Monday, June 12, 2006

Conroe getting really close to be busted

One month from Conroe launch, we finally have one independent 32 bit benchmark comparison between Conroe XE 6800 2.93GHZ 4MB and Athlon FX62 2.8GHZ 2x1MB. Remember I told you that Conroe will only do better in single threaded loads where the working set is comparable to 4MB cache?

First look at ScienceMark results,the Con E6800 is 9.2% faster than FX62 in Primordia, but they got almost identical scores in Molecular and Cypher (less than 1.5% difference).

Next, in PiFast and SuperPi, E6800 is 36.6% faster in 10M PiFast, and 88% faster in 1M SuperPI. Both PI codes have very little branching and memory usage has high locality.

In Cenema 4D, FX62 and Con E6800 scores were close, 11 and 10 seconds, E6800 is about 10% faster, but the time measurement is not precise enough to determine exact numbers. This is clearly a multithreaded test, as the single core CPUs have about half the performance.

In Kribibench, Con E6800 is over 100% faster than FX62. Seems to be another Super SuperPI case to me.

In Cinebenh 2003 CPU Render, Con E6800 is 47.9% faster than FX62. However, we have seen other 32 bit Cinebench tests, a 3GHZ Woodcrest was 22% faster than a 2.6GHZ Opteron. The version of Cinebench differs, but the discrepency between these two results is too big. A question mark here.

Doom3, Con E6800 39% faster. Doom is single threaded.

Quake 4, Con E6800 9% faster.

FEAR, Con E6800 5.8% faster.

One thing you should notice is that the above gaming tests were all done at low resolution of 640x480. At higher resolutions and high quality, the FX62 beats Con E6700 by 7% in this Intel arranged Farcry test. Higher resolution requires more memory, thus reduces the effectiveness of cache.

The Con E6800 is 2.93GHZ. The E6800 has 4.6% clockspeed advantage over FX62.

Judging from the results, the Con E6800 offers minor speedups in modern games. Except a few simplistic tests, the E6800's lead over FX62 is around 10% in majority of the tests.

What's missing from the picture is those SysMark tests and multi-tasking tests Intel is hiding from us. I expect FX62 to bust Con E6800 in such tests. I like to see some tests with quicker memory (lower latency).

Another missing item is 64 bit performance. Windows Vista is near and we expect AMD64 to achieve 10-40% performance gains running in 64 bit. Mooly Eden's claim of a major performance gap is closer to be busted. Those who bought Intel's hype will be disillusioned.

The FX62 is on sale now at $1390. Intel must be mad seeing that price tag,as its Pentium D chips are selling at 10% of that price. Intel has to wait 40 days to sell Con E6800 at $999. I think by then, AMD will have a few 65nm parts off its test runs just to push Con E6800 a bit.

4x4 will definitely set Intel back by 50%. The 4x4 tag will add a 20% premium to AMD chips, just like SLI cards sell at higher price.

Then comes AMD's next core, with 50% integer performance increase and 300% FP performance increase per core.

AMD has slashed prices on its single core CPUs. An Athlon 64 3500+ AM2 is now only $120 at newegg.com. I am glad that AMD is executing Operation Kill. AMD is going to deny Intel the oppurtunity to unload its legacy inventory. I expect AMD's ASP to fall by 15% in 3Q06 to $84 due to massive price cuts on desktop chips. However, both FA36 and Chartered FAB7 are cranking, AMD's capacity should increase by at least 40% from the 1Q06 level. The Turion 64 X2 and Opterons should continue to give AMD some nice profits.

52 Comments:

Anonymous Anonymous said...

The first Conroe benchmarks put the XE at 40% over the FX-62.

Then we started seeing bechmarks showing the XE at around 20% over the FX-62.

A month or so later this was down to a 15% lead.

Now we see that the lead is around 10%.

The performance of Intel's chips is supposed to increase as the silicon nears production, not decrease. Conroe will be around 10% faster than AMD at launch. All AMD needs to do is drop prices by 15% and they will lead the price/performance ratio. Next comes 65nm with massive clocks, then K8L; Intel is about to be fragged.

12:39 PM, June 12, 2006  
Anonymous Anonymous said...

