Printable View
It's more than likely a compatibility issue with the HT2100 chipset. AFAIK, this mobo is as close as you can come to a "full-blown" K10-ready solution atm. The only problem is that none of those features are even selectable in bios. As we see more boards appearing with the HT2100 chipset, we'll eventually see the true capabilities of such a mobo...Quote:
Originally Posted by PhilDoc
Quote:
Originally Posted by s7e9h3n
Thanks, looks like we'll just have to wait. Thanks for the info today and taking your time to feed the hungry multitude. It was very much appreciated.
Ops, so you are saying that there are problems still with NUMA and being only single channel memories, s7?
Thanks for the information again. ;)
Hi s7e9h3n,
The CB R10 single thread result, 1900, 3.0 GHz Penryn around 3500, so very slow here, perhaps somebody knows is CB hurt my memory performance? Excellent scaling.
As you pointed, memory bandwidth is poor, half of what it should be. Everest reports RAM is 333 Mhz. SDR speed.
Photoworxx score is very good (!!), other FPU scores seem to be about 5-10% slower, clock for clock than C2D.
Possible to run some benchies?
Folding @ Home, frame time for any SMP project
Super Pi 32 M
Wprime 1024 M
Mainconcept H264
Windows media encoder
A 3d game (unreal tournament 2004, serious sam 2, Quake 4)
PC Mark 05 (has a bunch of math and encoding/decoding tests)
Thanks.
Quote:
Originally Posted by JVguest
Point me in the right direction and I'll take a look@ it. Don't know if I really want to be listening to the whine of a 1u chassis for too long though :p:Quote:
Folding @ Home, frame time for any SMP project
I honestly tried to give this a go, but it was so painstakingly slow that I had to stop it. One thing I did find interesting about the run, though, was how the OS handled the single threaded Spi process. The calculations were all done on physical cpu #1 and cycled through all four cores relatively smoothly (unlike earlier). EDIT: LOL...oops...I forgot it was running in the background...just popped up a window saying "Pi calculation done." 38:11.60...yuk....Quote:
Super Pi 32 M
http://img341.imageshack.us/img341/2724/wprimezn7.jpgQuote:
Wprime 1024 M
Lead the wayQuote:
Mainconcept H264
Windows media encoder
I'm waiting to see if it's possible to run my 8800ultra in this rig....;)Quote:
A 3d game (unreal tournament 2004, serious sam 2, Quake 4)
Need to install 32-bit windows for this. Will try to soon.....Quote:
PC Mark 05 (has a bunch of math and encoding/decoding tests)
@Johanne: Yep :yepp:
s7, check out my sig for a link to F@H setup guides :) SMP client would be REALLY helpful, actually on that you could run two of them :D
s7, when you put it on a linux operation can I ask for some other specific benchies? ;)
Thanks a lot. :)
Lol....linux is still a bit away, but I guess I can always take requests - any particular version?Quote:
Originally Posted by JohannesRS
I hate to say it but... They sound like quad core K8's, eh?
Hopefully they'll shine in the unforseen areas. :D
Folding @ home, SMP client: http://folding.stanford.edu/download.htmlQuote:
Originally Posted by s7e9h3n
During install you'll need to be logged in as administrator and need to have a password. The install.bat file will request username and password, I think it needs to be the windows user/pass. You'll need a working internet connection to run F@H. A frame should finish in less than 25 minutes.
For super pi 32 M, set affinity to one CPU and priority to high or real time under windows task manager.
Forget the main concept, I think you need an add in DSP board to run the bench, but windows media encoder you can get from here: http://www.free-codecs.com/download/...r_9_Series.htm
Then of course you need a file to encode, and then need to compare to the same file encoded on a K8 or C2D.
How about in UNBUFFERED Sandra :wth: ?Quote:
Originally Posted by akaBruno
http://img406.imageshack.us/img406/1...aunbuffxb2.jpg
Same exact settings as earlier, just unchecked the "buffered" box in options....
Sheesh, looks like this chipset's gotta a few bugs to work out.
I have faith in you..You'll make that puppy sing!:up:
Seems like there's a few teething pains, but it'll be interesting to see how quickly they're sorted out.
