TechReport on Barcelona

[JumpingJack]

No, not of course. The L3 adds a L3 eviction and write back penalty on an L3 miss (in addition to L1 and L2), but the improved DRAM prefetchers, buffers and crossbar should be making up for it. It looks to me like 40 cpu clock cycles are being eaten up.

There is a penalty for having L3 cache if the miss goes straight through and requires access to memory, no way around that.... you will spend the extra time snooping the L3 that normally would not have occured if L3 was not there.... so it gets in the way a little.... the net benefit should be positive though so long as the compounded latency is much less than a call to main memory for all the hits.

[MaxxxRacer]

Well I must say, I'm not entirely surprised, but still disapointed that AMD couldnt come through with barcelona. I was hoping for atleast 1:1 performace clock for clock with core2, but that doesnt seem to be the case.

Unified core and lower power consumption will not win my heart when I design a new system.

[savantu]

Not correct.

From what I understand, what Jacky was getting at with the cache latencies being reduced at a high clockspeed is actually correct. When going to RAM, the processor has to go through all the cache levels of the processor, including the 'slow' L3. The latency of the L3 cache is proportional to a speed (not sure on the terminology) that is increased when the clockspeed is raised. Therefore, with figures quote on the Tech Report site, the cache latecies are 23ns on the slower 2.0Ghz chip and 19ns on the faster (2.5Ghz) chip.

Therefore, increasing the clockspeed could result in what would seem a higher IPC.

I think you need to read more about CPUs , somebody explained it nicely : http://www.xtremesystems.org/forums/...&postcount=107

I'll add this because hollo fails to account for the RAM latency :

Assuming RAM latency 40ns

At 2GHz we have total latency 63ns or 126 cycles.
At 2.5GHz we have a total latency of 59 or 147.5 cycles.

All of a sudden a cache miss isn't as minor as some think.You spend 17% more clock cycles time waiting for memory.As a result IPC is lower.

Also, TG Daily slamming AMD with a pro-Intel bias? Well I'll be damned...

EDIT: Also, after more careful reading,

And correct me if i'm wrong, but the motherboard doesn't support split power planes, so the memory controller would have been running at 1.6Ghz. Based on that fact alone, we should see improved performance from the new mobos.

The servers tested were supplied by AMD and used split power planes mobos.

[JVguest]

Savantu, how is 40 ns out of 59 cycles only 17% slower? Anyway, typically L2 latency is 15 cpu cycles on core 2/K8 while going to main memory is 150+ cycles. It's 10 times slower going to main memory, fortunately with branch prediction and caches it doesn't happen often. Main memory access can be generalized as 10x slower than cache. A cache miss always causes a major slowdown, but AMD minimizes that with IMC.

[savantu]

Savantu, how is 40 ns out of 59 cycles only 17% slower? Anyway, typically L2 latency is 15 cpu cycles on core 2/K8 while going to main memory is 150+ cycles. It's 10 times slower going to main memory, fortunately with branch prediction and caches it doesn't happen often. Main memory access can be generalized as 10x slower than cache. A cache miss always causes a major slowdown, but AMD minimizes that with IMC.

I'm not sure we're talking about the same thing ; I'm not comparing memory vs. cache access time , but how costly is a cache miss in clock cycles while running at different frequencies , hence lower IPC at higher frequencies.

[BeardyMan]

Well I must say, I'm not entirely surprised, but still disapointed that AMD couldnt come through with barcelona. I was hoping for atleast 1:1 performace clock for clock with core2, but that doesnt seem to be the case.

Unified core and lower power consumption will not win my heart when I design a new system.

It is the case actually, it's a lower clocked cpu vs a higher clockec cpu and it wins often so...

[accord99]

It is the case actually, it's a lower clocked cpu vs a higher clockec cpu and it wins often so...

Yes, the 2.33GHz Clovertown does win often against the 2.5GHz Barcelona in Techreport's review.

[saaya]

nice, techrepot really does nice articles!

[BeardyMan]

Yes, the 2.33GHz Clovertown does win often against the 2.5GHz Barcelona in Techreport's review.