"In Kribibench, Con E6800 is over 100% faster than FX62. Seems to be another Super SuperPI case to me."

Kribibench is multithreaded (cache won't get owned by one core only), and is demanding on brute SIMD performance.

Conroe is able to do 4 FP multiply+add ops every cycle, twice the K8, that's it.

Anyway, just like SuperPi, Kribibench is known for stressing only a subset of the cores' execution units. It's no ScienceMark.

12:47 PM, June 12, 2006  
Blogger Mad Mod Mike said...

What's most interesting is the performance in gaming of each Conroe, the 2.4 is performing just like the 2.9, so why would anybody even buy a extreme? Intel is going to run into problems, people who aren't too much of an idiot will buy the 2.4, overclock it, and all their other expensive chips nobody will buy. Intel is hurting themselves more with Conroe than it will help.

It's also interesting to see how the lead has fallen from 40% to Conroe actually losing some tests to the FX-62, it appears the wonder chip Intel fanboys love is no more. Even on Cinema4D, almost all the Conroe chips perform the same, even with 600MHz clock speed difference between them.

Important to note is the AMD system had 1GB RAM where the Conroe had 2GB, which is odd. The RAM used was 5-5-5-15 DDR2-1000....again, if you really look at it, the FX-62 would be beating Conroe on every test including DOOM 3 if it had 2GB RAM and 3-3-3-8 timings, so this test is another fraud BS that attempts to make Conroe look better than it is against the competition.

1:18 PM, June 12, 2006  
Anonymous Anonymous said...

what is the next AMD core specified? rev G or K8L that?

sunhing

1:21 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

the 2.4 is performing just like the 2.9
Clear indication that Conroe's performance is derived from Cache. Clockspeed doesn't matter that much.

1:25 PM, June 12, 2006  
Anonymous Anonymous said...

Pay attention that Quake 4 is quite possibly graphically limited in this test, just as are F.E.A.R. results.

I wonder if Intel really cheated on those early PuppetTech benchmarks. Looks like the picture is sunny for AMD. The consumer is the winner in the short term while the price war rages. Later on we might get ripped off by AMD if Sharikou is right abut Intel. Think massive layoffs, restructurisation, manufacturing reorganisation. Intel would be like a whale that has to adabt to life on land.

1:29 PM, June 12, 2006  
Anonymous Anonymous said...

WOW, you guys just don't stop.

Try this... Find a positive to Conroe.

There is no way that Conroe is as terrible as you state, even if its just 10% over an FX62, thats still a hell of alot better than netburst.

If you can not then you are purely biassed and have no business trying to inform anybody of what is to be available.

1:33 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

Try this... Find a positive to Conroe.

It is no small feat for Intel to catch up with K8 in 32 bit performance. Let's wait and see some more tests and some 64 bit tests. 64 bit is quite relevant, don't you think? You don't want to buy a CPU that will run Windows Vista 20 slower.

1:41 PM, June 12, 2006  
Anonymous Anonymous said...

I completely agree that 64bit is the future. I am also looking forward to more test, especially 64bit.

I will not pretend to know as much about chip archiecture as you do. That is why I read your blog, but a biassed opinion is not always informative.

Mad Mod Mike made a good point in regards to the 2.4 being equal to the 2.9 in gaming.

Can it really be because of the cache?

If it is then you may also be right about the 2MB cache performance that has still yet to surface.

Keep writing and I'll keep reading.

1:50 PM, June 12, 2006  
Anonymous Anonymous said...

Not exactly cache, since its speed scales along with the cpu, but the prefetch subsystem that loads data into it.

If the prefetcher performs too well for lower clockspeed Conroes, it will suck a lot off the available bandwidth, leaving few advantage to the higher clocked models for playing with.

1:50 PM, June 12, 2006  
Anonymous Anonymous said...

Guys, pay attention to those benchmarks. (640 X 480) !

In those (640 X 480) benchmark, Conroe has around 10% -25% performance lead. I saw some game benchmarks when the resolution is up to (1028*768), the performance lead has come down within 10%. I suspect the cache makes the difference.

If I spend $3,000 on a game machine, I defintely not buy it for games with resoultion of 640 X 480. Conroe does not look like an Operton killer.

