E6400 & X2 3800+ Discussion

[Kougar]

Since the original thread owner was getting annoyed at having his thread badly hijacked, I moved my reply here. ;-)

Smart move but you might have to wait a while before an DX10 are actually out and also lets hope Vista don't bomb as Vista isn't all that different that XP other than the bling factor; early tests show XP a bit faster but once DX10 hardware is out that could all change.

I am also curious what chip are you going to buy, AMD or Intel? Ever thought about the Core 2 Duo (Conroe)? It is damn fast.

Vista is pretty different under the hood, which people seem to ignore... infact very different with DX10, the completely redesigned tcp/ip stack, and how the kernel is kept better isolated. And regardless of all the issues it has on slower hardware that it was never designed for, Vista is also showing some pretty good performance improvements with x64-bit processors as well. I believe some people were saying similar things about XP and even older hardware... And if you want to use DX10, then you must upgrade to Vista. MS stated they decided it was not worth completely rebuilding XP from the floor up to work with DX10, and I think it was by far the better move. If they had tried it'd probably have always had a few bugs from forcing the compatibility.


And not to be rude, but don't go randomly saying things without doing the research Werty!

E6300 will be $183
E6400 will be $224
E6600 will be $316
E6700 will be $530
Prices in lots per 1,000

For the price getting a E6700 isn't justified for an extra 50FPS when a X2 at half that price is good enough as long as a fast video card it used with it. You won't notice 100FPS compared to 150FPS.

You need to be specific and not overgeneralize in your array of replies... I just grabbed this one of the lot to pick apart. ;) For what game are you comparing? If you were comparing Oblivion, any FPS improvement would be very much needed I think! I'm not sure you should be comparing the 2nd highest Conroe offering to the 2nd lowest X2 processor either... So back to the $183 E6300 then...

While the slashing of the cache will kill it's performance leads, a E6300 will still outperform a $170 X2 3800+, and that $14 price difference you will easily recover on your monthly power bill alone in energy savings since Conroe is a more efficient processor.

If you are thinking about the Energy Efficient line of X2 processors about now... well, those will not have any price cuts and would cost $323 or $364 depending on the TDP envelope you want.

Considering Ocing, I can only surmise from the results I've seen to date, but that bin part for Conroe part should at least reach 2.66ghz if not further, and would only pull further ahead of a X2 3800 as far as OCing comparisons go. ;)

[vfrex]

One thing that concerns me in this whole comparison is that I haven't seen benchmarks from the lower end conroe chips. As you mentioned, the decrease in cache certainly won't help the performance. If we are going to talk about performance in the lower range though, how accurately can we predict how those cache and clock cuts will change relative performance?

[werty316]

Research??? What research do I need to look up now eh? As of the release of beta2 it has no performance advantage over XP. There was a comparision done. Looks like I touched a nerve indirectly.

One thing that concerns me in this whole comparison is that I haven't seen benchmarks from the lower end conroe chips. As you mentioned, the decrease in cache certainly won't help the performance. If we are going to talk about performance in the lower range though, how accurately can we predict how those cache and clock cuts will change relative performance?

Most users want to see the best and how they perform compared to the competition but expect a full comparision of the entire Core 2 Duo range of CPUs once they are released but here is a bench of the E6400:

http://www.digit-life.com/articles2/cpu/intel-conroe-2-13-ghz.html

[Das Capitolin]

One thing that concerns me in this whole comparison is that I haven't seen benchmarks from the lower end conroe chips. As you mentioned, the decrease in cache certainly won't help the performance. If we are going to talk about performance in the lower range though, how accurately can we predict how those cache and clock cuts will change relative performance?

You will eventually see those comparisons, but until the chips are mainstream you will only see the flagship models compared to the flagship models of the competition. This is pretty common on net chips. Lower end comparisons are an afterthough.

