Tom posts new AM2 Conroe tests....... Conroe looks good

[BaronMatrix]

I just saw the newest comparison and had to post. It looks very interesting an seems like Intel is introing new SSE instructions because nonSSE code is not as friendly to Conroe.

I noticed how the gaming results were very close in most cases. I am getting sick of these partial reviews though. I'd rather wait until the damn thing comes out.

I also noticed that Intel was using DDR2 667 which may have slowed them down a little, but from what I've seen Core2 doesn't do a lot more with the extra bandwidth.

I still think though that all benchmarks should be run with things in the background. I never run games with everything else closed. Non of this clean room stuff. I mean it is relevant to some extent but I want to see how my chip and video do with 700MB laoded before running the game. I guess I could run all the benches myself though.

Anyway, again Conroe is looking good. It will finally give K8 some competition. But then if K8L is half what they say as a dual core, Conroe will catch MASSIVE BEATDOWN next year.

 

[MU_Engineer]

A regular Core 2 chip by itself can't utilize more than DDR2-533 in dual channel (1066 effective) as its FSB is 266 quad-pumped (1066 effective.) The chipset can utilize a little more RAM bandwidth for other things, particularly an integrated graphics core. So I'd say that in a Core 2 chip, you'll see DDR2-533 be slightly beaten by DDR2-667 and DDR2-800 being pretty much equal with DDR2-667, with all the latencies being equal.

AMD's AM2 chips are a much different story as their integrated memory controllers expect to see DDR2-800 and don't function that hot with RAM slower than that (as Tom's AM2 story showed, especially with the 2.2 GHz X2s and Athlon 64s.)

 

[Ycon]

Well, I think that the performance will improve since many memory makers are going for fast DDR2-800 which will reduce timings of the modules and that is a pretty good thing.

 

[mjp1618]

But then if K8L is half what they say as a dual core, Conroe will catch MASSIVE BEATDOWN next year.

Is next year good enough?

Intel will have 45nm and monolithic quad core next year. And a year from next, they will intro Nehalem...

All the AMD fanboys seem to be expecting miracles from AMD's transition from 90nm to 65nm. Going by that logic, the Core architecture could benefit a lot from going 45nm...

And, next year might have only server class K8Ls. Nothing is clear at this stage (the situation is similar to what it was last year with the NGMA hype); but many reports seem to suggest that K8L desktop chips will intro only in 2008. And that is wayyy late.

By the way, I agree that K8L would be very good. After all, it is a new marchitecture. But then again, wasn't it AMD's own Henri Richard himself who said that K8L will be only "evolutionary", rather than revolutionary?

 

[kickbutt]

Yeep it still looks good but its starting to look much more even now
The 2.66GHz Conroe beats the 2.8GHz FX62 but not with so much (in games and even the fx62 at 3GHz).
But its not with that much ...

So AMD will stay in the race untill there die shrink when they can pump up the Hertz some more.

And as long as they can get chips out there thats 10-15% faster in MHz then they can compete with conroe.

Also this was agains a 2.66GHz 4Mb cache conroe.
There will be slower conroes with only 2Mb and they will prob have a harder time agains amd as intel needs it cache (FSB anyone?) and then the diffrence might not even be that much (MHz/MHz).
Ohh but there will be a 3GHz EE version to...
Yeeh but hardly with more cache then 4Mb and then its the same diffrence that with the 2.66GHz version just higher clocked.

So AMD isnt dead yet (and will live to be the champs again then K8L comes out).

Hehe fanbooiie I am (well not really but Im pretty tired of conroe allready as the intel fanboys go on and on...)

 

[BaronMatrix]

But then if K8L is half what they say as a dual core, Conroe will catch MASSIVE BEATDOWN next year.

Is next year good enough?

Intel will have 45nm and monolithic quad core next year. And a year from next, they will intro Nehalem...

All the AMD fanboys seem to be expecting miracles from AMD's transition from 90nm to 65nm. Going by that logic, the Core architecture could benefit a lot from going 45nm...

And, next year might have only server class K8Ls. Nothing is clear at this stage (the situation is similar to what it was last year with the NGMA hype); but many reports seem to suggest that K8L desktop chips will intro only in 2008. And that is wayyy late.

