Some Actual Barcelona Benchmarks

Status
Not open for further replies.
T

tsuehpsyde

Supreme [H]ardness
Joined
Oct 22, 2004
Messages
6,604
Well, I posted these over on another forum, but I might as well toss them here since I don't see anyone else posting any. :)

Take note that this system is running in Single Channel, DDR2-533. The single channel mode is being reported in CPU-Z, so I'm assuming it's correct. The memory is clocked down because it won't boot into Windows unless I do so (not sure if the procs or memory or mobo is to blame, haven't had time to troubleshoot). I had to do some changes here and there to get it to cooperate, but at least the system is running. :)

First, our CPU-Z:
http://valid.x86-secret.com/show_oc.php?id=243088

First thing I'll note is the HT Link. Processor #1 says 400MHz (BIOS says 1400MHz), Processor #2 says 200MHz? :confused: Not sure if CPU-Z is showing it wrong or if the BIOS isn't doing what it should. Either way, take note of that.

Our ugly, ugly SuperPI time:
nastypi.png


Again, the memory is not helping matters there. Next, Cinebench R10:

cine10_32bit.png


If you guys want any other benchmarks, let me know. 3D is onboard since this is a server (work is letting me test it since this is a test box), so 3DMark** or AM3 are both out as they'd be pointless on onboard video. Hopefully I can get this thing stable so I can get some more honest numbers out of her...
 
yah really cant say much till you get the quirks worked out with the memory and whatnot. lookin forward to seeing what you can do wiith though :cool:

edit: btw, are you using registered ECC memory? if not maybe thats the problem
 
fluxion said:
yah really cant say much till you get the quirks worked out with the memory and whatnot. lookin forward to seeing what you can do wiith though :cool:

edit: btw, are you using registered ECC memory? if not maybe thats the problem
Click to expand...

Haha, yeah, that's Registered ECC memory. It wouldn't POST without it. ;)
 
Well... since SuperPI only uses 1 core.. that score is to be expected for 1.9Ghz.

As for the HT link and memory speed.... I bet the BIOS is buggy and CPU-Z is probably also not working quite properly.

I would try less RAM (the least you can use) to see if maybe you have a bad stick of RAM.

What are the rest of the specs of the system?
 
Folding@Home SMP client.
Just let it run for a few percent numbers to get logged into fahlog.txt.
Give us the protein it's working as well... or just heck paste the fahlog.txt here or PM it to me :)

Lots of folks are interested in the numbers the SMP client will put out. It seems as the Intels are very cache dependent and the speed at which the cores can intercommunicate plays a big role in the SMP client.
 
tsuehpsyde said:
Our ugly, ugly SuperPI time:
Click to expand...

I think that the superpi score is not too bad, at the current speed it shows its about 1 - 3 second slower than my K8 (20%ish faster than 4800+ x2(2.5mhz)clock for clock) also im using ddr2 800mhz @ 356mhz. Plus there is the HT 400mhz issue and the fact that early reports state that the chip will scale better than conroe/clovertown.

When i first read Anandtech's i was not too happy but after reading more about Barcelona i think it will hold its own.
 
New CPUs are so disappointing (Conroe included), I wish we got double the performance with each release like GPUs.
 
Nobi125 said:
New CPUs are so disappointing (Conroe included), I wish we got double the performance with each release like GPUs.
Click to expand...

the performance is over double, 2 cores to 4 cores with better IPC. The reason that chip makers have gone multi core instead of higher clock speeds is because if you go much over 3ghz there is a big heat/wattage penalty. They can however increase IPC as well as increasing core count.
 
j1979 said:
I think that the superpi score is not too bad, at the current speed it shows its about 1 - 3 second slower than my K8 (20%ish faster than 4800+ x2(2.5mhz)clock for clock) also im using ddr2 800mhz @ 356mhz. Plus there is the HT 400mhz issue and the fact that early reports state that the chip will scale better than conroe/clovertown.

When i first read Anandtech's i was not too happy but after reading more about Barcelona i think it will hold its own.
Click to expand...

I don't think there is any chance for Barcelona in SuperPi test, though it may be slightly better than K8. According to the test made by Coolaler, a Barcelona at 2GHz got a score of 39.6s, which shows the frequency/performance scale is just proper but not good enough to help AMD catch up when it might reach 3GHz the next year.
 
The SMP folding client only opens 4 threads, and while it may do some good, it may not. Plus, this isn't my rig so I don't want to abuse it too bad. It's a test rig for work and they're letting me run tests on it. :) That, and the general instability of the machine makes me cautious about folding or other intensive stuff (at the moment).
 
j1979 said:
the performance is over double, 2 cores to 4 cores with better IPC. The reason that chip makers have gone multi core instead of higher clock speeds is because if you go much over 3ghz there is a big heat/wattage penalty. They can however increase IPC as well as increasing core count.
Click to expand...

So how can GPUs become twice and nearly 3 times faster than previous generations without implementing multi cores... They must pack more transistors, or new technology. Why does CPU speeds and performance hardly increase over a period of 2 yrs. At one point of time it was different. Did developers reach the CPU ceiling or some crap.
 
because GPUs perform limited kinds of processes, plus they're designed differently to CPU's.

I can't remember if arstechnica or somewhere else wrote a report but just remember it's easy to come out with a spanking new chip that does a few specialized things extremely well, it's not easy to come out with a backwards-compatible processor that has to adhere to very restrictive conventions AND make it eat less power AND make it faster. Just look at SSE5/Altivec - or where Alpha or MIPS were back in the late 90's. These processors had screaming performance but what's the point when you have to re-engineer your entire software base to use them?
 