Thanks for the updates.
hehehe i don't think you need to give S7 tips on 32M :DQuote:
Originally Posted by JVguest
LOL! Well done s7, marvelous journey :D
CB single CPU FX 70 at 3.2GHz got 3,000 and 4 CPUs got 11,000.
Movieman ran his X5350 at 2GHz and got around 2,325 single and 14,260 for 8-cores IIRC.
Yup, applications depend on coding as much as CPU power nowadays. K10 doesn't look like a hotty in the CB area alone, but as for the rest, we'll wait and see...
CB is an area where K10 should theoretically shine...
but giving that this B1 + mobo has serious FP/SSE
performance penalties(its even a tad slower that K8
in CB:p), it was expected.
is anyone wiling to buy 2 retail BAs? :)
The problems aren't with the cpu's as much as the crippled chipsets. Throw a B2 on any current board and I promise, you'll still encounter quite a few bugs. All we can do is be patient.....(easier said than done :p)Quote:
Originally Posted by alayashu
nice to see its running up again.. now they need to fix up dual channel to work 1st.
i just got my Asus KFSN4-DRE board in this AM. will hook it up tonite and start running benches. still integrated video but, it's definitely a better looking board.
this DOES feature split plane and dual HT links. the chips tested will be B1s as BA's are still not here.
dave
Sounds great Dave, good to hear from you. Doesn't that board have a 16x pcie slot?Quote:
Originally Posted by dave_graham
Better looking than what? :p:Quote:
Originally Posted by dave_graham
:)Quote:
Originally Posted by s7e9h3n
Thanks, s7. But no specific request on a linux distribution.
But I would possibly make use of acml, atlas, fftw and gromacs benchmarks on linux. ;)
But I guess those will be still a bit "useless" before dual-channel and numa gets to work properly. When the time comes, I bother you with the links (unless you got "too curious", what I would not expect... :D ).
Edit: God, I will not see any Barcellona on my hands. Dave saying good things about and Asus board means that the armaggedon is just begining... :)
3992. :) which, incidentally, I should have the E version here somewhat soon and a 2915-E (SAS version) shortly after that. :)Quote:
Originally Posted by s7e9h3n
cheers,
dave
yes.....don't know if it supports video but, I'll give it a shot. I only have a 7800 GTX around the house so, you'll have to forgive me for any terrible 3d based benchmarks.Quote:
Originally Posted by PhilDoc
this is, however, a very good 1U board....
Asus's packing job, however, was not....we'll deal with that later.
dave
If support for those boards is anything like it is for the 3993, I wouldn't be getting my hopes up too high yet :(Quote:
Originally Posted by dave_graham
i've got the backing of a few higher-ups at Tyan and AMD on this...not too worried.Quote:
Originally Posted by s7e9h3n
I'm more worried about high i/o failures and the like.
btw, run the AMD platform stress test tool and give me a holler as to how the 3993 is running for you.
dave
3Dmark2001 and 06 would be enough to see it's power clock per clock.Quote:
Originally Posted by dave_graham
Ofourse we would have to take the slow mem in mind.
Looking forward to that though :)
and if you want something very quick, take aquamark 03
I've run that stress test a few times....anything in particular you'd like to know?Quote:
Originally Posted by dave_graham
s7e9h3n, where is 3dmark06 cpuscore? all we are waiting for it
Hehehe...I didn't know I was planning on running it, but I'm trying to figure out how to get my ultra into this server chasis without having to remove the mobo. Maybe I'll have to in the end....Quote:
Originally Posted by MAS
place your mobo onto the table and run 3dmark
many people are eager to know the results he-he
Quote:
Originally Posted by dave_graham
Anything you have is just fine. Good luck with the new board. If it runs good, I might think about buying a couple of the low end cpu's and this board, just for fun.
I'm patiently awaiting for a 2.6ghz+ Phenom model. :D I was going to buy another Intel quadcore, but I think I'll wait for some Phenom action.