[werty316]

I guess some people expect benchmarks of products when they aren't even in the market yet ;)

[vfrex]

*shrug*

First of all, the benchmark you linked to is tainted. In the first paragraph, it states that the sample has twice the L2 cache than what the actual retail version will. It is effectively just an underclocked E6700. Second, I don't knoww what you mean by "expect" benchmarks of products. Intel has been releasing samples of their flagship processors for weeks now for the purpose of benchmarking and reviews. If their architecture is actually as superior as their PR claims, why wouldn't they release REAL samples of their lower end Conroes? Are they scared they will confuse the consumer? Cannibalize their lines? Or maybe the lower end processors just aren't that good.

[Kougar]

*shrug*

First of all, the benchmark you linked to is tainted. In the first paragraph, it states that the sample has twice the L2 cache than what the actual retail version will. It is effectively just an underclocked E6700. Second, I don't knoww what you mean by "expect" benchmarks of products. Intel has been releasing samples of their flagship processors for weeks now for the purpose of benchmarking and reviews. If their architecture is actually as superior as their PR claims, why wouldn't they release REAL samples of their lower end Conroes? Are they scared they will confuse the consumer? Cannibalize their lines? Or maybe the lower end processors just aren't that good.

Thank you, you spotted what I was just about to point out! ;) The Engineering Sample E6400 in that test, and also the not linked to gaming tests they have done has the full cache size enabled.

However, to my knowledge Intel has NOT released samples for the intention of benchmarks and comparisons (yet). They were released for testing, compatibility purposes, and to make sure there were no last minute bugs like that last infamous floating-point flaw that was missed by Intel... There is still a NDA in effect, which is why there should be a pretty large second wave of "official" benchmarks of every kind rolling in as soon as the NDA on Conroe lifts. Of all the unsanctioned benchmarks I have seen I have not found a 2mb cache part yet, either engineering sample or retail sample.

If you doubt the validity of the benchmarks though, there are pages of unofficial and unstanctioned independant benchmarks mixed with official benchmarks run under Intel's supervision in an Intel facility. http://forumz.tomshardware.com/hardware/Collection-Conroe-Data-Core-Duo-Core-Extreme-ftopict183765.html (http://forumz.tomshardware.com/hardware/Collection-Conroe-Data-Core-Duo-Core-Extreme-ftopict183765.html)

I'll admit trying to wade through that first post was a problem when I first found it, now it's probably impossible... :mrgreen: However if you follow through those 20 pages you'll find the vast majority of the links I mentioned above...

[mousiness]

theres almost no way conroe chips will have lower benchies than X2's, theyre just too well built just like AMD vs. Intel currently, no matter what even without benchmarks of the slower conroe chips by the looks of the e6700 benchies you can tell conroes are just too fast for comparison, thats why ive been waiting for so long now

[Kougar]

theres almost no way conroe chips will have lower benchies than X2's, theyre just too well built just like AMD vs. Intel currently, no matter what even without benchmarks of the slower conroe chips by the looks of the e6700 benchies you can tell conroes are just too fast for comparison, thats why ive been waiting for so long now

I wish! The FX62 is about par with a 2.4ghz E6600 processor, although the E6600 will win the good majority of the benchmarks. Not even mentioning several of the newer samples can OC to 3.1ghz on stock voltage, where it becomes a not a completely different ballgame, but a completely different ballpark... My only question here is will Intel continue the trend with retail chips at this bin level, or will Intel modify the chip binning differently for one reason or another... Once Conroe hits retail channel I expect half the well-known review sites will go out and buy retail chips just to figure this out... :)

I fairly well expect the lower cache E6300, E6400 chips to hold their own around their price points, but beyond that I'm not sure if they'll be quite the same high-performance chip the 4mb cache Conroe is.

[mousiness]

and that 4mb conroe is the only conroe anyone should buy, at least for gamers, its amaizing OC'able, super fast even at stock voltages, and is much much faster than any AMD at time of release, but i agree with you, reviewers will be buying the lower end retail chips to figure out the performance capabilities compared to AMD's, and even for the e6600 equalling the fx-2, whoa thats quite a leap for intel, i never knew AMD's most valuable and fastest processor is equal to the lower end e6600, and kougar, is by any chance the e6700 the EE of intel? or will they make another EE?