Shottah_king said:
So how can GPUs become twice and nearly 3 times faster than previous generations without implementing multi cores... They must pack more transistors, or new technology. Why does CPU speeds and performance hardly increase over a period of 2 yrs. At one point of time it was different. Did developers reach the CPU ceiling or some crap.
Click to expand...
This post makes me very unhappy. The lack of research, the lack of common sense and the sheer arrogance.

So here it goes.

WHAT ARE YOU TALKING ABOUT?! DO SOME RESEARCH.

DO YOU KNOW HOW MANY PROCESSORS ARE ON THE G80 CORE? 128 SCALAR STREAM PROCESSORS. YES THAT'S 128.

Do you know how many are on the Barcelona? 4 General Purpose Processors

You think engineering is just some sort of magical dust they sprinkle on the wafers.
 
Keep on topic, boys. ;) No need to get all fired up over some Barcelona benchmarks.
 
CoW]8(0);1031436331 said:
You think engineering is just some sort of magical dust they sprinkle on the wafers.
Click to expand...

you mean its not?
then how come when you let the magic smoke out, it doesn't work?
J/K :p
 
what applications does superPI mimic, i mean is a good superPI score just for bragging rites or does it represent what some real world software would run like ?

i have always wondered that
 
Exactly, it's magic smoke.

And I always thought my friend made up that joke himself too. :rolleyes:
 
Shottah_king said:
So how can GPUs become twice and nearly 3 times faster than previous generations without implementing multi cores... They must pack more transistors, or new technology. Why does CPU speeds and performance hardly increase over a period of 2 yrs. At one point of time it was different. Did developers reach the CPU ceiling or some crap.
Click to expand...

A GPU(so far) is a very specialized piece of hardware, geared towards limited functions.
A CPU is master of nothing, but jack of all trades.
That and their differences in the physical layout, plus the fact that CPU's run in GHz and not MHz(unlike GPU's)

And if you don't think CPU has evolved over the last 2 year(hint :Core 2 Dou, Core 2 Quad and very soon 45nm) then you must have been living under a rock? ;)
 
APOLLO said:
If the processors are release units and not ES, then the appropriate stepping should be BA. B1 was a slower stepping and not intended for release.
Click to expand...

These were pre-release samples for testing purposes, but they didn't get to using them until yesterday. However, they are labeled normally, not as an ES or anything. :) But CPU-Z clearly shows the B1 stepping.
 
CPU-Z Latency:

cpuz_latency.png


In other news, I moved the 2347s over to another identical rig, but instead of 8 x 2GB PC2-5300, it had 4 x 1GB PC2-5300. The other rig had a set of 2218s in it and was working fine with the same BIOS. Tossed in the Barcelonas, and it wouldn't boot unless I lowered it to 266MHz. :( So the IMC or the BIOS is to blame, but I won't be getting any 333MHz benches sadly.
 
From what I've been reading on the web, notably the 2CPU.com forums, all motherboards released prior to the Barcelona launch will require a new BIOS flash to support the new processors, if at all. There are supposedly some socket F boards that won't fully support it.
 
APOLLO said:
From what I've been reading on the web, notably the 2CPU.com forums, all motherboards released prior to the Barcelona launch will require a new BIOS flash to support the new processors, if at all. There are supposedly some socket F boards that won't fully support it.
Click to expand...

This system has been flashed with the newest BIOS. The ones prior to, they system wouldn't even turn on. ;) So trust me, it has the newest one.
 
And some more....

everestlat.png

everestread.png

everestwrite.png

sandrabw.png

sandralat.png

sandracpu1.png

sandracpu2.png


The latency from everest was the biggest wtf. My first run was actually 27.7ns. I said no way, and re-did the test and got 29. Then I tested and got like 28.1, 33, 36.3, 32, 29, etc. I had to run it a few times and managed 27.9ns, but I did get 27.7ns on my initial test, which is a bit curious...
 
The question is, does the newest one have Barcelona support, or will it be added to some newer BIOS release?

If the newest one doesnt have Barcelona support then that would make perfect sense. My guess is that at some point in the future they will support it, but it isnt ready yet.
 
Those Sandra memory scores are horrible. With my Opty175 I am able to get 7.6Gb/sec.

Then again.. I am running at 2.83Ghz as well.

I bet the BIOS is the main issue.

Oh and look at the BIOS release notes... if they don't say anything about adding barcelona support, it probably doesn't.

You could also contact the manufacturer of the board and ask them if they have a beta bios that adds support for the barcelona. Chances are if they are going to officially support it, they will have one and would probably be willing to give you a copy.
 
Could you run the "Multi-Core Efficiency" test in Sandra?

Dual core K8 was getting 1/3-1/2 the performance of Conroe in that test. It would be interesting if Barcelona improves there.
 
Guys, let's not forget these processors aren't release stepping, and not fully reflective of the performance potential. Also, the machine is experiencing some hiccups, so take everything in stride. B2 stepping with frequencies of 2.5GHz and higher should be available in Q4.
 
pxc said:
Could you run the "Multi-Core Efficiency" test in Sandra?

Dual core K8 was getting 1/3-1/2 the performance of Conroe in that test. It would be interesting if Barcelona improves there.
Click to expand...

sandramulti.png


It's pretty piss poor too. ;)

And remember, the memory is single channel DDR2-533 with Registered ECC memory. Of course the results suck. ;) And the HT Link is (possibly) at 400MHz. Ouch. :(

At this point, I'm blaming the BIOS until I see otherwise.
 
Thanks for the effort, but I actually prefer the thread entitled "Some Fake Barcelona Benchmarks"

I'm thinking now would be a great time for you to invest in an additional server though, from either CPU company. :p
 
Status
Not open for further replies.
Back
Top