Conroe: The truth?

http://sharikou.blogspot.com/2006/04...ng-busted.html

Quote:

---

However, once you go over the 4MB limit, Conroe is slower than Athlon 64 at the same clock. Both the Cryptography and STREM tests use a lot more than 4MB, larger than Conroe's 4MB cache, and Conroe immediately falls below Athlon 64 on the performance curve.

---

Just as I predicted.. the measly 1066FSB becomes a huge bottleneck.
His point about cache thrashing makes sense as well.

Gamers probably won't notice much of a performance hit since most games are still single threaded, and unless you're running SLI or CF, the FSB doesn't become too much of a bottleneck. BUT... once multithreaded games come out + 2x 16x PCIe mobos and R600 CF or G80 SLI, the FSB will become a MAJOR bottleneck. Imagine the Quad Core Intel CPU coming out next year.... yikes.

When is Intel going to switch to an on die memory controller?? :rolleyes:

Can we say :woot:, because non of the next generation offering is a huge leap from the current gen. Good for people who just bought their comp.

lol, that dude is known to shout out bulls hit....

Please completely disregard anything he says... I'm a big AMD fan, but this is just fanboy bulls hit...

Not to say I don't agree with the old FSB technique to be a major bottleneck, also it is responsible for very hot northbridges which isn't something you want. Intel should switch to a more serial technique like HyperTransport.

He's claiming Conroe can't run 64 bit windows..... :X

But but, I so want conroe to be no better than my X2.

Honestly Conroe has a few more flaws than that.. :p:

Quote:

---

Sharikou, Ph. D said...

First, anything done at IDF was a setup by Intel cowards.

---

Mmmm sounds unbiased to me!

Quote:

---

Sharikou, Ph. D said...

I read some comments saying the 4MB cache is not a gimmick but can be a real boost for performance in some cases. In some sense, that's true, if you are running something like Super Pi 1M for fun.

---

So I suppose that explains the terrible results obtained with the Celeron-M and it's mighty 1MB cache.

Oh and the "measly 1066MHz FSB" is a 1333MHz FSB. It would be better if it wasn't an FSB for sure.

Quote:

---

Originally Posted by :banana::banana::banana:in

Mmmm sounds unbiased to me!


So I suppose that explains the terrible results obtained with the Celeron-M and it's mighty 1MB cache.

Oh and the "measly 1066MHz FSB" is a 1333MHz FSB. It would be better if it wasn't an FSB for sure.

---

Well he does have a point in that Intel was anal about not sending a box for Anandtech to do independant tests.... If conroe was bulletproof, what are they afraid of?

The 1333Mhz FSB is only available on the Conroe XE CPUs, and even if it was run on a 2Ghz FSB, there will still be at least a 20-25% increase in memory latency compared to the on die approach, there is NO getting around that.

Don't get me wrong, Conroe will whip AMD's ass in almost all gaming benches when it comes out. It's the aging Netburst FSB (need I remind you that it's been around since the FIRST Pentium4?) that I'm worried about especially once SMP games + quad cores come out.

Basically, these benchmarks were able to expose the weakness of the FSB bottleneck whereas most current gaming benches probably won't.

Well, after everybody saw the brute force of Conroe CPU, people are trying to find it weaknesses. I agree that HT is better when it comes to four cores, but I still don't understand, why people are complaining for dualcores. HT doesn't bring a huge increase in bandwich against FSB, but lower latencies. Just think a little, bandwich isn't the real problem for todays procesors. With a dual channel controller, you have almost 3000MB read and over 1000MB write speed per core. 3000Mb means almost 400 milions 64bits longint variables or alomst 300 milions 10bytes reals. The CPU has also to execute jump and compare instructions, and 300mil floating point multiplication/division is too much event for the fastest which will be released this year. And remember, 80% or more bought AMD 64 for gaming. If scores will be the same in July, those people will buy Conroe, nomatter the better performance in ScienceMark.

What currently needs more than the 4mb anyway. Sure if your running stress test designed to fill out the cache then you have problems but 4mb is a BIG cache and I don't forsee many problems at the moment. Saying that in the future it will not perform as well makes perfect sense. That happens to every processor when the next gen comes out.

This is called pure "hateration". Looks like the fanboys are already into mega spin mode. Just so sad... yet amusing.

Good god. I started reading some of his other articles, I had to restrain myself from punching my laptop in the face.

This guy is a babbling idiot.

Jup, as I said, this guy is known to shout out bulls hit....

But honestly, come on. I seen the numbers.. conroe isn't THAT much faster than an FX-60 for example. I mean compare xeons to opterons, thats a big difference. Compare A64 to Conroe or AM2, they all look to be pretty much the same.

Besides if there were any real hardcore gamers, they would use high quality settings. And that means pretty much little to no CPU bottleneck just because the video card can barely keep up.

You're right, but think about marketing. In my country, if you ask most people (with a few knowledge in IT) about a gaming sistem, they will tell you to choose AMD, although a gamer with a fast video card will always play at high quality settings, where there is no difference between an Intel and an AMD cpu (both "averange" cpus). When CPU does matter, it's useless, because human eye doesn't tell the difference between 200FPS and 300FPS. With Conroe, the situation will be the same, all will tell you to choose Conroe, although there will be no diference (unless you will use CPU power to "emulate" Ageia's phisics engine in future games :p: ).