[Kougar]

and that 4mb conroe is the only conroe anyone should buy, at least for gamers, its amaizing OC'able, super fast even at stock voltages, and is much much faster than any AMD at time of release, but i agree with you, reviewers will be buying the lower end retail chips to figure out the performance capabilities compared to AMD's, and even for the e6600 equalling the fx-2, whoa thats quite a leap for intel, i never knew AMD's most valuable and fastest processor is equal to the lower end e6600, and kougar, is by any chance the e6700 the EE of intel? or will they make another EE?

Nice summation. ;)

I'm not entirely sure what you were asking... Intel will release the E6600 and E6700 for the mid to high-end range, with the X6800 clocking in at 2.93ghz. The "X" in X6800 signifies the new "EE" chip. Later this fall Intel will release a X6900 at 3.2ghz... and if not in January then in the first half of 2007 will release Kentsfield as the third "EE" processor.

[vfrex]

What? Aren't games more often bottlenecked by video cards than processors? I fail to see how spending more on a processor that will go under-utilized is a better idea than spending more on a video card.

[mousiness]

nobody understands what im saying its just the age difference i bet

[werty316]

What? Aren't games more often bottlenecked by video cards than processors? I fail to see how spending more on a processor that will go under-utilized is a better idea than spending more on a video card.

Thats what I was thinking too and thats why I OC to try to less the bottleneck. Everything you said is what I got by. Todays games reply much more on video card power.

[vfrex]

nobody understands what im saying its just the age difference i bet

I see what you are saying, but let me extend my counter. Gamers don't need the most cutting edge processors. Their money is better spent on more moderately priced processors that will deliver close to the same level of performance. That trend isn't going to go away. Yeah, the higher end conroe will OC well and give you a few more FPS...but there are many ways to sneak in a few more. More, the ways vary based on how any particular game is designed. It seems silly to get a high end processor when one a few steps down will perform just as well.

[mousiness]

if you are a true gamer you will buy a processor like the 3000 or 3500 and OC it to beat out an fx-53, thats the way gamers do, but if you are not a true game or a gamer with money you will want the name, gamers spend more money on cards

[werty316]

if you are a true gamer you will buy a processor like the 3000 or 3500 and OC it to beat out an fx-53, thats the way gamers do, but if you are not a true game or a gamer with money you will want the name, gamers spend more money on cards

Gamers don't have to be cheap with hardware if they are rich. Maybe you should reword that like some gamers prefer to spend less and buy a slower CPU and OC it to the speeds of a more expensive CPU for spending the extra cash on a faster video card which games reply on more.

[mousiness]

isnt that basically what i just said except for the rich gamers part? IMHo if i was rich id spend money on the best cpu out there instead of OC'ing, but as you said the true gamer will OC, or enthusiast

[werty316]

Not all hardcore gamers but a slower chip. The last part of your post (http://www.bjorn3d.com/forum/showpost.php?p=36219&postcount=16) didn't really make sense.

[mousiness]

yes it did, what i said was, true gamers (such as you OC'ers) that are rich, might not buy the best CPU out there, if tehy are true to their name they will OC the most OC'able CPU out there to beat out the more expensive CPUs they can afford, and instead spend their money on video cards

[werty316]

If a gamer is rich they will buy the most expensive parts.

[GIBSON]

i personally do think the smaller cache of the lower-end segment will make quite the difference, after all doesn't conroe get a lot of it's speed because of it's large cache?

[mousiness]

yes it does if it wasnt for conroes cache it wouldnt be as fast as it is

[GIBSON]

i'm actually getting very curious what those lower-end conroes will do with the smaller cache

[mousiness]

probably not as fast because conroe is built off of the shared 4mb cache, thats why reviewrs are buying the slower conroes first (lower end ones i mean) to test them out against am2 and current s939 AMD's, theres no doubt in my mind theyll still be stupid fast though

[werty316]

That and it a unified cache and not 2x2MB to make 4MB.