Opteron 2350 2.0Ghz vs Xeon L5335 2.0Ghz

1)Sandra Cache and memory bandwith
*11534 vs 5179
higher is better
Opteron 2350 2.0Ghz


2)Cpuz memory access latency
*91 vs 95
lower is better
Opteron 2350 2.0Ghz


3)SPECjbb
*88949 vs 87099
higher is better
Opteron 2350 2.0Ghz


4)Valve VRAD map build time
*121 vs 107
lower is better
Xeon L5335 2.0Ghz

5)Cinebench
*12623 vs 14129
higher is better
Xeon L5335 2.0Ghz

6)POV ray rendering
*77 vs 80
lower is better
Opteron 2350 2.0Ghz


7)Merimatch benchmark 8 threads
*372 vs 379
lower is better
Opteron 2350 2.0Ghz


8)Folding @ home
* 4 wins vs 2
Opteron 2350 2.0Ghz


9)The panorama factory
*23.05 vs 20.41
lower is better
Xeon L5335 2.0Ghz

10)PicColor
8.09 vs 10.11
higher is better
Xeon L5335 2.0Ghz

11)windows media encoder
*543 vs 510
lower is better
Xeon L5335 2.0Ghz

12)Sisoft sandra
*296230 VS 335126
higher is better
Xeon L5335 2.0Ghz

13)power consumption
* 2 wins vs 0
Opteron 2350 2.0Ghz


Is it me or is 7 more then 6...

[Shintai]

I wouldnt even look at 1+2+12.

Also for folding, almost all WUs are gromacs. The bench also takes old outdated WUs in place without SSE support. x87 is dead in 64bit too. Kinda like you also see a K8 beats a K10 in x87 tho 2.5 vs 2.6Ghz.

But all the rest, yes.

[accord99]

Opteron 2350 2.0Ghz vs Xeon L5335 2.0Ghz

1)Sandra Cache and memory bandwith
2)Cpuz memory access latency

These are characteristics of the platform, not benchmarks of performance.

6)POV ray rendering
*77 vs 80
lower is better
Opteron 2350 2.0Ghz

There are two POV tests. The L5335 is faster in the second. I'd call this a draw.

8)Folding @ home
* 4 wins vs 2
Opteron 2350 2.0Ghz

There are only 4 tests, with L5335 winning in the most important gromacs cores. Considering Tinkers and Amber units are hardly processed anymore the averaging score is suspect and it's a draw at best for the Opteron.

12)Sisoft sandra
*296230 VS 335126
higher is better
Xeon L5335 2.0Ghz

I wouldn't even bother with synthetic tests like these.

13)power consumption
* 2 wins vs 0
Opteron 2350 2.0Ghz

Not a test of speed.

By my count, the Opteron 2350 wins 2 tests, the Xeon L5335 wins 6 and there's a draw in 2 with the Sisoft test excluded.

[BeardyMan]

I wouldnt even look at 1+2+12.

Also for folding, almost all WUs are gromacs. The bench also takes old outdated WUs in place without SSE support. x87 is dead in 64bit too.

Shintai it seems you and other don't get the point
It's not about it wins there and there and it gets pwned there and there..

it's about AMD being back with a nice and good competitive cpu.
Many l seem to trash it while i just prooved that it's uncalled for

If but i doub't it really, K10 scales on a different way other cpu's do, then i see them take more wins at 3.0ghz... but that can be wishfull thinking

K8 beating K10 is pretty sad yes, but aren't there test were P4 beats a C2D as well ....

[BeardyMan]

T

Not a test of speed.

LOL weight your words carefully
Power consumption is very very important in teh server market.. or do you think electricty is free

[BeardyMan]

SmartASS it doesn't matter really
It seems youre reading selective and only what you wanne read.

This is what i stated in a previous post

It's not about it wins there and there and it gets pwned there and there..

it's about AMD being back with a nice and good competitive cpu.
Many seem to trash it while i just prooved that it's uncalled for


period

[savantu]

SmartASS - the sandra score shows actual performance the same way memory bandwidth does, even less

Umh no...

It shows what the chip is actually capable of if its units are used to full extent.

