and push down the prices of the already high-yielding (smaller die than K10.5) and well scaling (clocks!) penryns. But that's just an economic prediction, I don't want to steal their thunder. Good job AMD.

Quote:

---

Originally Posted by Jacky

JJ stated they are not accurate, I concur. Saying they tell us nothing is exaggeration.

---

His calculations tell us nothing because they are grossly inaccurate.
How inaccurate do you want it to be, before you call such calculations useless?

For me being off +- 7% when even one % makes a big difference is no longer in the realm of useful in any way, but that's just me.

Link

Quote:

---

IBM posts leadership 4-processor blade score on SPECjbb2005 benchmark
November 13, 2008 ... IBM® BladeCenter® LS42 using IBM JavaTM6 Runtime Environment,
achieved a leadership 4-processor blade result of 721,843 SPECjbb2005® business operations
per second (SPECjbb2005 bops) and 180,461 SPECjbb2005 bops/JVM, running SPECjbb2005
(Java Business Benchmark), SPEC’s benchmark for evaluating the performance of servers
running typical Java applications.
The LS42 was configured with the AMD Opteron™ Model 8384 quad-core processor at 2.7GHz
with 2MB L2 cache and 6MB L3 cache (4 chips/16 cores/4 cores per chip), 64GB of memory, one
36.4GB disk drive, and IBM Java 6 (using a 1875MB heap), and Microsoft® Windows® Server
2008 Enterprise x64 Edition. (1)
The LS42’s score demonstrates the highest performance achieved to date by a 4-processor
blade server, surpassing the score of 383,456 SPECjbb2005 bops of the Dell PowerEdge M905,
which ran BEA JRockit® 6.0, and used the AMD Opteron Model 8356 quad-core processor at
2.3GHz (4 chips/16 cores/4 cores per chip). (2)
The LS42’s score also handily beats—by 42%—the 508,240 SPECjbb2005 bops of the Dell
PowerEdge R900, which ran BEA JRockit 6.0, and used the Intel® Xeon® Processor X7460 at
2.66GHz (4 chips/24 cores/6 cores per chip). (3)
BladeCenter LS42 blade servers, coupled with the BladeCenter chassis, deliver advanced
application serving with performance, power efficiency, and scalability ideal for enterprise
environments.
Results referenced are current as of November 13, 2008. The SPECjbb2005 results have been
submitted to SPEC for review. Upon successful review, the result will be posted at www.spec.org,
which contains a complete list of published SPECjbb2005 results.
(1) The LS42 model using the AMD Opteron Model 8384 quad-core processor is planned to be
generally available on November 30, 2008.
(2) Dell PowerEdge M905: 383,456 SPECjbb2005 bops and 95,854 SPECjbb2005 bops/JVM,
using four AMD Opteron 8356 quad-core processors at 2.3GHz (4 chips/16 cores/4 cores per
chip), 32GB of memory, one 36GB disk drive, and BEA JRockit 6.0 P27.5.0. The comparison is
based on Dell’s best SPECjbb2005 score for a 4-processor blade server published at SPEC as of
November 13, 2008.
(3) Dell PowerEdge R900: 508,240 SPECjbb2005 bops and 127,060 SPECjbb2005 bops/JVM,
using four Intel Xeon Processor X7460 at 2.66GHz (4 chips/24 cores/6 cores per chip), 64GB of
memory, two 36GB disk drives, and Oracle JRockit 6.0 P27.5.0. The comparison is based on
Dell’s best SPECjbb2005 score for a 4-processor server published at SPEC as of November 13,
2008.View all published results at www.spec.org/jbb2005/results/jbb2005.html
IBM and System x are trademarks or registered trademarks of IBM Corporation.
BEA JRockit is a registered trademark of BEA Systems, Inc.
Intel and Xeon are registered trademarks of Intel Corporation.
Java and all Java-based trademarks are trademarks of Sun Microsystems, Inc., in the United
States, other countries, or both.
Microsoft and Windows are registered trademarks of Microsoft Corporation.
SPEC and SPECjbb2005 are trademarks or registered trademarks of Standard Performance
Evaluation Corporation (SPEC).
All other company/product names and service marks may be trademarks or registered
trademarks of their respective companies.