[mousiness]

yep i forgot to add that part, thats why conroe is so fast both cores can access one cache at the same time

[werty316]

yep i forgot to add that part, thats why conroe is so fast both cores can access one cache at the same time

Why mention it when thats pretty much what I said? E6700 isn't a lower end Conroe.

[mousiness]

sorry bout that ive been on my pc all day, im known to be a complete idiot and say exactly what the last person said in real life i dotn comprehend things the same as you guys

[GIBSON]

as far as i know there haven't been any reviews yet about the lower-end ones so we can't know yet right? Maybe they might be performing on par with amd then?

[mousiness]

never know until they come out

[Kougar]

I see what you are saying, but let me extend my counter. Gamers don't need the most cutting edge processors. Their money is better spent on more moderately priced processors that will deliver close to the same level of performance. That trend isn't going to go away. Yeah, the higher end conroe will OC well and give you a few more FPS...but there are many ways to sneak in a few more. More, the ways vary based on how any particular game is designed. It seems silly to get a high end processor when one a few steps down will perform just as well.

The big deal about Conroe is that the 4mb cache parts are showing not just a few handfuls of FPS leads, but leads as much as 70+FPS over the FX62 using the exact same graphics card or Crossfire/SLI setups. 20+FPS might as well be a higher level GPU graphics card, and 50FPS will at the minimum be the same as having purchased not the next higher GPU, but two GPUs above your current model. This is especially true if you game at 1024x768 or 1280x1024, as the lead diminishes gradually as you increase the GPU dependence with the bigger resolutions. The more CPU limited your current game is, the bigger the gains with Conroe will be.

Basically I'm trying to say that Conroe seems to give most users far more performance than say, from switching to the next highest level of graphics card. With Conroe it will be tricky to match your graphics card to the right processor to get the best possible combination of both to get the best results. That's why I think the rule to spend more on your GPU and skimp of the CPU is no longer applicable... ;)

Some random preliminary numbers:http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2771&p=5 (http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2771&p=5)

All the numbers you could possible need and them some 8-) HKEPC.com (http://translate.google.com/translate?u=http%3A%2F%2Fhkepc.com%2Fhwdb%2Fx6800v sfx62-12.htm&langpair=zh-CN%7Cen&hl=en&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools) Infact I linked straight to the Crossfire review page... they have to date done the most thorough analysis run I've seen on Intel's upcoming flagship compared to AMD's FX62 flagship. I'd suggest reading the whole thing, or at least just the graphs and tables ;)

[Kougar]

as far as i know there haven't been any reviews yet about the lower-end ones so we can't know yet right? Maybe they might be performing on par with amd then?

There are none that I have seen as of yet, because there isn't a way to turn off half the L2 cache. And you are right, dropping the cache from 4mb down to 2mb will have a noticeable impact, but from what I have seen I would extrapolate it won't turn the cache strapped parts down to sub-AMD levels.

As Werty pointed out Intel was the first to use a unified cache design, a "true" dual-core. It is not 2x1mb of cache like AMD's FX62, but just one large 2mb pool, that either core can use the entirety of. While this would leave the second core hanging, it'd still produce a bigger performance advantage over two cores foreced to use half the cache all the time.

[mousiness]

exactly and the two cores are working together to provide the best performance

[werty316]

Isn't it technically two cores but on one die?

[vfrex]

The more CPU limited your current game is, the bigger the gains with Conroe will be.

So if you change the settings of a game as to decrease the load on the GPU thus removing the bottleneck, FPS will depend more on the CPU...I mean, that's like saying the sky is blue because it has a blueish color. It is obvious and mostly irrelavent to the topic. The game is still GPU limited, you are just avoiding it by disabling any possible eye candy you can.

How many people actually play games at the minimum resolution with all of the graphical options disabled? We're probably talking about a minority of gamers here playing FPS where those extra 30-40 FPS might make a difference. Otherwise, we're still talking about having a graphics bottleneck. And so long as we're talking about game performance bottlenecking in the graphics card long before the CPU slows things down, Conroe's performance is irrelavent to the majority of gamers.

[werty316]

Here are some CPU scaling benches.