1:52 PM, June 12, 2006  
Anonymous Anonymous said...

Conroe systems use sATA HD, but older intel and AMD systems don't. However I ignore how much influence that has on those benchmark performances

1:52 PM, June 12, 2006  
Anonymous Anonymous said...

Well, if Intel hadn't hyped the Conroe so much with it's 40%+ performance claims, no one would be quite so down on it now.

Sucks when the pendulum swings the other way.

1:52 PM, June 12, 2006  
Blogger Ajay S. said...

"There is no way that Conroe is as terrible as you state, even if its just 10% over an FX62, thats still a hell of alot better than netburst."

Intel fan boys are welcome to bow before the altar of Paul Otenelli and get ripped off at the temple of Intel. Ever since the Intel made review sites press buttons at the IDF, it has been one biased review after another.

With every little standardistion of the test benches, the performance diff with AMD keeps dropping. So much for the 20-40% performance jump over AMD. Is this all Intel can do with their brand new core and 65nm process?

Lets see how conroe holds up in a full review that also does 64bit in a unbiased setup. Guess the chips will have to be available in the open market wihout NDAs before anybody can do that. But how many sites will be left do a fair review that does not contradict their own earlier "review"

A company I know recently bought a bunch of Xeon racks. After a few hard disk failures and system shutdowns within a month, it occured to the perplexed admins that they could actually bake something with the heat that was coming out of the new systems.

For such idiots who eat whatever Intel throws at them, Conroe and woodcrest should be a nice promotion. Not that it makes a diff to them anyway.

As for which architecture is superior, Conroe is the final proof that the chip giant with 1,00,000 employees has lost its vigour.

Intel finally has a product that has a 5% to 40% performance advantage depending on who did the review and when. Sad to see that the comparison is with a 3 year old design begin manufactured on a one generation old process.

2:54 PM, June 12, 2006  
Anonymous Anonymous said...

Just found this benchmark today too.
Intels performance lead shrinks everyday, by the time it really comes out, there wont be any left.
Note that the test systems were not exactly the same, and it could be better (and definatly the ram, cl4 isn't that hard to find these days).
And why do they allways test (only) low quality? I doubt that anyone runs an fx on that quality level. Oh yes thats true, conroe performs better on that level.
And allways those wierd benchmarks. I read a lot of reviews and some of these benchmarks I rarely see.

Does anyone really know if you use a low latency low speed ddr2 (like 533mhz cl3) instead of high latency high speed ddr2, what the difference would be, i haven't found any benchmarks that try this. I thought i read somewhere, that am2 doesn't fully uses the high bandwith, so it would much more benifit from low latency.

3:19 PM, June 12, 2006  
Anonymous Anonymous said...

Nice! Intel's new chips are faster, cheaper, use less power, and run cooler. I can't really see any negatives there.

3:58 PM, June 12, 2006  
Blogger Mad Mod Mike said...

http://www.legitreviews.com/article/347/3/ - proves lower latencies and more RAM give AM2 20% boost in performance.....hear that? That's Conroe being knocked around by the FX-62, but it's behind closed doors....for now.....

4:26 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

But how many sites will be left do a fair review that does not contradict their own earlier "review"

Go read my "Busting Anand" article, read AnandTech's self-defense and how they care more about saving face than the truth. If I were them, I would admit the mistake and correct it. No one knows everything and no one is always right.

4:39 PM, June 12, 2006  
Anonymous Anonymous said...

".....hear that? That's Conroe being knocked around by the FX-62, but it's behind closed doors....for now....."

LMAO

4:48 PM, June 12, 2006  
Anonymous Edward said...

"Nice! Intel's new chips are faster, cheaper, use less power, and run cooler. I can't really see any negatives there."

Intel's new chips are 10-15% faster, that's it. The rest are just marketing make-believe.

It might not be cheaper with AMD's pending pricecut, and it certainly does't use less power when including the northbridge.