[alayashu]

How nice and competitive is this alpha BIOS?

not at all

[JVguest]

How nice and competitive is this?

That's one reason I think the Anand preview didn't really show the full performance of the K10. It also doesn't seem that AMD will be shipping a lot of these systems now if the AMD supplied Anandtech system has such problems. AMD was rushed to get barc out the door by their stated release date, and they didn't quite get it ready. Hopefully, yet again AMD will just manage to hang on.

[GoThr3k]

after reading this thread with a lot of crap in it,thx to savuntu and mr smartass...

juste making a few point

Server <-> desktop
Desktop parts will be faster,do you really think with wider prefetch and 4 cores DDR2 667 ECC can provide enough bandwidth?

No clear winner
in the benches i've read,there is no clear winner here
barcalona sometimes beats a 2.33 clover, and sometimes it's the other way around.

AMD is back
looking at the benches,K10 isn't as bad as some people had wished for..
It's actually quite good imo
finally bringing quadcore to the scene,which is really important in the server market. They don't need high clocked parts,cause who is insane enough to buy the most expensive?
they just need fresh money,new better products provides this
2.5 ghz will be sufficient if you ask me, and those parts are coming real soon (i hope )

[Shintai]

Actually Intel is only stronger in the desktop. K8 already do quite well against Core 2 in the server space. In the desktop it gets molested by the Core 2.

So if anything, its only gonna be worse when the Core 2 is released from FB-DIMMs and the 5000 platform. So if you are hoping for something..dont..it will just backfire as the K10 expectations alot had.

But again, highend is low volume.

[GoThr3k]

Actually Intel is only stronger in the desktop. K8 already do quite well against Core 2 in the server space. In the desktop it gets molested by the Core 2.

So if anything, its only gonna be worse when the Core 2 is released from FB-DIMMs and the 5000 platform. So if you are hoping for something..dont..it will just backfire as the K10 expectations alot had.

But again, highend is low volume.

dualcore against dualcore they went well,but AMD didn't have a quadcore for a very long time, now they have one,and its a decent product,not a miracle but a good and solid product with a wel thought concept.
I desktop AMD will need have more horsepower, so hopefully it will have higher clocked parts
and a better B3 revision soon

[StealthyFish]

I would love to see a quote of any somewhat knowledgeable person who was being serious about an increase greater than ~40-70% over C2Q.
I'm guessing some noob in the 30k 3dMark06 K10 thread said the 170%.

I was expecting about equal to C2Q, hoping for a 10-20% advantage but we still have yet to see Phenom on an AM2+ board. Still won't push Phenom way ahead of C2Q but probably equal, possibly slightly ahead.

I'm still holding out to see how Phenom OCs and if Penryn is worth it over a Q6600.

I agree. I believe AMD's executives specifically said Barcelona would be 40% faster than the Xeon 51xx Clovertowns, and I've yet to see a 40% gain in performance over a clovertown in a benchmark, though, there haven't been many reviews or benchmarks run on their chips.

But, do keep in mind, there has to be a substantial overall performance gain over C2D for AMD to continue being competitive. It's much like the 8800 series vs HD 2900XTs. Yeah, the HD2900XT is a pretty good graphics card, but becuase it wasn't that much better than the 8800GTS//GTX//Ultras, their sales still suffered. If Phenom isn't going to be spectacular in comparison to a Q6600//QX6700//QX6800//QX6850, I really don't see any point in buying an AMD system. Price would be the final determining point; and seeing how AMD's in financial troubles and the whole "monolithic core" thing going on, I don't see how AMD can afford to play another price war unless they manage to live for another 2 years and hope to win that lawsuit against intel

[cupholder2.0]

Sure i'm not blind , i see and read what i see and read

It's clearly that AMD has regained the performance crown, it's just a hard pill for the intel boys to swallow.

Sorry kid, K10 can't even touch Conroe let alone penyrn. K10 is much more expensive to manufacture than K8. It is inferior in performance even to the Conroe, so there is no hope of matching penyrn. The penyrn will be much cheaper to produce and it will have 5-10% performance increase along with the lower TDP and superior transistors. AMD is far from taking the performance crown. AMD right now needs to worry about staying alive more than trying to take the performance crown.
For 3 years AMD beat Intel, all it did was made Intel get a bloody nose. In one year, AMD ended up in the intensive care unit.