F.E.A.R. CPU Shootout: AMD vs. Intel (http://www.firingsquad.com/hardware/fear_cpu_performance/)

A 7800GTX 256MB was used for these tests.
http://www.firingsquad.com/hardware/fear_cpu_performance/images/mainhigh.gif http://www.firingsquad.com/hardware/fear_cpu_performance/images/dualhigh.gif

CPU Scaling ATI’s and NVIDIA’s Best (http://www.hardocp.com/article.html?art=NjMy)

[vfrex]

3dMark06
0.919021213
0.923160601
0.92066338
0.932631867

Doom3
0.883287671
0.898741419
0.961128527
0.962213225

Half-Life2
0.9231088
0.906950328
0.910609538
0.772043126

BattleField2
0.856324754
0.891815833
0.895366741
0.920306112

Splinter Cell CT
0.859735164
0.833591731
0.822247065
0.821835231


That is the ratio of the FX-62 peformance to the Conroe based on HKPEC's benchmarks. The top number in a game will be 1024x768 and the bottom is 1600x1200 4AA. I left out the numbers for 3dmark03 and 05 because they are pointless.

You can see that as the resolution increases, the numbers in 3dmark06, Doom3, and BF2 get closer together. That is, the FX-62 gains on the conroe in relative performance. The explanation in those instances is a system bottleneck not related to the CPU...most likely a graphics bottleneck, DESPITE THE FACT THAT THEY ARE RUNNING HIGH END GRAPHICS CARDS IN CROSSFIRE MODE.

That leaves HL2, whose results are a mixed bag. Does it make sense to anybody that its performance drops off so heavily when 4AA is introduced in this game, but not in any of the others? I wonder if there was an error in testing? Splinter Cell is the only game here where we see the conroe actually gaining as resolutions increase, and those gains are minimal.

[GIBSON]

The big deal about Conroe is that the 4mb cache parts are showing not just a few handfuls of FPS leads, but leads as much as 70+FPS over the FX62 using the exact same graphics card or Crossfire/SLI setups. 20+FPS might as well be a higher level GPU graphics card, and 50FPS will at the minimum be the same as having purchased not the next higher GPU, but two GPUs above your current model. This is especially true if you game at 1024x768 or 1280x1024, as the lead diminishes gradually as you increase the GPU dependence with the bigger resolutions. The more CPU limited your current game is, the bigger the gains with Conroe will be.


this doesn't really make sense now does it, if almost all today's games are GPU bottlenecked, why would a conroe make that much of a difference then?

[Kougar]

Here are some CPU scaling benches.

CPU Scaling ATI’s and NVIDIA’s Best (http://www.hardocp.com/article.html?art=NjMy)

Werty, I love you... But that's not quite enough for me to let you get away with another 7900GT :mrgreen:

Even though this is a dated article it is a very, very good one. vFrex, this is basically my case in a nutshell for why you shouldn't skimp on the CPU when running with a good GPU. Yes, there are games that are solely GPU dependant, but the same can be said about the reverse, some games love faster CPUs more.

That is the ratio of the FX-62 peformance to the Conroe based on HKPEC's benchmarks. The top number in a game will be 1024x768 and the bottom is 1600x1200 4AA. I left out the numbers for 3dmark03 and 05 because they are pointless.

You can see that as the resolution increases, the numbers in 3dmark06, Doom3, and BF2 get closer together. That is, the FX-62 gains on the conroe in relative performance. The explanation in those instances is a system bottleneck not related to the CPU...most likely a graphics bottleneck, DESPITE THE FACT THAT THEY ARE RUNNING HIGH END GRAPHICS CARDS IN CROSSFIRE MODE.

I would agree with this. But I don't think these numbers demonstrate anything about CPU scaling, you would need different Intel and AMD CPUs run to compare them to the above X6800 and FX62 numbers. The link I quoted from Werty is pretty much the best evidence I could have shown for my arguement. In particular, the Apples to Apples comarisons with the 6 different CPUs, spanned across three different GPUs.