By the way, I agree that K8L would be very good. After all, it is a new marchitecture. But then again, wasn't it AMD's own Henri Richard himself who said that K8L will be only "evolutionary", rather than revolutionary?

Yeah, AMD has been really tight lipped, but at Analyst Day, they did say that K8L WOULD BE DUAL CORE also. What it looks like they are doing is Quad for servers dual for desktops and mobile. It's just good to see the inventor of x86 not look like Cyrix. But then they di run several manufs out of business so I guess AMD is payback.

When you think about evolutionary, it means that their simulations say K8 is not dead. Tweaking the decoders to take larger streams, tweaking the pipeline to do more in a cycle (superscalar). WIth the improvements that can be made to K8, they don't need a K9 or K10.

ANd as far as Socket F being pushed back I woul dsay that they are making most of them for Dell, Sun and HP, so by delaying it they can get more chips to them for system testing. In my opinion.

 

[306maxi]

Yeep it still looks good but its starting to look much more even now
The 2.66GHz Conroe beats the 2.8GHz FX62 but not with so much (in games and even the fx62 at 3GHz).
But its not with that much ...

So AMD will stay in the race untill there die shrink when they can pump up the Hertz some more.

And as long as they can get chips out there thats 10-15% faster in MHz then they can compete with conroe.

Also this was agains a 2.66GHz 4Mb cache conroe.
There will be slower conroes with only 2Mb and they will prob have a harder time agains amd as intel needs it cache (FSB anyone?) and then the diffrence might not even be that much (MHz/MHz).
Ohh but there will be a 3GHz EE version to...
Yeeh but hardly with more cache then 4Mb and then its the same diffrence that with the 2.66GHz version just higher clocked.

So AMD isnt dead yet (and will live to be the champs again then K8L comes out).

Hehe fanbooiie I am (well not really but Im pretty tired of conroe allready as the intel fanboys go on and on...)

What you fail to notice is that this processor that is near enough to the FX-62 in gaming benchmarks isn't the top of the range processor. So if you want performance that almost matches AMD's top range processor you only need to pay roughly 50% of the price of an FX-62. if you want to buy a more expensive processor you can get the Extreme Edition. Also don't forget that as the review says.... this wasn't an ideal test setup for Conroe and you will see improvements in Conroe performance as time goes on and different steppings are released. These will be small of course but still important.

We will need to wait till the lower spec processors are compared to lower specced AMD processors before we can conclude that the Conroe's with only 2mb of cache aren't going to perform.

 

[Makaveli]

While i'm glad intel finally has something worth buying! what interests me the most is how much performance do these conroe chips actually get from that 4mb Cache. I know they made lot of improvement to the core, but most of the gaming benchmarks I believe its all cache!

It will be very interested in seeing how the 2mb cache version do.

It about time they started to push AMD after sitting on K8 for the last 3 years with no answer from intel.

As for K8L we might just see it in 2007, but I believe that will be the server chips. The desktop K8L will come in dual core first then Quad.

As it is right now i'm still gonna hold on to my opteron 146 and go to a Dual core 170 after a few more prices drops. I will be more interested in doing a full system upgrade when the 170 is to slow for me.

seeing conroe win in the games is impressive but no one plays sh*t in low detail most everyone plays stuff in 1280x1024 or higher or atleast I do. so my upgrades will be videocards until then. CPU bottlenecks aren't as great as they use to be.

 

[MU_Engineer]

I'd be dumb to take you up on your bet as you are likely right. The FSB and northbridge on the Intel setup limit what different RAM can do to the CPU, both for good (high-latency, slow RAM doesn't affect the CPU that much) and for bad (low-latency, fast RAM doesn't help the CPU out much.) We've seen it with the P4 lineage, why should it be different as the chipsets talk to the processor the same way?

 

[BaronMatrix]

Yeep it still looks good but its starting to look much more even now
The 2.66GHz Conroe beats the 2.8GHz FX62 but not with so much (in games and even the fx62 at 3GHz).
But its not with that much ...