[iddqd]

Correct me if I'm wrong, but I've heard server mobos aren't the best choice for overclocking

I also wonder what kind of chips are those - B1 or B2

The 3GHz CPU was obviously multiplier-unlocked, which makes it much easier to overclock in any old board..

IIRC, at launch, Opteron was only doing 1.8GHz, which of course was nowhere near as fast as 3.4GHz Xeons/P4s... oh, how the things changed a few months later. So, it's still too early to tell. Maybe they'll be able to compete with 2x2-core Pentium-M chips.

Also, I think their reverse-HT gambit might pay off to improve IPC on a single thread, once properly supported by the OS and enabled on the chips.

[GoThr3k]

@Gothrak

After reading your poorly written BS without facts, data and logical explanations and my nick-name in the same post, it is my pleasure to respond. I don't believe that I can change your blind fanboyism about your favorite company, nor I think that you'll learn something. But its kind of fun to frustrate people like you.

is it my fault you're always full of BS,and sorry english is only my third language....

It is not the RAM that provides bandwidth, but it is the memory controller(s). For example K10 with DDR2-667 provides more bandwidth than K8 with DDR2-1000.
Memory bandwidth is not CPU performance. It can help in certain cases, but until we have a proof(backed up with DATA) we can't conclude anything.
The fact that K10 has two optimized(each much more advanced than the one on the K8) memory controllers and a shared L3(which all can be dedicated to a single core if needed) leads to the conclusion that K10 should perform well even with the high latency rDDR2-667.
In a case you didn't know, the cache memory reduces the impact of the RAM latency and bandwidth(i.e. Core2).

So anyone with some basic knowledge about CPU architectures won't expect a dramatical performance improvements from a faster RAM.

i have basic knowledge about cpu architectures
indeed,we need proof backe up with data,so proof me there is no bandwidth difference between DDR2 667 ECC and DDR2 1066 on K10...

Then you must be blind. When you are talking about performance in 2P environment, the clear winner is Clovertown(X5365). You can cry as loud as you want about the clock advantage it has, but it is almost a year old CPU and blows away the Opteron 2350.
What makes the Opteron 2350 attractive is:
1) the overall system energy efficiency
2) overall system price
Thats why it will gain some market share for AMD. Those who are looking for a TOP performance are looking at Intel.
No matter how hard AMD tries with their K10 barely reaching Core2 IPC and their 65nm SOI they are not going to come close to the performance crown in 1P and 2P.

i was talking about 2GHz barcelona vs 2.33 ghz Clover,and there is no clear winner there...

It is VERY important for AMD, but not for the mobile/desktop/workstation/server markets. Quadcores already exist for many years, and if you talk about x86 Intel have sold millions already.
If no body buys them, then no body will produce them. So, there are people and companies who need the best performance and are willing to pay a premium for that.

so you say quadcore isnt important for the server market??

Yes! AMD needs money or it'll die sooner than most people here expect. When AMD will make some profits and pay off the loans, we can say that they are saved from the disaster. Until then, we can only hope they survive.

Sufficient to compete in 2P and 4P environment until Nehalem.

Don't be surprised if AMD paper-launch the parts you hope to arrive soon.

so you have crystall ball and knows how nehalem will perform?

[xVeinx]

The 3GHz CPU was obviously multiplier-unlocked, which makes it much easier to overclock in any old board..

IIRC, at launch, Opteron was only doing 1.8GHz, which of course was nowhere near as fast as 3.4GHz Xeons/P4s... oh, how the things changed a few months later. So, it's still too early to tell. Maybe they'll be able to compete with 2x2-core Pentium-M chips.

Also, I think their reverse-HT gambit might pay off to improve IPC on a single thread, once properly supported by the OS and enabled on the chips.

reverse-HT.... reverse-HT... he said... *falls down to the floor laughing*

No offense, but that doesn't exist. There are no whitepapers, no evidence at all that sometone has even figured out how to do it much less implement it.