Well, one thing I can tell you for sure is that there's no way possible to run my 8800Ultra on this rig because I just realized that the card is too long. Installed, the Ultra would run into the heatsink on CPU1. But the thought of running 3d on this rig still intrigued me to the point where I started messing around with the chasis, did a few quick mods, pulled together some necessary parts, and I ended up with this (yes, it's a bit messy, but what do you expect for ~20 minutes work :p: ):Quote:
Originally Posted by MAS
http://img125.imageshack.us/img125/2519/wrignp8.jpg
And in case you're wondering, YES, that's a dedicated Meanwell 12v 320W PSU I had to add since the 7950GTKO I cannibalized from my HTPC doesn't have enough juice supplied from the PCIE slot.
This is one of those times where you guys need to be thankful that I'm just as geeky, if not geekier than the rest of you when it comes to hardware :D .
http://img125.imageshack.us/img125/8251/nvji0.jpg
3D to come a bit later.......:woot:
Love those extra bigg-ass fans over those small cpu coolers ;) Will be very intresting to see 3dmark06 Cpu scores later S7, Good job and pic´s as always! Keep it up!
Looking sweet S7, thanks for the big effort.
WOW!
Nice 1U mod s7e9h3n! :p:
BTW what's that big white thing in background? UPS?? :shrug:
looks lika a
http://g-ec2.images-amazon.com/image...AL._AA160_.jpg
APC Back-UPS RS - UPS ( external )
Yep...good eyes :up: It's actually 2 of those.....That silver box next to them is an older Promise NAS Box and the 1u is sitting on top of my Silverstone HTPC....and if only you guys could see what's in the rest of the room :cool:Quote:
Originally Posted by funoshi
I'll get round to playing with 3dMark in a bit......
EDIT: @ FUNOSHI - BTW, let me be the first to welcome you to XS....nice first post...lol:toast:
s7e9h3n, thanks. Mankind will never forget you =)
hey MAS who are you talking about?
msimax, see several posts above
i was joking:)Quote:
Originally Posted by MAS
Check out Anand's weblog:
Quote:
..Those were my highlights from IDF, now on to Toronto. I was there visiting AMD to talk about you know what. There's not much I can say, which is unfortunate.
http://anandtech.com/weblog/default.aspx?bid=301Quote:
I have a feeling Phenom may end up being a little more competitive than expected, especially because Penryn won't hit the mainstream until next year. Once dual-core Penryn reaches the market in early '08 then AMD will be in for a rough ride. I seriously doubt we'll be seeing an upset on the level of what Intel did with the Core 2 launch however; that processor and microarchitecture look like they'll remain on top for the near future.
Hmmmmm.. Interesting comments.. Not sure what to make out of this tho.
now it's clear K10 won't be a monster.
Of course 1066MHz dual-channel low-timing memory mode will greatly help Phenom in its fighting against fast new Intel's cpu, but only at stock frequencies...
it looks like there is no sence to wait for X4, you can buy NOW Q6600 @3600MHz - and it will be faster than any AMD K10 CPU during the next half a year
MAS: *even* disconsidering the fact that this topic seems to be a bit more "server" than "desktop": I still think no one can say a word for sure about Phenom. Even for XS users.
I mean it not because of price or stock performance. But with a stock performance comparable do intels clock-per-clock, here at XS people would still be interested in how well that thing will OC. And, for dear Lord, absolutelly nobody knows it now. All options are still opended, and it can OC as well as, worst OR better.