---

16 Shanghai cores @2.7Ghz are 42% faster on SPECjbb2005 benchmark than 24 Dunnington cores running at 2.66Ghz...With perfect frequency scaling,intel would need 24 cores @ 3.7Ghz to just match teh Shaghai 4P server....

Who cares?Are you kidding me??Well maybe "Xtreme OCer" doesn't...
These are server CPUs and in this particular Java benchmark they just scream.60% faster per clock than Barcelona which was beat by Dunnington.Now it is reversed again.

Quote:

---

Originally Posted by informal

Link



16 Shanghai cores @2.7Ghz are 42% faster on SPECjbb2005 benchmark than 24 Dunnington cores running at 2.66Ghz...With perfect frequency scaling,intel would need 24 cores @ 3.7Ghz to just match teh Shaghai 4P server....

---

Im sure you could also see the software differences aswell.

Note the speed of the Dell Opteron....:yepp:

Its funny the 2.7Ghz 45nm is 75% faster than the 2.3Ghz 65nm...

Sounds like the new IBM java run-time is either really good, or IBM has found an optimization trick with SPECjbb2005 much like Sun's compiler did with SPECfp2000.

Quote:

---

Originally Posted by Miss Banana

His calculations tell us nothing because they are grossly inaccurate.
How inaccurate do you want it to be, before you call such calculations useless?

For me being off +- 7% when even one % makes a big difference is no longer in the realm of useful in any way, but that's just me.

---

Yeah just ignore my post and the numbers I have provided to back up at least 4 of them, whatever. :D

I have to admit, the new 45nm chips look good. But what bout Core I7 Xeon based chips. That will be the true test for AMD in the server market.

Quote:

---

Originally Posted by Shintai

Im sure you could also see the software differences aswell.

Note the speed of the Dell Opteron....:yepp:

Its funny the 2.7Ghz 45nm is 75% faster than the 2.3Ghz 65nm...

---

Funny? the difference in scores shows an 88% improvement for ~18% more clock speed, implying thr 45nm parts offer more than just improved speed.

Quote:

---

Originally Posted by ryboto

Funny? the difference in scores shows an 88% improvement for ~18% more clock speed, implying thr 45nm parts offer more than just improved speed.

---

No...you should look closer ;)

Quote:

---

Originally Posted by vengance_01

I have to admit, the new 45nm chips look good. But what bout Core I7 Xeon based chips. That will be the true test for AMD in the server market.

---

there are no xeon based i7 at the moment, intel choose to introduce desktop part first, amd choose to introduce the server part first (which looks like smarter move the me financial-wise)
but maybe AMD could only do this because only the cpu's had to be validated, not the intire platform

Quote:

---

Originally Posted by Shintai

No...you should look closer ;)

---

Dell's system used 2 harddrives? :rolleyes:

Dell uses the Oracle JRockit 6.0 P27.5.0 + unknow OS
IBM uses the BEA JRockit 6.0 P27.5.0. + windows server 2008 x64

New spec submissions are in for Opteron 2.7Ghz (2384 and 8384).For now only the int_rate and fp_rate:
2384:
http://spec.org/cpu2006/results/res2...024-05683.html
http://spec.org/cpu2006/results/res2...024-05684.html

8384:
http://spec.org/cpu2006/results/res2...024-05685.html
http://spec.org/cpu2006/results/res2...024-05686.html

Before shanghai,top scores for AMD in these two tests were held by 2360 Opteron model.From my early calculations and clock normalization(8% difference between 2.5 Barc and 2.7Ghz Shanghai;best submitted scores for Barcelona are used;2P scores are used) you can see that the 45nm part is ~18% faster per clock in int_rate and ~20% per clock fp_rate than Barcelona.
Using the 8xxx series scores,in int_rate the lead is ~19% ,while in fp_rate the lead is ~17% (per clock of course).