Be careful to note that when they change from the AMD 3500+ (socket754) to the AMD 3000+ (socket939) that the ghz drops back from 2.45ghz to 2.21ghz, respectively. This muddles up the numbers... but if you page over to the three Intel CPUs, where they scale the clock speeds from 2.4ghz, 3ghz, and 3.4ghz, you can see where the game goes from being CPU limited to becoming GPU limited on the FarCry benchmarks. While BF Vietnam mostly has little to no improvements, Flight Sim 2004 again demonstrates my point.

To quote their conclusion:
There is no question that if you want the best visual gaming experience from your high end video card you need the fastest CPU you can get your hands on to experience the high level of gameplay these video cards are capable of producing. While it has become evident over the past couple of years that video chipset technology is extremely important in producing a great gaming environment, we can't forget about the CPUs at the heart of our machines. Quite simply put, upgrading to some of today's CPU technology can directly benefit your gaming experience.

Yes this isn't true with all games (And in particular some of the latest offerings, I'll admit), as Werty's link to those FEAR results show... (Those were surprising). Some CPU scaling game benches should demonstrate this... here are several intentionally CPU limited game benchmarks. Notice how badly Conroe wipes the floor with the FX62... http://www.matbe.com/articles/lire/306/merom-et-conroe-test-des-core-2-duo/page21.php (http://www.matbe.com/articles/lire/306/merom-et-conroe-test-des-core-2-duo/page21.php) Now realize that these numbers were based solely on the strangth of the CPU using the same GPU. The CPU matters, and Conroe can do for todays games what AMD did for the last couple years of games, bring a large FPS jump from the previous ruling processor line.

Here is a THG forum post that should interest anyone interested in figuring out some rough numbers for when AMD's hardware can again match Intel's... although it seems more of a best case scenario to me since these are only take from FEAR numbers. I've seen worse... http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&p=1096547#1096547 (http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&p=1096547#1096547)

Again, thanks Werty! :)

[Kougar]

Since this deserves it's own post:

Here's a very large range of benchmarks across a compltely massive database of processors. In particular, the CPU limited game benchmarks looked interesting! (so far, I've not had a chance to read it all yet!) I did not notice anything to point out in the setup section, except to clarify that the the E6300 appears to be a legitimate 2mb cache processor! So now everyone can see what Intel's sub-$200 part can do... And do, it certainly does...

The review is in French, translated via google: http://translate.google.com/translate?u=http%3A%2F%2Fwww.matbe.com%2Farticles% 2Flire%2F306%2Fmerom-et-conroe-test-des-core-2-duo%2Fpage1.php&langpair=fr%7Cen&hl=en&safe=off&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools (http://translate.google.com/translate?u=http%3A%2F%2Fwww.matbe.com%2Farticles% 2Flire%2F306%2Fmerom-et-conroe-test-des-core-2-duo%2Fpage1.php&langpair=fr%7Cen&hl=en&safe=off&ie=UTF-8&oe=UTF-8&prev=%2Flanguage_tools)

[vfrex]

Kougar, you are proving what I said. As long as they take away the graphics bottleneck by decreasing graphics settings, Conroe wins hands down. The gap widens if you consider older games that have less eye candy that you can't disable. But that doesn't matter. Are you suggesting that people buying a high end Conroe will be playing old games at low resolutions? If so, the difference between 150 might not even be perceptible by the human eye. The debate on that rages on..

Newer games will stress both the processor and the video card more than today's games, right? Increasingly the video card. Further, there aren't going to be any real changes to video cards until DX10 hits. THEN, maybe the bottleneck from the graphics cards will be lifted and high end Conroes (and AMD's for that matter) will shine in gaming scenarios that actually matter. However, at that point, today's top end Conroes won't be top anymore. Their prices probably will have dropped.

[Kougar]

Kougar, you are proving what I said. As long as they take away the graphics bottleneck by decreasing graphics settings, Conroe wins hands down.

It still wins at GPU limited settings as well.

The gap widens if you consider older games that have less eye candy that you can't disable. But that doesn't matter.