5:30 PM, June 12, 2006  
Anonymous Anonymous said...

mmm, that legit reviews PROVES NOTHING when it comes to comparing to conroe. as i have told you before, that review only proves that memory timing,speed, and size MATTER. there was not a single conroe system in that review and no comparison can be made. you say 20% faster, great, but what if the 20% does not come close to conroe, which it probably does not. 20% increase over the slowest config in a particular case, so what? what if the slowest is 30% slower than conroe, then things are not looking great. why don't you stop using that review and find one review that proves conroe is just barely on par for everything. i haven't seen a review like this, has anyone else? also these "reviews" of conroe in which the reviewer does not even have a conroe system does not count, and none of this theoretical math crap cause it is too easy to twist the truth to suit the motive.

5:32 PM, June 12, 2006  
Anonymous Edward said...

"Lets see how conroe holds up in a full review that also does 64bit in a unbiased setup."

Mind you, the more powerful the architecture is, the more it is subject to (mis)optimization.

My first-hand experience shows 64-bit encryption on Athlon64 be 60% faster on some algorithms, but 40% slower on others. The same implementation, the same everything else, but one reason: optimization.

Benchmarking is hard. Fair benchmarking is impossible.

5:34 PM, June 12, 2006  
Anonymous Anonymous said...

Mad Mod Mike and Sharikou read this article and explain wtf AMD's processors only gain ~5% from the lower latencies. http://www.simhq.com/_technology2/tech
nology_085a.html

I have always thought that simulations are the most CPU bound games and I would thus expect AMD to shine with the lower latencies but it is now obvious to me that the upgrade to CAS3 certainly isn't worth it especially considering it means doubling the price paid for the memory.

6:28 PM, June 12, 2006  
Anonymous Anonymous said...

"You don't want to buy a CPU that will run Windows Vista 20 slower."
i really can't find any benchmark supports that besides a white paper released by Microsoft. do you care to back it up?

6:42 PM, June 12, 2006  
Anonymous John said...

"What's most interesting is the performance in gaming of each Conroe, the 2.4 is performing just like the 2.9, so why would anybody even buy a extreme?"
humm... both of them have performance leads up to 50%, while cost 50% cheaper. the question you should really be asking is... "why would anybody even buy an AMD FX?"

"Important to note is the AMD system had 1GB RAM where the Conroe had 2GB, which is odd."
it is because the test was done in Computex.

"the FX-62 would be beating Conroe on every test including DOOM 3 if it had 2GB RAM and 3-3-3-8 timings, so this test is another fraud BS that attempts to make Conroe look better than it is against the competition. "
um... how about no? if you increase the timing on FX's system, don't forget to increase the timing on Conroe's system too. it is not likely for FX to beat Conroe, even with 3-3-3-8 timing.

"Clear indication that Conroe's performance is derived from Cache. Clockspeed doesn't matter that much."
somehow the benchmarks tell me the otherway.

"Intel fan boys are welcome to bow before the altar of Paul Otenelli and get ripped off at the temple of Intel."
AMD fanboys are welcome to lick their wounds at the asylum of sharikou.

ok.. let's be nice to AMD fanboys. even IF the FX-62 outperforms Conroe XE (by 5%, according to your logic), wouldn't that be a "must" thing? FX-62 cost roughly $200 more than XE, and that's without the expensive low latency RAM. however, most benchmarks show that FX got annhilated by XE.

7:08 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

ven IF the FX-62 outperforms Conroe XE (by 5%, according to your logic), wouldn't that be a "must" thing?

Well, if you want 4x4, you got to have AM2

7:15 PM, June 12, 2006  
Blogger Mad Mod Mike said...

"ok.. let's be nice to AMD fanboys. even IF the FX-62 outperforms Conroe XE (by 5%, according to your logic), wouldn't that be a "must" thing? FX-62 cost roughly $200 more than XE, and that's without the expensive low latency RAM. however, most benchmarks show that FX got annhilated by XE."

There aint no cure for stupid.

7:44 PM, June 12, 2006  
Anonymous Anonymous said...

"Well, if you want 4x4, you got to have AM2" Errmm... out of topic here!

Did you noticed that they used the mid range 965 chipset mobo? Did you noticed that this is a single card setup (performance is becoming GPU limited instead)? The previous better Conroe benchies were done on 975 chipset mobo and with SLI/Crossfire...

Conclusion is.. Conroe still pulled ahead of FX and costs much cheaper!

7:54 PM, June 12, 2006  
Anonymous Anonymous said...