Quote:

---

Originally Posted by Shintai

No...you should look closer ;)

---

i hate smiley's. What are you implying? The only thing I can think of is that the IBM system is using SPEC's software, and not Java's?

edit: Bellisimo got to the question before me, ignore mine.

Quote:

---

Originally Posted by Bellisimo

Dell's system used 2 harddrives? :rolleyes:

Dell uses the Oracle JRockit 6.0 P27.5.0 + unknow OS
IBM uses the BEA JRockit 6.0 P27.5.0. + windows server 2008 x64

---

No...

Quote:

---

The LS42 was configured with the AMD Opteron™ Model 8384 quad-core processor at 2.7GHz
with 2MB L2 cache and 6MB L3 cache (4 chips/16 cores/4 cores per chip), 64GB of memory, one
36.4GB disk drive, and IBM Java 6 (using a 1875MB heap), and Microsoft® Windows® Server
2008 Enterprise x64 Edition. (1)

---

Dell uses BEA/Oracle Rockit 6 JVM + Windows 2003 SP1.
IBM uses IBM Java 6 JVM + Windows 2008 SP1.

I just went through the numbers for 24 cores Dunnington @ 2.66Ghz systems for int_rate/fp_rate and compared them to 16 cores Shanghai server @ 2.7Ghz.The Dunnington systems leads by 16% in int_rate test and falls behind by 44% in fp_rate test(24 versus 16 cores,clocks are essentially the same with 1% difference).

Quote:

---

Originally Posted by informal

New spec submissions are in for Opteron 2.7Ghz (2384 and 8384).For now only the int_rate and fp_rate:
2384:
http://spec.org/cpu2006/results/res2...024-05683.html
http://spec.org/cpu2006/results/res2...024-05684.html

8384:
http://spec.org/cpu2006/results/res2...024-05685.html
http://spec.org/cpu2006/results/res2...024-05686.html

Before shanghai,top scores for AMD in these two tests were held by 2360 Opteron model.From my early calculations and clock normalization(8% difference between 2.5 Barc and 2.7Ghz Shanghai;best submitted scores for Barcelona are used;2P scores are used) you can see that the 45nm part is ~18% faster per clock in int_rate and ~20% per clock fp_rate than Barcelona.
Using the 8xxx series scores,in int_rate the lead is ~19% ,while in fp_rate the lead is ~17% (per clock of course).

---

meh... only rate benches (only really relevant for high throughput apps), are there any none rate (SPECfp2006/SPECint2006) available?

Patience young Skywalker :),they will be available soon.The rate scores are also valid for clock/clock evaluation.

It's interesting to see how well these things scale.I'm willing to bet that 6 core Istanbul will scale great too so we can "predict" the scores using present Shanghai numbers(should scale pretty much in line with core numbers and if it does,it will simply destroy dunnington which is already pretty much matched with Shanghai ).

Quote:

---

Originally Posted by duploxxx

lol nice crap post, check the benches again against 2,66ghz xeon and 2,7 shangai, i would say nice performance, yeah the 3,33ghz is better but so is it price and power and it has minimal volume against the other parts.

btw 45nm prices are also way lower then current barcelona. you'll get 2,5 for the same price as 2,3 so that also brings them back in competition.

---

I said that I found the performance disappointing but the power improvements are excellent, How it it a crap post? Are you happy with the ipc improvement? What was your guess for ipc improvement and what is it in reality? Does every opinion have to be 100% positive for AMD for you not to be insulting?