Again I will say read the article (http://www.hardocp.com/article.html?art=NjMy), as you obviously didn't. They took some of the best processors and GPUs from 2004, and compared the CPU scaling of both AMD and Intel CPUs, across the same three graphics cards. The settings were at 1600x1200 4xAA+16xAF. Considering this is 2004s best hardware I'm sure you would have agreed these were GPU dependent settings for GPUs back then? Then how about explaining the linear CPU scaling I see with FarCry and Flight Simulator 2004, while BF:V appears to be GPU dependant at those settings.

Are you suggesting that people buying a high end Conroe will be playing old games at low resolutions? If so, the difference between 150 might not even be perceptible by the human eye. The debate on that rages on...

For the second time, I am not. I am actually suggesting the average person will not have SLI/Crossfire, or the topmost end graphics hardware. I would agree the difference between 150 FPS and 220FPS wouldn't be noticeable, but since you thought I was referring to SLI/Crossfire rigs I can see why you thought I was saying this. Again however, just because a game is played at 1600by1200 with 4xAA+16xAF does not automatically make it a GPU dependent setting either.

Newer games will stress both the processor and the video card more than today's games, right? Increasingly the video card.

I agree that newer games would stress either one or both, yes. It's possible this could be happening regarding the video card... but if GPUs have spare cycles going unused so that nVidia and ATI plan to devote parts of the GPU to physics rendering, then wouldn't this imply otherwise? There are still many games where the GPU loses cycles while the CPU is working away, which is the core idea behind running physics processing on the GPU.

Further, there aren't going to be any real changes to video cards until DX10 hits. THEN, maybe the bottleneck from the graphics cards will be lifted and high end Conroes (and AMD's for that matter) will shine in gaming scenarios that actually matter. However, at that point, today's top end Conroes won't be top anymore. Their prices probably will have dropped.

I would agree completely with this. ;) But if this came true, then having the most powerful CPU uArch for gaming would be critical, which means we're back to Conroe. Not just the X6800, but the E6600 and E6300 even. Regardless of the resolutions run, GPU or CPU dependancy, I have not once seen AMD put forth a better FPS number than a similarly priced Conroe part, at any price point.

[vfrex]

I'll respond to this more when I get to work, but for now:

hen how about explaining the linear CPU scaling I see with FarCry and Flight Simulator 2004, while BF:V appears to be GPU dependant at those settings.


Where do you see linear scaling in FarCry? There NO difference between the S754 3000 and the 939 3500. The only difference is going from the XP 2500+ to the S754 3000. Look at Flight Sim. Same deal.

[GIBSON]

I'll respond to this more when I get to work, but for now:



Where do you see linear scaling in FarCry? There NO difference between the S754 3000 and the 939 3500. The only difference is going from the XP 2500+ to the S754 3000. Look at Flight Sim. Same deal.
indeed, i just put the two graphs next to each other (they are both at the same settings) and there's nearly no difference, only the intel system seems to favor the fastest of the three

[Kougar]

I'll respond to this more when I get to work, but for now:



Where do you see linear scaling in FarCry? There NO difference between the S754 3000 and the 939 3500. The only difference is going from the XP 2500+ to the S754 3000. Look at Flight Sim. Same deal.

As I also pointed in my my 2nd post above this one... This is because of the clock scaling differences. ;)

From my earlier post:

Be careful to note that when they change from the AMD 3500+ (socket754) to the AMD 3000+ (socket939) that the ghz drops back from 2.45ghz to 2.21ghz, respectively. This muddles up the numbers... but if you page over to the three Intel CPUs, where they scale the clock speeds from 2.4ghz, 3ghz, and 3.4ghz, you can see where the game goes from being CPU limited to becoming GPU limited on the FarCry benchmarks. While BF Vietnam mostly has little to no improvements, Flight Sim 2004 again demonstrates my point.

[vfrex]

You are confusing me.

AMD Platforms:

MSI K8N Neo 2 Platinum, Socket 939 Athlon64 3500+ @ 2.21GHz, 2 X 512MB Corsair XMS PC3200LL TwinX Dual Channel DDR400, Maxtor 40GB ATA/133, Windows XP Professional SP1 with DirectX 9.0b.