"http://www.legitreviews.com/article/347/3/ - proves lower latencies and more RAM give AM2 20% boost in performance.....hear that? That's Conroe being knocked around by the FX-62, but it's behind closed doors....for now..... "

No offense, but could you try to be a little bit less misleading? Someone has already pointed out the following benchmark:

http://www.simhq.com/_technology2/technology_085a.html

They tested at DDR2 800 with 4-4-4 timings and 3-3-3 timings. I believe the performance benefit is pretty much always below 5% with around 3% being the norm.

Furthermore, even in the Legit Review article they show no such 20% benefit. It seems that people are always trying to point out the bias and unfairness of the Conroe reviews, but the least you can do is uphold your own high ground. Even if you took the X2 4800+ results on DDR2 667 at 4-4-4 timings and compared it to the X2 4800+ DDR2 800 3-3-3 timings you only gain 15% in Doom 3. However, touting that 15% benefit around to prove the superior performance of the FX62 is a lie. Even the Anandtech review that is taking so much flak uses DDR2 800 at 5-5-5 timings. This means we will never see that 15% improvement just from RAM much less 20%.

What's more, looking within DDR2 800 timings we see the following quote from the Legit Review page you linked.

"With a 6% performance gain from using low latency memory it really goes to show how big of a role memory plays on these new AMD AM2 processors thanks to the fact that the memory controller is on the processor."

Moving from DDR2 800 5-5-5 to 3-3-3 gains you 6%. However, even that is misleading. Why? The 5-5-5 timings was down at 2T and the 3-3-3 and 4-4-4 timings were down at 1T. They introduced 2 variables which artificially inflated the latency benefit for AM2. This means that the real benefit of moving from DDR2 800 5-5-5 to DDR2 800 3-3-3 is a lot closer to the average 3-4% found by simhq. Those are the facts, and to keep whining about how lower latencies will yield 20% benefits is unrealistic and deceptive.

From the original blog post:
"One month from Conroe launch, we finally have one independent 32 bit benchmark comparison between Conroe XE 6800 2.93GHZ 4MB and Athlon FX62 2.8GHZ 2x1MB"

Now I know you've questioned some of these, but officially there have been 4 independant reviews up to this point. 5 really since the TweakTown review was down with Hard Info so it could really count as 2.

They are in order of appearance:

http://www.hexus.net/content/item.php?item=5692

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2771

http://firingsquad.com/hardware/intel_core_conroe_benchmarks/

http://www.tweaktown.com/articles/914/benchmarking_intel_conroe_core_2_x6800_e6700_and_e6600/index.html

The results are starting to filter in. It should also be noted that it's possible the reason why Conroe appeared to do poorly in the games and didn't scale very well was because it was graphics card limited. It may be unlikely, but for the sake of argument TweakTown only used a 7800GTX with 256MB of RAM while all the other reviews up to date have at least used a 7900GTX 512MB while Intel of course likes to use a X1900XT Crossfire combination. More reviews, especially 64-bit ones with optimal configurations for both systems are of course still needed.

For Sharikou, I was hoping you could respond to AMD's recent massive price cut announcement.

http://www.hkepc.com/bbs/itnews.php?tid=613657

I know you may feel differently, but a 50% price cut for the X2 doesn't inspire confidence. You were strongly criticising Intel for their price cuts on the Pentium Ds, and now it seems AMD is doing the same. They may be in an even worst position since at least Intel has new products coming in to justify it while AMD is just cutting prices. AMD is probably in an even worse position than Intel to handle these cuts since Intel is using 65nm and 300mm wafers while AMD is still mainly using 90nm and 200mm wafers. I'm not sure what they're contract with Chartered is but they may be paying Chartered a fixed price per chip which would mean that AMD is now eating the difference there too. I know you may point to 65nm production, but I don't think it's to wise to rely on that too much. Even AMD has pointed out in their recent Technology Analyst meeting that they won't be having any 65nm production shipments ready until Q4 2006. This would be perfectly inline with their December launch plans. It's not likely that they would miraculously turn out with 65nm parts in the next few months. Certainly not enough to offset the cost of these major price cuts. Chartered itself won't even begin to transition to 65nm until mid 2007 so we shouldn't rely on anything from them until 2008.

