![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
AMD Down 4 the Count? Future Map I vs A
- Thread starter MC FLMJIG
- Start date
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
those results were actually kinda surprising. the 2.6ghz K8 managed to stay really close to the 3ghz woodcrest considering how much a 3ghz conroe kills a 2.6ghz K8 in performance, and woodcrest/conroe are basically identical..Donnie27 said:
(cf)Eclipse said:those results were actually kinda surprising. the 2.6ghz K8 managed to stay really close to the 3ghz woodcrest considering how much a 3ghz conroe kills a 2.6ghz K8 in performance, and woodcrest/conroe are basically identical..
I agree... If that Opteron system was 3ghz it would have taken most of those benches..... hmm that is very interesting.
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
yeah, especially with the FSB at 1333mhz, it shouldn't have been tooooo much of a bottleneck compared to amd's coherent ht links. perhaps the shared L2 is causing THAT much of a hit in multithreaded apps? there were quite a few cases where woodcrest was way ahead in the single threaded performance, but by the time it got up to 4, K8 was hardly behind.duby229 said:
so possible causes:
1. shared L2
2. FSB limitations
3. single memory controller vs NUMA for dual K8
4.
either way, it appears that amd can still pull something off. even a 20% increase in ipc and 10% increase in clocks will give it the edge at the dual-dual core level. just imagine what'll happen with 4 quad cores
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
a different memory standard isn't gonna knock off 10-30% of the performance. hell, going from single channel ddr1 to dual channel ddr2 did like.. 5-10%. going from dual channel ddr2 to dual channel fb ddr2 shouldn't be more than 1-3% AT MOST.enumae said:
(cf)Eclipse said:
I clearly dont know as much about memory as you do, but I also remember hearing that FB-DIMM was the cause... Something about latency that the buffer causes...
Maybe it's latency sensitive?
But then again I dont think I'll fall for that line either becouse I read an article about a core2 capable DDR board that performed on par with core2 DDR2 boards. So clearly latency shouldnt matter too much.
Maybe it's bandwidth starved?
Maybe But then again maybe not. It is a dual FSB system, it is more or less 4 dual core dies talkin across two busses.
Maybe it's cache coherency?
Most likely. These systems support MESI, but they do not have the same capability MOESI..... They actually have at least one hop any where it goes, and has do a lot more work to keep the cache coherent through the use of cache snooping. Cache snooping is not required on die, but only when traffic goes off die. I suspect this is the root of the problem right here.
(cf)Eclipse said:those results were actually kinda surprising. the 2.6ghz K8 managed to stay really close to the 3ghz woodcrest considering how much a 3ghz conroe kills a 2.6ghz K8 in performance, and woodcrest/conroe are basically identical..
FBDIMMs causes a hit in certain situations. That's old news and it has Intel looking at DDR3, DDR1 and even RAMBUS again LOL! Plus, some Apps just run better on each architecture=P
http://www.anandtech.com/mac/showdoc.aspx?i=2816&p=11
Anand said:
So, with Two independent FSBs and NOT using FBDIMMs, Bensley would see a boost for sure.
To the others guy, there's not speculation at frickin' all. The Quad Core Intel chips have been tested and overclocked for months not LOL! Please go to Xtremesystems.org.
Donnie27 said:
Keep reaching for those straws....
Latency is not really that bad. At least not bad enough to cause that much disparity....... Like I said I suspect it is a problem with cache coherency, which is probably compounded by bandwidth starvation.
It's when Anand makes comments like that you KNOW he is completely just simply making shit up. Benches just simply dont agree with him. At all..... Anand has a tendancy to make shit up.... He pulls things right out of thin air.... (aka Intel surely pays him well for it too.)
duby229 said:Keep reaching for those straws....
Latency is not really that bad. At least not bad enough to cause that much disparity....... Like I said I suspect it is a problem with cache coherency, which is probably compounded by bandwidth starvation.
It's when Anand makes comments like that you KNOW he is completely just simply making shit up. Benches just simply dont agree with him. At all..... Anand has a tendancy to make shit up.... He pulls things right out of thin air.... (aka Intel surely pays him well for it too.)
Sure, I'll send you an e-mail on my 2 month old system. According to you I would have to wait for another month to do so? No one has to reach for straws with you. If you have a problem with Anand, please kindly take it up with him. You do know that he is NOT the only one saying that, right? Aren't you and Duvie the same person?
Here some more for you to feast on.
http://www.dailytech.com/article.aspx?newsid=4588
EternityZX9
Limp Gawd
- Joined
- Sep 15, 2004
- Messages
- 236
Thanks, drizzt81, at least someone took notice.
I think speculating can be fun an entertaining, but lately it's just out of control.
Donnie27 said:Sure, I'll send you an e-mail on my 2 month old system. According to you I would have to wait for another month to do so? No one has to reach for straws with you. If you have a problem with Anand, please kindly take it up with him. You do know that he is NOT the only one saying that, right? Aren't you and Duvie the same person?
Here some more for you to feast on.
http://www.dailytech.com/article.aspx?newsid=4588
Nope Every site I post on, I go by the same name duby229.
OK, I'm gonna draw a comparison.......
The VAST majority of american news networks sit on the extreme left. The question is does that mean that the vast majority of americans sit on the extreme left? The answer is simply no.
In the same vein..... The VAST majority of computer enthusiast sites are BIASED towards Intel.... After all Intel PAYS them MILLIONS every single year.......... Does that make these !!!!!! sites accurate? Does that mean that the benches they run are encompassing, and complete? Again the simple answer is no. The truth is that they get PAID to write what Intel wants them to write.... Need I mention Babco?