Well, maybe there exists a few people around the world who should be able to say, but are prohibited for NDAs, but that's another point. ;)
Running 3d05 and 3d06 didn't go as smoothly as I had hoped but I finally was able to get a couple results uploaded to my project manager. I get the feeling that "Pro" users must have a much easier time obtaining and uploading results to thieir accounts, but I don't run 3d enough to warrant paying for the prog :p:
I guess Futuremark interprets my results as invalid since I can't publish any of them and therefore I don't have any comparison links. Regardless, the results - in my eyes - were far from spectacular due to problems with Futuremark's engine not having the ability to correctly identify my hardware (and it doesn't help that my own hardware seems to have problems identifying itself :p: ) So here are caps of all the pages from my resultanalyzers:
Everything @ default clocks.....(well as close as we can get to default)
3d06:
http://img267.imageshack.us/img267/8620/3d06ass3.jpg
http://img267.imageshack.us/img267/297/3d06byl9.jpg
http://img239.imageshack.us/img239/5189/3d06cyf8.jpg
http://img441.imageshack.us/img441/1187/3d06dhv5.jpg
*Nice to know that my system isn't Vista capable :rolleyes: *
http://img441.imageshack.us/img441/1936/3d06efv7.jpg
3d05:
http://img441.imageshack.us/img441/831/3d05ahi7.jpg
http://img441.imageshack.us/img441/2411/3d05blw7.jpg
http://img441.imageshack.us/img441/5850/3d05cih9.jpg
http://img441.imageshack.us/img441/9151/3d05dxs5.jpg
http://img441.imageshack.us/img441/7936/3d05epy4.jpg
I had some major issues trying to get a result on 05. For some reason, when the benchmark finished, 3d05 would just disappear and if I restarted windows immediately afterwards, an error report pops up from windows regarding 3d05. The result from above was the only one I was able to upload out of close to 20 runs.
*On a side note: I don't know how you guys can stand to watch the CPU tests from 3d06. The bench moves along @ way under 5fps. There were moments when I began wondering if my Etch-A-Sketch could get better frame-rates than this rig :D
Hey, thanks, S7. Not bad really for a crippled system. You think 5fps is bad try the 2 fps for my opteron.
:lol2: true that!Quote:
Originally Posted by PhilDoc
I think you guys misunderstood me....take a look at my avg. fps. @ 2fps, your opteron beats my octals :D
LOL, I was being kind to my Opty. I think it was actually less then one. You're cpu score is about double the one I had. It's like watching grass grow, reminds me of a few plastic surgery cases I scrubed in on.Quote:
Originally Posted by s7e9h3n
Rgr that Phil,in dual core rigs,mark06 CPU tests crawl around 0.5-0.8fps :p: ,so Steven's octo rig is mighty fine in around 2fps :)
3800 X2s OC nice.
nvm, i feel stupid...
Quote:
Originally Posted by informal
Looks pretty tasty to me. I just don't know how much having dual cpus adds to the mix
For comparison my X2 5600+ CPU score @3174MHz is 2402.
CPU1 - Red Valley 0.755 FPS
CPU2 - Red Valley 1.222 FPS
Here's a couple woodcrest numbers if they help comparison for you at all steve ;)
2x 5120 @ 7x402.5 (2.817ghz)
4x 1gb FB-DIMM
X1900GT 648/900
3d06 cpu score: 4403
test 1: 1.458fps
test 2: 2.127fps
Thats all I can get out of them at the moment.. but there's a bag of dry ice in my freezer :D
So Penryn on X38 and SkullTrial boards gets:
3DM 06 (CPU)
1x QC 3GHz (1333FSB) : 4569 (unreleased part)
2x QC 3.4GHz (1600FSB) : 6359 (unreleased part)
1x QC 4.14GHz (1840FSB) : 7000 (Kinc)
1x QC 5.2GHz (1600FSB) : 8092 (Kinc)
2x K10 2GHz : 4242
According to Intel at IDF. Not bad, not bad at all.
Thanks s7. Must be a good workout :p:
s7e9h3n, gracias
can you run tests with only one cpu installed? it's not very hard to temporarily remove second cpu from socket (and its memory modules) :)
PS see my post http://www.xtremesystems.org/forums/...&postcount=743
tbh, for some reason I'm thinking that 3d06 is calculating the cpu scores using only 4 cores max regardless of platform. also, everyone remember that I'm running the bench @ 200fsb vs. Intel's 333+ speeds...Quote:
Originally Posted by MAS
hahahaha sooo trueQuote:
Originally Posted by s7e9h3n
The CPU score is pretty good actually, almost doubled up on my X2@2.95Ghz, which is pretty good considering the 950mhz disadvantage, so even assuming that quad cores double performance, thats quite the clock per clock advantage.