So AMD will stay in the race untill there die shrink when they can pump up the Hertz some more.

And as long as they can get chips out there thats 10-15% faster in MHz then they can compete with conroe.

Also this was agains a 2.66GHz 4Mb cache conroe.
There will be slower conroes with only 2Mb and they will prob have a harder time agains amd as intel needs it cache (FSB anyone?) and then the diffrence might not even be that much (MHz/MHz).
Ohh but there will be a 3GHz EE version to...
Yeeh but hardly with more cache then 4Mb and then its the same diffrence that with the 2.66GHz version just higher clocked.

So AMD isnt dead yet (and will live to be the champs again then K8L comes out).

Hehe fanbooiie I am (well not really but Im pretty tired of conroe allready as the intel fanboys go on and on...)

What you fail to notice is that this processor that is near enough to the FX-62 in gaming benchmarks isn't the top of the range processor. So if you want performance that almost matches AMD's top range processor you only need to pay roughly 50% of the price of an FX-62. if you want to buy a more expensive processor you can get the Extreme Edition. Also don't forget that as the review says.... this wasn't an ideal test setup for Conroe and you will see improvements in Conroe performance as time goes on and different steppings are released. These will be small of course but still important.

We will need to wait till the lower spec processors are compared to lower specced AMD processors before we can conclude that the Conroe's with only 2mb of cache aren't going to perform.

True but the 4800+ and 5000+ are not THAT much slower than the FX62 at 300-400 less. I'm sure the Core 2 with 2MB will be competitive, at least I hope Intel didn't pull a cache trick even though they tend to do that a lot.

AM2 is looking good. AMD will jsut be in a position of needing to be 300MHz faster at a similar price. They definitely have a low power chip - except for the FX62. Both companies look good right now. Let's hope it continues so prices will go down.

 

[Action_Man]

I concur.

 

[306maxi]

Yeep it still looks good but its starting to look much more even now
The 2.66GHz Conroe beats the 2.8GHz FX62 but not with so much (in games and even the fx62 at 3GHz).
But its not with that much ...

So AMD will stay in the race untill there die shrink when they can pump up the Hertz some more.

And as long as they can get chips out there thats 10-15% faster in MHz then they can compete with conroe.

Also this was agains a 2.66GHz 4Mb cache conroe.
There will be slower conroes with only 2Mb and they will prob have a harder time agains amd as intel needs it cache (FSB anyone?) and then the diffrence might not even be that much (MHz/MHz).
Ohh but there will be a 3GHz EE version to...
Yeeh but hardly with more cache then 4Mb and then its the same diffrence that with the 2.66GHz version just higher clocked.

So AMD isnt dead yet (and will live to be the champs again then K8L comes out).

Hehe fanbooiie I am (well not really but Im pretty tired of conroe allready as the intel fanboys go on and on...)

What you fail to notice is that this processor that is near enough to the FX-62 in gaming benchmarks isn't the top of the range processor. So if you want performance that almost matches AMD's top range processor you only need to pay roughly 50% of the price of an FX-62. if you want to buy a more expensive processor you can get the Extreme Edition. Also don't forget that as the review says.... this wasn't an ideal test setup for Conroe and you will see improvements in Conroe performance as time goes on and different steppings are released. These will be small of course but still important.

We will need to wait till the lower spec processors are compared to lower specced AMD processors before we can conclude that the Conroe's with only 2mb of cache aren't going to perform.

True but the 4800+ and 5000+ are not THAT much slower than the FX62 at 300-400 less. I'm sure the Core 2 with 2MB will be competitive, at least I hope Intel didn't pull a cache trick even though they tend to do that a lot.

AM2 is looking good. AMD will jsut be in a position of needing to be 300MHz faster at a similar price. They definitely have a low power chip - except for the FX62. Both companies look good right now. Let's hope it continues so prices will go down.

I concur also. I care not for the sticker that I put on the front of my PC. Only how fast it makes my PC does stuff and how much it costs me.

 

[Action_Man]

Wuzy's post.

