I've come across some benchmarks done by AMD from a feb 2007 presentation.
They compare 2/4 Opteron 3GHz with 2/4 K10 2.3GHz.
Speed advantage for K10 : 53,3%
Most benchmarks scale by 55-66% considering a 53% speed advantage => K10 IPC over K8 is 0-20%.
Thank to enumae for the link.link
Last edited by savantu; 07-02-2007 at 05:25 AM.
And how did you come by a 53% advantage?In those graphs,most results are 66% advantage with 30% clock deficit for K10.And you said most benchmarks scale ~60%...So you are guessing here.
Last edited by informal; 07-02-2007 at 06:07 AM.
Because for 2P you are comparing 4 3GHz K8 cores vs. 8 2.3GHz K10 cores ; 12GHz vs. 18.4Ghz.Originally Posted by informal
For 4P yuou double that.
Got link?
Crunching for Comrades and the Common good of the People.
0-20% also sounds like what is expectable IPC increase. We just need abit more validation.
After checking the extensive powerpoint it seems ok. Just a shame that AMD is going the FUD marketing route. I really hope that powerpoint is the work of a single idiotic department. They even needed synthetic benchmarks to show an improvement over Intel.
Their sole focus on virtulization and theorectical numbers along with power consumption and synthetic benchmarks mixed with real also puts the nail in the coffin. K10 is a lackbuster in performance.
Its a sad year indeed for AMD. Hopefully they can get the ship turned around in 2008 or 2009.
Last edited by Shintai; 07-02-2007 at 06:30 AM.
Crunching for Comrades and the Common good of the People.
Man...s7even has been running those on Asus Quad Fx board like 4 moths ago.Search the forums.
I think comparison DC vs QC is little out of place for any IPC performance evalutations. We don´t know how effective are those benchmarks in taking advantage of two additional cores hence we can only guess. DC vs DC desktop comparison would come handy.
http://tweakers.net/reviews/661/3
Based on this summary and the images posted above, we can infer that 2.3 K10 compared to 2.66 Clovertown..to trail in SPECjbb2005, to be beat in TPC-C, and finally match in SAP-SD. The others aren't worth mentioning for obvious reasons.
One word. Some stupid person had to put the company in a lock on Native quad. A K8 MCM quadcore at 2.5-2.7Ghz (With maybe a few special 3Ghz) would have been a very good thing (And massively cheaper to make). And AMD seems to have lots of issues scaling clocks on 65nm.
K10=prescott.
Crunching for Comrades and the Common good of the People.
K10=prescott.
You are joking,aren't you?
Yes He's joking, prescott works higher in frequency
You are joking,aren't you?
No, looking on the scores and failed ramping. And remember the 2.3Ghz is a 120W, not 95W. With 0-20% boost a possible 120-130W 65nm K8 MCM Quad at 2.7-3Ghz would be a better option. So its a step backwards.
You are joking,aren't you?
Last edited by Shintai; 07-02-2007 at 06:50 AM.
Crunching for Comrades and the Common good of the People.
Why ? AMD is saying that not me.
K8 is very strong to start with ; you simply can't leap forward by a mile when the bar is already set high.
If the graph is right, it got NDA all over it, so lets assume it is, K10 is like 10-15% faster then K8 clock for clock.
Not enough for desktop, but for servers AMD will put up a fight.
AMD has an official statement, that its K10 core is 40% faster than Clovertown clock for clock. That's been said, those graphics are likely to be fake.
be patient, its not much to wait till August :P
Really? Please let me see "We will beat Clovertown by 40% clock for clock in general".
http://www.channelregister.co.uk/200...arcelona_2ghz/
"We will be seeing a performance boost of 40-50 per cent above our highest frequency dual-core products that are available today," AMD VP Randy Allen told us.
The 40% is from 1 single benchmark, they never claimed clock for clock. And K8 already beats Intel in the same bench with about 20%.
Crunching for Comrades and the Common good of the People.
Or about 5-20% faster clock for clock than K8. In other words, about Core 2 performance.
Crunching for Comrades and the Common good of the People.
Or 10-15% faster than C2D,since we don't know for sure how it will perform in desktop usage model.
Uhm..no. In the best case it might be 5-10% faster on the serverside due to the FB-DIMM issue. On the desktop side it will be about the same +/-5%. So AMD needs to raise the clockspeed alot.
How did you even come up with the 10-15% when its K8 vs K10 on an equal platform o.O
You need to use the K8 baseline for all comparisions with Core 2 and add K10.
How can it be 10-15% faster than Core 2, when its 5-20% faster than K8. And Core 2 is 5-20% faster than K8 on the desktop.
Last edited by Shintai; 07-02-2007 at 12:19 PM.
Crunching for Comrades and the Common good of the People.
And how do you know that??Please,stop guessing around when AMD is in question.Who knows,maybe it will be on par.But you make it sound like you personally tried it and it is set in stone.
So,did you try the desktop version with unreg. memory at 1066Mhz?No?So stop guessing.
Uhm. You do know the AMD powerpoint with K10 vs K8 right? Everything is detailed in it. What K10 speed, what K8 speed.
So the guessing part is more or less over. There is no maybe´s or if's anymore if that powerpoint is correct.
I know you dont like the news, but get over it and look forward.
Crunching for Comrades and the Common good of the People.
You mean the optimization guide?
Seems we made our greatest error when we named it at the start
for though we called it "Human Nature" - it was cancer of the heart
CPU: AMD X3 720BE@ 3,4Ghz
Cooler: Xigmatek S1283(Terrible mounting system for AM2/3)
Motherboard: Gigabyte 790FXT-UD5P(F4) RAM: 2x 2GB OCZ DDR3 1600Mhz Gold 8-8-8-24
GPU:HD5850 1GB
PSU: Seasonic M12D 750W Case: Coolermaster HAF932(aka Dusty)
Crunching for Comrades and the Common good of the People.
What PDF?Those are for Barcelona (with reg ram) and in SPecRate...And those russian slides are also for server parts with the same reg. ram and with God knows what scaling going to multiple processors.
You yourself were bashing those specrate tests as unreliable for performance comparisons.So you don't know for sure what IPC gain will K10 have over K8 in int and fp workloads.Just get over it and look forward.
Last edited by informal; 07-02-2007 at 12:35 PM.
Posting Permissions
Bookmarks