They almost match up with Stevil's dual woodcrests at 2.8 which again suggests a sizeable clock/clock advantage. I would love to see some oced numbers on these chips.
Yes, it would be impressive.....if I was running 4 cores. But remember....2xBarcelona = 8 cores ;)Quote:
Originally Posted by funnyperson1
Oh wow, I don't know how I missed that, for 8 cores thats kind of sad unless the benchmark only uses 4 cores.
s7e9h3n: cant Ati 2900 or Nvidia 8800 be modded to fitt in this board?, is one of the cpus coolers to close to the pci-e ? cant the cooler be modded on the cpu? so it can fitt.
So 4k for 8 cores @ 2GHz?
I get 2k with 2 cores @ 2.8GHz. ;_;
If this is correct, even it is server mobo & quite early server CPU, I would say that at least the current K10's are doomed, unless they clock to something like 5-6GHz.
Talking about extreme-end! ;D
Hello old friend,Quote:
Originally Posted by Lastviking
Nope, the card would make it impossible to install any type of cooler on cpu1. BUT I possibly have a solution to end all questions about the true performance of these cpu's - A WORKING BIOS WHICH SUPPORTS BARCELONA ON THE ASUS L1N64!
Prepare to see some damage done to the ORB ;)
:eek: :up:Quote:
Originally Posted by s7e9h3n
My hope is: like good wine, K10 gets better with time :D
Nice results for a CPU in a , definitively, non-optimised platform. :up:
Is correct to assume that the desktop version of K10 will be much "better" than the server part?
EDIT: s7e9h3n we´re all expecting great things after those words...keep it coming...
For comparison heres what I get if I clock my X2 at 2Gs:
CPU Score 1489 Marks
I hope for AMD's sake that 3DMark doesn't use all 8 cores, because if it does there is a performance gap that cannot be explained away by poor memory bandwith.
:confused:Quote:
Originally Posted by Blasar
I see no reason for performance difference being any other than zero percent (0%).
Just like with K8 server vs. desktop, it's the exact same chip in the end.
Very good news s7e9h3n :).How soon can you start testing ? :)
@largon
In case you didn't notice,both s7e9h3n and Dave Graham had various issues with their rigs,memory running in single channel mode and @poor clocks,cas latency is 5 and the memory is regged...Also these are B1s ,not BAs etc.
largon, because of memory
Phenom's conventional memory is ddr2 1066; as you know K8 (and obviously K10) performance is extremely depends upon memory speed. You can get 10 per cent or even more performance boost when replacing registered ddr2 667 with low-timing ddr2 1066
we are waiting for a new bios :)
s7e9h3n, how hot are the chips?
I don't know, doesn't look that bad to me. Maybe I'm missing something. Look at the 2 x QC 3.4G (1600 fsb). What's S7 running, a crippled 2 x 2G at 667 fsb in single channel. Remember these are B1's. I would expect the gap to close considerable as clocks got higher and memory systems improved. Of course there's no way to know for sure until we get a BA system running full memory compared to one running 2.5G full memory.Quote:
Originally Posted by KTE
MAS,
10% plus?
Sure, in bandwidth benchies.
No really, since when did Hammer start being "extremely" bandwidth/latency dependant?
Largon,is it that hard to understand:regged memory,cas 5 latency,single channel.All of these affected K8,plus we have 4 cores now instead of 2 and they are more bandwidth hungry than K8 wasQuote:
Originally Posted by largon
well you could force 3d06 to use just one cpu so we can compare easier to intel and clear the 4/8 cores usage drama
Finally new bios... Good Luck :up:Quote:
Originally Posted by s7e9h3n
just when i think i can take a break from this thread.....
E5310 2x 2GHz QC: 4335 http://service.futuremark.com/compare?3dm06=2982103
Supermicro X7DA8
E5320 2x 2.33GHz: 4688 http://www.xtremesystems.org/forums/...1&postcount=21
X5535 2x 2.66GHz: 5662 http://service.futuremark.com/compare?3dm06=746484
Shintai QC 2x 2.66Ghz: 5772 http://www.xtremesystems.org/forums/...0&postcount=25
Q6600 2.4GHz: 3846 http://service.futuremark.com/compare?3dm06=2424217
Asus Striker
How dat for a multi-core efficiency comparison? :D
BTW the scores differ a lot on the motherboard. The P965 boards and the modern X38 board have like a 200-450point gap for the same processor and setup.