I'm betting Conroe doesn't need DDR2-800 while timing lower than 4-4-4-x 1T won't show much increase either. Anyone wanna place money on it?
Unlike S-AM2 which has been proven (unlike some idiotic comments here) it needs lowest possible timing (preferbly 3-2-2-x) and not DDR2-800.

 

[Action_Man]

Bah to expensive memory! Unless you're overclocking.

 

[BaronMatrix]

tbh i can't say i care. i would have thought intel with all there marketing people would stop mucking around with this pre setup tests that no one ( hopefully) believes. i mean what are they thinking. i would have thought they would have been all too happy to allow some independant tests by now. it is nearly launch time and they are still acting suspiciously. what gives.

It definitely is fishy. I believe the marks but I never liked benchmarks run on clean machines. Intel should be letting systems out. I mean eventually, Tom will get all the speeds and do a comparison.

 

[Ramsjunior]

Dont some of those tests (the games) show that the systems were bottlenecked by the gfx cards. On the higher resolutions the FPS were about the same, but at lower resolutions, the Conroe was ahead.

 

[conroe]

A $180 E6300 competes with the FX-60? I can't see AMD matching that price to performance ratio. They go dual socket and double CPU price and power consumption? Silly. Power consumption, price, and performance, Conroe is the leader.

DDR2-800 will show a minimal improvement over DDR2-667 with Conroe. Best of all it will run 1:1 at 3.5-4GHz.

 

[Makaveli]

We have a winner. Yes most games are bottlenecked by video performance. Cpu's bottles necks aren't as great as they use to be these days.

And unless you plan on playing at 800x600 with your new conroe, it isn't gonna be much faster at the resolutions and settings people actually play at.

 

[Ramsjunior]

"We have a winner."

oooo what do i win

I thk I will be sticking with my system for a little bit longer, especially since the only game I am playing at the moment is wow ( didnt like oblivion, so wasted £25 ).

 

[MrsD]

Yes Conroe looks really good. Although they admit the tests were run on preconfigured systems from Intel.
Looks like we have a winner but not by the margins that we were led to believe.
This will really bump up the cpu war and only means good things for us. :lol:

 

[Heyyou27]

It all looks very nice, but when playing at a resolution higher than 640x480 I see no reason to upgrade to a Conroe this year. :lol:

 

[Ramsjunior]

Regarding the the comment "we have a winner", that was Makaveil nicely pointing out the obviousness to my statement of bottlenecking

 

[gr8mikey]

I was somewhat disappointed in these benchmarks. Mainly because it doesnt really give us enough information about the test systems to be able to draw definite conclusions. It was interesting to see the FX-62 win the Far Cry high quality tests.

Just for the sake of fair comparison, though you should only compare the non-overclocked benches. Otherwise, you would need to consider what the conroe scores would be if similarly overclocked.

I also realized thiere is a bit of disparity between these results and the early IDF benchmarks. This is likely to be due to this test being more GPU limited as the IDF scores were obtained using an ATI crossfire setup as opposed to the single card that was used for this test. The low res scores in this review pretty much confirm this.

 

[Ramsjunior]

I wouldnt worry too much about the benchmarks just yet, like you have just pointed out for me, it was just done on a single gfx card for a start, so lets hope for some duel gfx tests when conroe is released to remove any gfx bottlenecks.

 

[GloriosoSLB]

I think that if Intel is limited by its FSB speed then AMD is also,
if we look at Anandtech review, the AMD FSB OC from 200 to 250, we also see the same trend.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2741&p=6

I was expecting AMD going to use the 266mhz bus with Am2 (even if it was 266x3, instead of 200x5).

Maybe they lack chipset support like Intel is also lacking (using P4 modified ones).

The results that Anandtech took from the 250Mhz FSB, AMD results are also much better, in games however doesn’t turn out so good.

22% better latency, 39% better bandwidth, 54% better memory write, 20% better Pi result,… even Far Cry goes from 62 to 75 fps.

I think that we have been looking at benchmarks with low detail settings where Conroe fly because of the huge 4MB (when using one core). That fit very well because of the lower number of instructions that has to compute, and more stay in cache improving the results.

I think with more details settings both will perform on par because of Video cards processing limitation and because of more accesses to the system RAM.