Actually a better CPU ensures slightly higher lows and although the game's not consistently CPU bottlenecked, it is always at times. A few FPS is a lot in the long run. It could make the difference between playable and unplayable gameplay, or even noticing a frame drop. Better safe than sorry, I sayeth.

Yonah/conroe/merom have proven themselves in pi, even 32m by a decent margin. However one must remember that all the top 03/05/06 scores are all amd, and IMHO getting onto "almost the first page" really means nothing.

It remains to be seen if conroe can deal with the lack of an on die memmory controller and Hypertransport bus. All of the 3dmark scores I have seen so far have been rather lackluster, things that the A64 has already acomplished.

Only time will tell.

That a heavily overclocked a64 just beats a stock midrange conroe is not something that even remotely suggests that amd will have an advnatage is 3dmark - wait until we see some overclocked conroe before you compare it with the 3.5ghz+ phase changed fx60's

But we have seen phase changed yonahs, and they are nowhere near capable of what some of the people have been doing with fx-60s and crossfire, because two of the big problems of netburst have been brought along -- memmory latency from no on die memmory controller and still using a FSB and not something like HTT.

Yonah -> conroe wont change that.

Hell, someone with x1900 Crossfire and a 3+ghz Yonah was only 600 points above Afireinside's 7800GT+170 3d05 scores! Only time will tell.

but where are the yonahcrossfire/sli motherboards? Hell even yonah motherboards are pretty scarse- and overclcking ones are pretty much non existant - their hardly an apples to apples comparison. A maxed out yonah on phase would probably beat a maxed out fx60, and conroe will be a furtherstep beyond yonah in virtually all departments, plus there won't be a cold bug...

Quote:

---

Originally Posted by onewingedangel

but where are the yonahcrossfire/sli motherboards?

---

The aopen i975 is retail in Japan right now.

Quote:

---

Hell even yonah motherboards are pretty scarse- and overclcking ones are pretty much non existant - their hardly an apples to apples comparison.

---

In the US mabye, but not everyone @ xs is in the USA :slap: We will get it soon. The voltmods for the Aopen board are even out! I would call that pretty apples to apples.

Quote:

---

A maxed out yonah on phase would probably beat a maxed out fx60, and

---

Oh Probably eh?

http://img53.imageshack.us/img53/3969/untitled3ua.jpg

I do not think Jason agrees with you.

Quote:

---

conroe will be a furtherstep beyond yonah in virtually all departments, plus there won't be a cold bug...

---

Conroe just gets wider execution cores and a different FSB 1066 vs 667.

And you know it wont coldbug, but that isnt that large of an issue.

Quote:

---

Originally Posted by kyosen

My latest results
T2600(ES)@3.1G, X1900CF@792.00~796.50/891

3DMark2001SE: 48425
http://service.futuremark.com/compare?2k1=8946909

3DMark03: 39511
http://service.futuremark.com/compare?2k3=4730996

3DMark05: 17578
http://service.futuremark.com/compare?3dm05=1946510

3DMark06: 11687
http://service.futuremark.com/compare?3dm06=238228

AquaMark3: 135796

Screenshots are here:
http://www.oohashi.jp/c-board/c-boar...ne;no=4324;id=

I received ClockGen-1.0.4.4 from Franck:)
I appreciate his great work as usual:toast:
...I hope formal release in near future...

---

I dont think we've seen anywhere near the maximum potential of yonah yet, its still relatively new and motherboard are not widely avaliable wheras the 939 platform is well estabished. Give it some time and I'm sure yonah will beat fx60 - and as a desktop platform conroe will have more options avaliable.

As I posted in the other thread, this blog entry has somewhat of a point.

As "Core" is still not featuring an integrated memory controller, it will slow down more severely than AMD64 once both trash their caches.

How many applications will trash a 4 MB cache is a different question.

Unreal tournament 2007?

this guy is pathetic, plagiator. He took Victor Wang result and post it as his result.

Can you all see- that all is VictorWang number, minus two benchmark with conroe have advantage.

I said, it's a good hoax.

Sigh...:rolleyes:

First rig: Currently Top of 3DMark05 20667

AMD FX60 @ 3.707Ghz "Used Dry Ice" Can't run that 24/7
2x ATI X1900XTX SLI "890/850"

Everyone agrees this is a heavily OC system


Now we take VictorWang's setup
3DMark 05 11258

Conroe @ 2.4Ghz "STOCK"
1x 7900GTX 512MB "622/1880"

This Conroe scores more than half that of a heavily OC system due to this not being in SLI only one card is running.

Just for fun lets add another card and double the score 22516.Allready higher than the the TOP 05 AMD setup. Now it should probably score a little lower or higher "not sure" in SLI than just multiplying the score like I did by two but this setup is still on a BETA BIOS and a STOCK Conroe and for $500 cheaper than FX60 just on the chip alone:)

Conroe will be very interesting :cool:

Intel Leap Years ahead ^_________^