I mean, look at this from the ORB:
Q6600 3607MHz: 5285 (Striker)
Q6600 3420Mhz: 5194 (P5K-E)
And QX6800 on BX2 and P5K:
BX2: 4136
P5K: 4448
basically you mean this bench is almost useless... :)
Quote:
Originally Posted by KTE
Thanks KTE for taking the time to look up these scores. I had trouble understanding the gloom and doom anyway, but this certainly puts things in perspective. :up:
The gaps are ridiculous you can tell nothing from the differences showing in processors (Extreme vs Non) and from different chipsets. My two 975s would bench versus a P35 chipset like a Bearlake.Quote:
Originally Posted by KTE
No worries Phil. It's good to see it all in perspective. Added some more from other thread... ;)
It's not a 8 core benchmark by the looks of it, it looks like it can scale only to around 6 cores from all the benches I'm seeing. It does look like using more than 4 cores but not all 8. There's definitely a setup affect on the CPU scores too which makes comparisons a little "inaccurate" to say the least, since a single chipset change seems to make a big difference.
Just woke up....I'll get to it ASAP. BTW, I'm taking a stance somewhere in between yours and largon's :p: I still say 3d06's cpu bench is optimized for 4 cores or less....Quote:
Originally Posted by informal
EDIT: Oh, I forgot to mention...I should have ~6-10 BA 2350's to play with later on today. We'll see how much difference there is in the steppings.....;)
KTE did post some interesting observations concerning mark06 scaling to more cores(according to him,only ~6 cores is used by the test-and probably not at 100% utilization..)Quote:
Originally Posted by s7e9h3n
Good news about the new chips.Any news on the Asus QuadFX front? :)
Quote:
Originally Posted by alfaunits
Yep, just caught that post...it's similar to what I was thinking as well....Quote:
Originally Posted by informal
I'll flash my board in a bit...I'm just as excited as everyone else, but ATM I'm just not as awake. Need some :coffee: before I :comp10: :D
Hate to do this to you guys but I'll be busy for a few hours. Testing the bios won't be possible until I've wrapped up some prior obligations. At least I'll be testing with my new 2350's :D
Keep on with the good work of testing :up: , i still using nf4 and bh5 and i want to see if its soon time to change the stuff to something newer :), but i think i wait for good ddr3 boards.Quote:
Originally Posted by s7e9h3n
That question really should be rephrased: "How COOL are the chips?" and my answer would be that I've yet to see 30C in bios for either cpu ;)Quote:
Originally Posted by alfaunits
Good plan as I'm waiting for a good one as well.....Quote:
Originally Posted by Lastviking
better than seeing another B1 test (even with the right BIOS the B1s areQuote:
Hate to do this to you guys but I'll be busy for a few hours. Testing the bios won't be possible until I've wrapped up some prior obligations. At least I'll be testing with my new 2350's
said to still have performance penalties) :P
Cinebench 9.5 and 10, Stephen, please!:D
autocascade today?:DQuote:
Originally Posted by s7e9h3n
autocascade today?:D
autocascade today?:D
autocascade today?:D
autocascade today?:D
autocascade today?:D
autocascade today?:D
autocascade today?:D
autocascade today?:D
autocascade today?:D
why Cinebench?
First of all 3dmark and Sandra please.
If it's possible - test both dual cpu and single cpu modes.
And overclockability of Barcelona http://www.xtremesystems.org/forums/...es/biggrin.gif
I need a new render machine eh:DQuote:
why Cinebench?
3dmark, Cinebench (X64) and Sandra then:P
Exciting stuff, let us know man. Meanwhile, I'm wondering what I was thinking cashing in on one of those DDR2 deals now...$60 of unused 2GB haha.
Not to diminish your previous results, but now things get really exciting :up:Quote:
Originally Posted by s7e9h3n