8:04 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

http://www.hexus.net/content/item.php?item=5692

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2771

AnandTech is a proven paid pumper. Hexus.net is very suspicious. These IDF button pushers got quoted by Intel as proof that Conroe took the crown in Intel's powerpoint to WS. I bet they got paid somehow.

As for AMD's price cuts, so far we have only seen cuts on single core. Cuts on dual core is to execute the "kill" strategy -- preventing Intel from dumping legacy parts at a good price. Most customers shop on price. Those Pentium D 945 at $160 will be attractive to many. AMD has to respond. I suggested this long time ago: a low priced X2 3800+ is enough to halt all Pentium D sales.

8:12 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

I recall I saw the FS test. But there wasn't any system config info, and the test was a kinda of quicky.

8:17 PM, June 12, 2006  
Anonymous Anonymous said...

"There aint no cure for stupid. "
no wonder you got banned from the forum...

8:17 PM, June 12, 2006  
Anonymous Anonymous said...

I just came back from Computex. Strolled around the intel floor. Lots of NGMA BS. I happened upon one Clownroe demo that was showing persistent utilization and temperature.

There were two Clownroe desktops, one running at 60% and one running at 80% CPU utilization. These were 2.8GHz devices.

What was shocking was they running at 58.5 and 69.8 degrees C!!! Holy crap! I run my FX-57 wide open for a day and if it gets to 50 degrees I'm really checking my fans, heatsink, ambient air temperature (I'm worried). Happened once when we let the house get to 85 degrees F. And then it was running wide open.

69.8 at 80%??? I told the intel pawn that this was a very, very, very bad demo if you're showing temperatures that high. Honestly. That chip is dead in a year or less at that rate.

But AMD stock will continue to suffer because intel will continue to dump. Pure and simple. intel sucking is actually bad news for AMD. Oh well, I still want to see that pig die!

8:25 PM, June 12, 2006  
Anonymous Anonymous said...

"What was shocking was they running at 58.5 and 69.8 degrees C!!! Holy crap! I run my FX-57 wide open for a day and if it gets to 50 degrees I'm really checking my fans, heatsink, ambient air temperature (I'm worried). Happened once when we let the house get to 85 degrees F. And then it was running wide open."
did you also see Conroe being passively cooled on the far left?
however, i did see the high temperature Conroe was radiating. i think there was something wrong with the temperature monitor.

8:30 PM, June 12, 2006  
Blogger Sharikou, Ph. D said...

however, i did see the high temperature Conroe was radiating. i think there was something wrong with the temperature monitor.

I couldn't help laughing....

8:31 PM, June 12, 2006  
Anonymous Anonymous said...

The Core 2 has not killed 5000+ in these tests. The closer comparison pricewise s 5000+ which put up some really respectable numbers at $400 less than FX62

8:58 PM, June 12, 2006  
Anonymous Anonymous said...

http://www.dailytech.com/article.aspx?newsid=2800

I saw this today and I got excited.

That said I read elsewhere that conroe will be delayed, so how long do i have to wait for the price cuts?

oh and from all the "benchmarks" I have seen here and elsewhere, it looks as thou Conroe has caught up and now beats the A64 (clock for clock) which is good for intel, its a shame it took so long. Anyway with new pricing and a cpu that is competative to say teh least it will be good for the consumer.

I personally plan on buying an AMD 64 AM2 4200+ in the near future (given price range). Even if Intel is 10% faster. I like AMD's tech and given upcoming price changes those like me that live outside the US will be able to afford. on and bring on the high res/64bit os benchmarks as thats where the future lies :)

Danny (AU)

9:25 PM, June 12, 2006  
Anonymous Anonymous said...

Conroe getting close to being busted ? So far it has hardly lost a bench, summary of this review

"Not once did AMD’s current highest performing and most expensive FX-62 processor come close to beating the Conroe. "

They must be paid Intel pumpers as well...

A while ago it was said that AMD do not have to slash prices as they had the better product, now it's said that it's great that they have slashed prices. There is no consistency here apart from AMD is always right. That sort of stance is untenable.