ABIT KV8 Pro, Socket 754 Athlon64 3000+ @ 2.45GHz, 2 X 512MB Corsair XMS PC3200LL TwinX Dual Channel DDR400, Maxtor 40GB ATA/133, Windows XP Professional SP1 with DirectX 9.0b.

ABIT NF7-S, AthlonXP “Barton” 2500+ @ 1.83GHz, 2 X 512MB Corsair XMS PC3200LL TwinX Dual Channel DDR400, Maxtor 40GB ATA/133, Windows XP Professional SP1 with DirectX 9.0b.

You got it backwards. The 3000 is 754. The 3500 is 939. The 3000 has a higher clock on S754, maybe because it doesn't support dual channel memory?

[Kougar]

You are confusing me.



You got it backwards. The 3000 is 754. The 3500 is 939. The 3000 has a higher clock on S754, maybe because it doesn't support dual channel memory?

Ack, you're right. I goofed the names in my posts... but it doesn't change my point! :mrgreen: The socket939 part is suffering from a 240mhz loss in clock speed, yet shows numbers on par with the lower performance 754 part.

I am sure you are correct in that this is at least in some part due to the new dual-channel memory capability that socket939 brought to the table... I wish they'd included a higher clocked socket939 part though, if just to not backtrack and muddle the numbers...

[mousiness]

i bet its the pins too, isnt an s754 3000+ a repinned s939 3000+?

[GIBSON]

i bet its the pins too, isnt an s754 3000+ a repinned s939 3000+?
LOL i didn't know that is actually possible? (you do mean taking existing 939s and making 754s of it right?)

[Kougar]

i bet its the pins too, isnt an s754 3000+ a repinned s939 3000+?

Pins don't increase the performance difference, as I'm sure the 754socket wasn't shortchanged on pins... ;) The change was due to the core modifications for dual-channel memory, which required the majority of the new pins. I don't recall the rest of it, although there was a performance improvement for the CPU itself if I remember correctly, that wasn't due to the increased memory bandwidth? I could be wrong.



Anyway, more on track with the thread subject, for anyone still interested!

Here is a new review on the FX62, 5000+, FX60, Pentium 955 EE, and Conroe E6400 (2.4ghz). http://www.tbreak.com/reviews/article.php?id=458 (http://www.tbreak.com/reviews/article.php?id=458)

These numbers seem a little higher than usual in the E6400's favor, but still within the overall norms I think. A E6400 at ~$316 is par to marginally ahead of a $1,200 FX62, which will NOT be getting any price cuts...

[werty316]

Yep the E6400 edges out a FX-62 at 1/4 of the cost but don't forget the Core2Duo is next gen compare to the FX-62 which is current gen but it is atill amazing how fast the Conroe chip is. Intel is gonna be a on good roll once it is out and the Kentsfield is out.

[mousiness]

you never know until AMD replies to intel with an equal competitor, but intel ha the edge now and i think will lead for the conroe generation

[werty316]

Thats what the K8L is for ;),... I think

[mousiness]

yes it is

[mousiness]

i just realised for $50 more of the price of an x2 3800+ you could by an intel 950 3.4ghz... gees thats a really expensive processor, im surprised anyone can afford that thing

[GIBSON]

i just realised for $50 more of the price of an x2 3800+ you could by an intel 950 3.4ghz... gees thats a really expensive processor, im surprised anyone can afford that thing
Which one of the two are you saying to be expensive now :confused:
Anyhow, I still don't understand why amd made am2 without changing directly to a new cpu architecture while they were at it. They had the lead and they just gave it away?

[werty316]

Which one of the two are you saying to be expensive now :confused:
Anyhow, I still don't understand why amd made am2 without changing directly to a new cpu architecture while they were at it. They had the lead and they just gave it away?

One step at a time I guess since its their first time into the DDR2 market. I think Intel had a hard time jumping into DDR2 but I can't remember as it was way too long ago.