Quote:

---

Originally Posted by informal

New spec submissions are in for Opteron 2.7Ghz (2384 and 8384).For now only the int_rate and fp_rate:
2384:
http://spec.org/cpu2006/results/res2...024-05683.html
http://spec.org/cpu2006/results/res2...024-05684.html

8384:
http://spec.org/cpu2006/results/res2...024-05685.html
http://spec.org/cpu2006/results/res2...024-05686.html

Before shanghai,top scores for AMD in these two tests were held by 2360 Opteron model.From my early calculations and clock normalization(8% difference between 2.5 Barc and 2.7Ghz Shanghai;best submitted scores for Barcelona are used;2P scores are used) you can see that the 45nm part is ~18% faster per clock in int_rate and ~20% per clock fp_rate than Barcelona.
Using the 8xxx series scores,in int_rate the lead is ~19% ,while in fp_rate the lead is ~17% (per clock of course).

---

Quite amazing.

Quote:

---

Originally Posted by Bellisimo

there are no xeon based i7 at the moment, intel choose to introduce desktop part first, amd choose to introduce the server part first (which looks like smarter move the me financial-wise)
but maybe AMD could only do this because only the cpu's had to be validated, not the intire platform

---

But they will next year and I am not sure AMD will have anything by then.

Quote:

---

Originally Posted by vengance_01

But they will next year and I am not sure AMD will have anything by then.

---

Not really. Then again, Intel does not have now. Though, if I was getting server hardware, I'd wait until Nehalem for servers and then get the AMD if the price is low enough. Sure Intel would be fine aswell, but I doubt that they need to start fiddling with price to beat AMD's offerings.

Looks like the int results are about what was expected (adjusting for 2.7 vs 2.8), while the fp results fall short a bit. Looking at the 2P projections from back in february, here:


http://www.xtremesystems.org/forums/...d.php?t=178416

In short: Nehalem is substantially faster, as expected.

For even more fun, note that a *single* (1P) Nehalem can score 110 in spec_int_rate_base. :)

Quote:

---

Originally Posted by terrace215

Looks like the int results are about what was expected (adjusting for 2.7 vs 2.8), while the fp results fall short a bit. Looking at the 2P projections from back in february, here:


http://www.xtremesystems.org/forums/...d.php?t=178416

In short: Nehalem is substantially faster, as expected.

---

lets remeber shintai's wise words

http://www.xtremesystems.org/forums/...35&postcount=3

Quote:

---

Originally Posted by Shintai

People should just remember rates benchmarks are very deceptive. Nehalem basicly just wins this so much due to trichannel and CSI.

---

Quote:

---

Originally Posted by terrace215

Looks like the int results are about what was expected (adjusting for 2.7 vs 2.8), while the fp results fall short a bit. Looking at the 2P projections from back in february, here:


http://www.xtremesystems.org/forums/...d.php?t=178416

In short: Nehalem is substantially faster, as expected.

For even more fun, note that a *single* (1P) Nehalem can score 110 in spec_int_rate_base. :)

---

Figure it...

Suddenly, projected scores are valid...


LOL

Cool thanks for the post

Quote:

---

Originally Posted by Shintai

No...



Dell uses BEA/Oracle Rockit 6 JVM + Windows 2003 SP1.
IBM uses IBM Java 6 JVM + Windows 2008 SP1.

---

Those poor guys at Dell, maybe they couldn't afford a Windows 2008 license and so their Intel machine was crippled? And IBM refused to sell their JVM to Dell coz they're so meeeeean? :rofl:

And did you hear of the W2k8 SP1 security fix that cripples Dunningtons by turning their L3, L2 AND L1 caches off?

The real surprise is how well Intel scores even with its hands tied! Right Shintai? :clap:

Quote:

---

Originally Posted by justthefax

Those poor guys at Dell, maybe they couldn't afford a Windows 2008 license and so their Intel machine was crippled? And IBM refused to sell their JVM to Dell coz they're so meeeeean? :rofl:

And did you hear of the W2k8 SP1 security fix that cripples Dunningtons by turning their L3, L2 AND L1 caches off?

The real surprise is how well Intel scores even with its hands tied! Right Shintai? :clap:

---

Seriously? What a monumental failure from Microsoft. How is that even possible?

Quote:

---

Originally Posted by justthefax

Those poor guys at Dell, maybe they couldn't afford a Windows 2008 license and so their Intel machine was crippled? And IBM refused to sell their JVM to Dell coz they're so meeeeean? :rofl:

And did you hear of the W2k8 SP1 security fix that cripples Dunningtons by turning their L3, L2 AND L1 caches off?