The other good thing about Conroe is it's overclockability. Now stepping 5 production chips are appearing it seems 3.8Ghz on air is a distinct possibility. 3.3Ghz on air for FX-62 it seems is about the limit so as well as the 25% you can add for Conroe over Fx-62 add another 15% for the overclocking part as well.

AMD will certainly need 4x4 and then it will win in all those applications that require 4 cpu's ...

Oh dear.

10:46 PM, June 12, 2006  
Anonymous Anonymous said...

The biggest problem of Conroe, I think, is that it will not be available until the fall at least. Alas for Intel's fanboys.

1:04 AM, June 13, 2006  
Blogger duploxxx said...

"Not once did AMD’s current highest performing and most expensive FX-62 processor come close to beating the Conroe. "

well i am happy for conroe that he can beat a 3 year old design.... lets see in 3 years from now.

when you check the 64bit applications the performance gain of a conroe is less then 20% the amd to almost 40% so conroe is no longer faster then a k8 clock/clock.

OC you certainly have a point, but then you come to the fact that all pre release and se models are most of the time the best ones (marketing hype....) i had one of the first venice cores that was introduced a 0512dpmw there is still yet none that can beat these.

fyi on the 4x4 a standard fx62 will outperform any clocked conroe available when the program supports multicore. how do i know: looking at 3dmark cpu score my opteron 280 system is faster than an oc'ed conroe at 3.8. clock to 2.8 gives a 20% faster performance. kyoson on XS did a run ln2 on 4,34GHZ. givin g aprox 3700points. so i need a 2.95 opteron to beat that score on air..

ok it might cost more, but thats a stable all day rig, no high level oc knowledge needed

2:30 AM, June 13, 2006  
Anonymous Anonymous said...

Dump Intel shares and run, says analyst
Dead money

http://www.theinquirer.net/?article=32362

What a title???

5:01 AM, June 13, 2006  
Anonymous Anonymous said...

Well, well, Conroe is not that great after all. It runs fast on low-resolution games, but almost at par with FX-62 in high-resolution games( <5%)

http://www.tomshardware.co.uk/2006/06/05/first_benchmarks_conroe_vs_fx-62_uk/page5.html

You want me go out and buy those expensive CPUs for 640X480 games. Get real !

9:16 AM, June 13, 2006  
Anonymous Celebrandil said...

So Intel will gain the performance crown after moving to 65nm? Wouldn't it be rather surprising if they didn't? Intel might have a few nice quarters ahead of them, but the future definitely speaks in AMDs favor. Processing cores hunger for bandwidth and we'll soon reach a point when bandwidth means everything. You might get great scores in artificial CPU benchmarks, but in the end of the day you want to do some processing, processing on images, videos, 3D models, etc. The only reason why Intel still is in business is because they have been quicker making transitions from one manufactoring technology to the next. I give them credit for that.

Analyzing the performance of an architecture isn't easy. Some applications are memory bound, others are CPU bound. In some cases you may fit your memory requirements to the caches, sometimes not. Caches do matter, but only as long as you can exploit them. In multi-threaded scenarios optimizing your cache utilization becomes a nightmare, and as more cores are added it soon becomes impossible. Most of the benchmarks I've seen so far are either fundamentally CPU bound or fit perfectly into the caches. Few benchmarks are on applications that are fundamentally memory bound. We need more multi-threaded memory bound benchmarks to really know.

We might end up in a situation where Intel is preferable for some applications, while AMD is suitable for others. I believe that is good news for consumers.

2:33 PM, June 13, 2006  
Blogger Entrophos said...

I just want the Intel fanboys to admit what the Con-roe really is:

A suped-up Pentium-M with another core duct-taped onto it and a huge cache to make up for their aging memory-cpu interface technology.

Yes, it looks great in artificial benchmarks. Yes, it seems to run single-threaded 32-bit applications quite well (assisted by the huge L2 cache). Congratulations Intel for making a CPU that can finally compete with AMD on these specific applications programmed and compiled for legacy 32-bit CPUs on a single processor system.

Unfortunately, these applications are short for this world. The future is 64-bit SMP processing and Vista is going to pave the way for the masses. AMD has the technology lead here (better 64-bit implementation, better SMP performance, HTT, etc) and Intel is playing catch-up.