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
i'm not entirely sure how accurate this is, but it's something we can neither prove or disprove.duby229 said:
i'm eating dinner now, but i'll try to produce some ddr2 vs fb ddr2 benches soon
duby229 said:
Doesn't matter!? I'm interested in how you'd make a claim like that....I know that places like Tomshardware come up with some wacky benches, but to be paid off?
Does anyone know if AMD has scheduled anything in the near future to show off Barcelona, or is this too early?
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
perhaps i should have said find. i don't have an fbdimm based rig to do testing on. and hw comes first
MrWizard6600
Supreme [H]ardness
- Joined
- Jan 15, 2006
- Messages
- 5,791
few more price cuts due, a few months apart.Theli said:
absolutly. but then thats a highly subjective question: aftermarket cooing? liquid cooling? stock multiplier? stock NB cooling? system airflow? Same chipsets?
duby229 said:
Until you realize that by removing the credibility of your site (which relies on trust), you will soon lose readers.
I'm just saying that rather than assuming that people are being paid to say things, why not just show some evidence?
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
mine isn't, it's just a hobby for meduby229 said:
http://www.phoronix.com/scan.php?page=article&item=495&num=1
curious.
Perhaps that isn't the best word, maybe "integrity" would be a better one. There is a reason people trust the WSJ over the National Enquirer, and I think there is a degree of similarity between that and sites like Anand, Toms Hardware, [H]OCP...duby229 said:
duby229 said:
You see, nothing is that simple. That same Vast majority of news networks are owned by wealthy folks like Rupert Murock
Even in tiny Little Rock Arkansas Walter Hussman's paper has supported every republican since at Least Nixon IIRC.
brucedeluxe169
2[H]4U
- Joined
- Aug 24, 2004
- Messages
- 3,720
lol, this is getting enteraining
i do have to say that the assertion that hardware sites are being paid by intel to skew benchmarks is just ridiculous
i do have to say that the assertion that hardware sites are being paid by intel to skew benchmarks is just ridiculous
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
guys, this isn't the soapbox in genmay, let's keep politics out of the amd forum, eh?
duby, unneeded
donnie, don't be stupid
bruce, it happens. sad, but true.
with all that sorted out, let's get back on topic..
duby, unneeded
donnie, don't be stupid
bruce, it happens. sad, but true.
with all that sorted out, let's get back on topic..
(cf)Eclipse said:guys, this isn't the soapbox in genmay, let's keep politics out of the amd forum, eh?
duby, unneeded
donnie, don't be stupid
bruce, it happens. sad, but true.
with all that sorted out, let's get back on topic..
Oh please, I'm not falling that very weak childish attempt to start at flame.
Oops, too late. Though, it wasn't flamebait at all. As usual, eclipse is just trying to get people to calm down.Donnie27 said:
----
I'm still amazed that most people can't seem to realize that this roadmap isn't set in stone. People keep saying things like "Intel will have this" and " AMD will release that". Both companies slip; you need to plan on it.
Also, the 45nm process isn't magic, people. What did 90nm -> 65nm give Intel in clocks with the same core? A couple hundred MHz. Why do you think 45nm will be any different? Again, hate to break it to you, but it won't. AMD's 90nm -> 65nm will give them 100 - 200 MHz at most. Why does Intel's 45nm give them a free 400 - 500 MHz? Oh, yeah, because their name is Intel, I forgot.
Again, I really think that anyone with an open mind who looks at the linked roadmap will come to the conclusion that Intel is making unreasonable claims abnout future products. I offered a bet to anyone who really thinks they will deliver and have not received any takers. I think that shows just how much faith people place in the roadmap; funny, because they still defend it. Anyone want to take my bet? I have $100 to the first person to take the bet that says Intel will not be able to do 4.0GHz stock on wolfdale by 3Q2007 as the roadmap claims? Anyone insane enough to take my bet?
----
Bah, unfortunately, we need rules. Once someone agrees to my bet, we both send $100 via paypal (or similar), in advance, to a trustworthy third part member of the forums. When a winner is decided, the third party sends payment to the winner. We need a system such as this, because I have had people refuse to pay me when I win, when I would have paid if I lost. This would keep everyone honest and everything fair.
visaris said:
The only reason I believe Intel will reach those clock speeds is they have a very large R&D budget, and it would seem that they have quite a bit of headroom for the X6800. From what I have read the 3.2GHz number isn't hard to obtain on a stock cooler.
Xtreme systems has Kentsfield at stock voltages on air at 3.2GHz -[Link]- also here is the CPU-Z validator -[Link]-.
Also The Inquirer sometimes gets things right.
In regards to your bet, keep in mind that Intel is working on its own water cooler, add that and the 4GHz isn't really out of reach, and it would still be stock.-[Link]-
visaris said:
You must have a PhD in BS.
Did Intel try to ramp Cedar Mill ? The answer is easy : no.
So altough the 90->65nm gave only 400MHz ( 533MHz 965XE ) it had the potential to go much higher.
But Intel did not beat a dead horse to get up , they waited for Conroe.
I don't buy 4GHz , but 3.66/1333 GHz.
visaris said:
(cf)Eclipse said:
Then it should have been;
duby, don't be stupid, that was uncalled for, don't be a jerk or etc.......
donnie, don't fall for it, don't be suckered into it or etc....
bruce, it happens. sad, but true
I wasn't talking about what (cf)Eclipse said as being flamebait but what Duby229 said, or did you miss that too?