The real surprise is how well Intel scores even with its hands tied! Right Shintai? :clap:

---

Acutaly he has a point, those two test run on different OSs with different JVM on different hardware.

Hell is there anything that they even have in common? :p:

Quote:

---

Originally Posted by Hornet331

Acutaly he has a point, those two test run on different OSs with different JVM on different hardware.

Hell is there anything that they even have in common? :p:

---

Size?? :p:

Quote:

---

Originally Posted by justthefax

Those poor guys at Dell, maybe they couldn't afford a Windows 2008 license and so their Intel machine was crippled? And IBM refused to sell their JVM to Dell coz they're so meeeeean? :rofl:

And did you hear of the W2k8 SP1 security fix that cripples Dunningtons by turning their L3, L2 AND L1 caches off?

The real surprise is how well Intel scores even with its hands tied! Right Shintai? :clap:

---

What? That is total BS, 2008 does not turn off the cache.

Quit making things up.

Quote:

---

Originally Posted by Loudwater

What? That is total BS, 2008 does not turn off the cache.

Quit making things up.

---

It's called sarcasm if you didn't catch it in his post...

btw, the numbers of my post are correct, but i'am not gonna help you find the review

I will back up Bellisimo and post the link for the review :D :
http://www.hardware.info/en-US/produ...84WAL4DGIWOF/#

BTW,nice find Bellisimo :)

Quote:

---

Originally Posted by informal

I will back up Bellisimo and post the link for the review :D :
http://www.hardware.info/en-US/produ...84WAL4DGIWOF/#

BTW,nice find Bellisimo :)

---

hehe :)
correct, it was quite fun playing with the URL's figuring out how to get to the benches :)
hope you guys were satisfied with results 2 days prior to launch

Quote:

---

Originally Posted by Bellisimo

hehe :)
correct, it was quite fun playing with the URL's figuring out how to get to the benches :)
hope you guys were satisfied with results 2 days prior to launch

---

It did smack a bit of genius:D Fair play.

Yeah great job man :D.I never came across that site before ,thanks for sharing it's a great site!

Quote:

---

Originally Posted by Bellisimo

dont forget the server parts of amd always have downclocked NB/L3 and lame memory

too bad there arent any desktop benches available :(

---

true, phenom 9850 has higher HT/memory
opterons has lower speed why? its needs to have lower power usage to be used in 1U/2U servers, who wants to run a 140W Phenom 9950 in their server, it will overheat easy

given the fact that the new deneb phenoms are 125W vs 75W on opterons says something that the deneb is faster, due to its using more power

got it?

@Bellisimo, a x264 test would be nice :)

Quote:

---

Originally Posted by wiak

true, phenom 9850 has higher HT/memory
opterons has lower speed why? its needs to have lower power usage to be used in 1U/2U servers, who wants to run a 140W Phenom 9950 in their server, it will overheat easy

given the fact that the new deneb phenoms are 125W vs 75W on opterons says something that the deneb is faster, due to its using more power

got it?

@Bellisimo, a x264 test would be nice :)

---

No thats bull. The reason is chipsets and such. And its 125W TDP desktops vs 75W ACP Opterons. The 75W ACP is 105W TDP for 300mhz less. And that fits quite well with the later 2.8Ghz Deneb at 95W.

Quote:

---

Originally Posted by Shintai

No thats bull. The reason is chipsets and such. And its 125W TDP desktops vs 75W ACP Opterons. The 75W ACP is 105W TDP for 300mhz less. And that fits quite well with the later 2.8Ghz Deneb at 95W.

---

You could of have said that more... nice though.

Ive to agree on power consumption, but I dont see how chipsets limits the NB clocks at all.

http://img408.imageshack.us/img408/2...7630432jw3.png
http://www.amdzone.com/phpbb3/viewto...?f=52&t=135802