I fully expect that the Conroes will suddenly lose a lot of their appeal when we see more fully optimized 64-bit code designed for SMP become more common as it will leave them in the dust, as AMD64 CPUs benefit more from this environment than they do. Considering that in this environment the current AMD lineup will probably outperform them from what little we've seen of benchmarketing, what the hell is Intel going to do when Rev G. and K8L come out?

Buying a Conroe now might get you a few more FPS on a 32-bit non-SMP optimized game in Windows XP today, but what will it do for you tomorrow? You're buying an obsolete piece of technology right out of the box to feed a slimy megacorp that cares more about shifty marketing than true innovation.

11:05 AM, June 14, 2006  
Anonymous Anonymous said...

Its running on 640 * 480 because they are testing the CPU if it was running at a higher resolution it will be GPU bound. That is why all graphics tests for CPUs are done at 640 * 480.

Way to inform the crowd sharikou and Mad mod mike were you planning on leaving these fools to their ignorance.......

11:07 AM, June 14, 2006  
Anonymous Anonymous said...

Conroe was never touted as being 40 % better than any amd, if you read properly without any fan boy bias it was 20 % better than amd and 40 % better than the current netburst architecture which it roughly is. Do AMD fan boys not know how to read now as well....

11:09 AM, June 14, 2006  
Blogger Mad Mod Mike said...

"Conroe was never touted as being 40 % better than any amd, if you read properly without any fan boy bias it was 20 % better than amd and 40 % better than the current netburst architecture which it roughly is. Do AMD fan boys not know how to read now as well...."

Please show me a fair benchmark that shows Conroe beating an equally clocked AM2 w/ 3-3-3-8 DDR2-800, uh oh...you can't!

12:13 PM, June 14, 2006  
Blogger Entrophos said...

Conroe was never touted as being 40 % better than any amd, if you read properly without any fan boy bias it was 20 % better than amd and 40 % better than the current netburst architecture which it roughly is. Do AMD fan boys not know how to read now as well....

Gee that's funny because in the current round of benchmarketing released to the masses, it has less than a 20% advantage. Actually more like 10%, and loses some benchmarks that are optimized for SMP and exceed the huge 4MB cache.

As for an AMD fanboy, I'm not particularly, although I use its products at home and I enjoy the fast performance and value that they represent (e.g. AMD64 x2 3800+ O/C to 2700MHz on air @ 40C peak).

I do, however, despise that Intel stands for a lot of the worst that the corporate world is capable of. It is a clasic example of what happens when you let idiot marketers and shifty sales execs take over a company that should be driven by engineering and innovation.

12:27 PM, June 14, 2006  
Anonymous Anonymous said...

Hmm... I'm all for AMD's AM2 offering being superior to the Intel, and making tests level. I know this is a pro-AMD website; shoot, I haven't even bought/made a computer with an intel chip for myself since '97. But let's give the conroe architecture at least some credit. Now, if you're talking about the quality of the marketing/placement of conroe... ;-)

2:22 PM, June 15, 2006  
Anonymous Anonymous said...

Its running on 640 * 480 because they are testing the CPU if it was running at a higher resolution it will be GPU bound. That is why all graphics tests for CPUs are done at 640 * 480

It's not true. When a system is tested at high resolutions, it is tested as a whole (CPU, Memory Subsystem, Graphics), not just a cachesize of CPU. That is why Intel is so afraid to publish this benchmarks.
When somebody buys Ferrari, he expects driving at least 90mph not 50mph.

In reality nobody knows now, how Conroe behaves under heavy loading or in multi-threaded environment ( I mean not just multi-threaded apps but whole OS, perhaps the shared cache will lead to large penalties).

12:27 AM, June 16, 2006  
Anonymous Anonymous said...

So you say that Intel and every credible benchmarker misrepresented numbers.. Ok let's believe that for a second.. Now, Can you tell us about your PhD Sharikou?? Be honest..

9:23 PM, June 17, 2006  
Blogger Maurice said...

Take your hat off before you talk through it next time. As a disinterested observer watching "fanbois" subjectively invent their own facts is a little tired. Conroe is well and truly out and about now and showing it is a very fast and stable cpu, I don't get why a "fair" observer would be afraid to acknowledge this.

2:48 AM, October 02, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home