Intels Conroe benchmarked!

chavvdarrr said:
Yes, running AMD system with old BIOS which can't run FX60 or better CPU with 2-2-2-5 1T timings is crippling , no ? ;)

Anyway, I'd expect someone here to get a FX-60 system with decent MB, update the bios :), then to run few becnhmarks with 2 x X1900 ... that should be quite decisive no?

PS btw I don't think that 20% on average, as claimed by Intel is too much.
Suppose AMD can get 10% from going DDR2-800 and add 5-10% from doubling the cache (the test was comparing Conroe EE with 4MB shared cache vs FX with 2MB total L2)
Although I prefer AMD to lower the prices for X2 3800+ ;)


intel is mammoth. you're really underestimating the gains conroe technology is showing. We're talking about a very noticable boost to CPU/GPU intensive situations (not some UT benchmark at 800x600 junk), at a slower speed then AMDs top end chip. I'm sorry to inform you of this but AMD really has very little plan to change their chip technology in terms of work per clock this year, but a 10% boost across the board from going to DDR2 800 would be amazing. Memory timings also play a very little part with AMD chips because latency is so low due to onchip memory controllers. Many people have run tests showing that you can run timings tight or loose and the gaming performance difference is practically zero. Intel has always been more problematic with having to keep timings as tight as possible. And cache gains in AMD chips wont do too much to single threaded program performance unfortunetly. But multi-tasking should be a bit more snappier.

AMD is still the king, but conroe is showing a very serious reason to upgrade. Even 15% would be nothing to snicker at. Lets not forget that this is 15% over AMD's best, compared to their own Prescott and yonah chips, this is a HUGE leap for intel. However, we must not forget AMD still hasnt gone to 65nm which conroe is, so AMD can very well continue to dominate in power areas.
 
Last edited by a moderator:
Well given how overclockable AMD chips generally are, and the fact they are moving production to their new fab 35, which surely has an improved process, I suspect they may have a good amount of room to jump their clocks up. Plus, arn't they doubling the FPUs in their monster socket F Opteron line with the 4 memory channels? There's nothing stopping them from doing that on the desktops too.
 
DudeMiester said:
Well given how overclockable AMD chips generally are, and the fact they are moving production to their new fab 35, which surely has an improved process, I suspect they may have a good amount of room to jump their clocks up. Plus, arn't they doubling the FPUs in their monster socket F Opteron line with the 4 memory channels? There's nothing stopping them from doing that on the desktops too.
try overclocking an FX 57/ FX 60 ;)
a64s hit a wall at 2.8~ ghz for singlecores and dual cores pretty much unless you use exotic means to cool it.

What he have here.. is a failure to communicate.
Intel we be king untill amd updates their per clock performance and of course gets on 65nm.
Just admit it.
 
Last edited by a moderator:
Well given I haven't seen benches of AMD's AM2 CPU, I'll just wait and see before jumping to conclusions. The Conroe is only about 15% to 20% faster, I think AMD could catch up in the 1/2 year it will take for the Conroe to come out. Also, no one knows the yeilds on the Conroe, so AMD may be more cost effective anyways. Then again, if Intel ramps up performance much farther from this current prototype, I may have Intel Inside my next computer.
 
DudeMiester said:
Also, no one knows the yeilds on the Conroe, so AMD may be more cost effective anyways.

The core is about the size of a Northwood, way smaller than a Prescott. Intel also ditched the use of high performance high leakage transistors. So overall yields should be excellent (cache has build in redundancy, so doesn't affect yield).

AMD need K8L ready when Conroe launches.

Cheers
 
ANova said:
I see, so the old bios can't run the FX-60. How is it that they even managed to run the computer to get those benchmark scores then? Whether or not it detects the cpu correctly matters not in memory settings, nor performance in most cases.
==========
Where's that popcorn smiley.
So, do you claim that updated BIOS decreased K8's score ? :oops:
Or someone expected 30% boost from updated BIOS ? :)

10% going to DDR2 and an extra 5-10% doubling the cache huh? I'm sure AMD would love it if those numbers were true. Tell me, if cache is so important, why is it that the 1 MB cache A64s beat the 2 MB cache P4s in gaming by as much as 10%?
Ok, what about a bet ?
I say that K8 CPu running on AM2 + DDR2-800 will be 5 to 10% faster than K8 CPu with same frequency and cache sizes running on s.939 with DDR400

Agree or disagree ?
 
ANova said:
Where's that popcorn smiley.


popcorn.gif


This one!
f_eyebrows.gif
 
AMD/Intel battle aside I think it's amazingly cool that anyone got a 40% bump on the current leader. If AMD can match then all the better for us consumers, but I'm giddy over 40% even though I tend toward the green camp.
 
This is the first time since Norhtwood that I'm looking forward to an Intel launch.
Those numbers are absolutely amazing!
 
Heck yeah, I may finally get another boost in CPU performance, the slow increase in speed has been frustrating. In the past I used to upgrade when speed doubled roughly, but I had to stop that :)
 
Tom's has a test of a am2 preview board.

chavvdarrr said:
Ok, what about a bet ?
I say that K8 CPu running on AM2 + DDR2-800 will be 5 to 10% faster than K8 CPu with same frequency and cache sizes running on s.939 with DDR400
Agree or disagree ?
This test compares the memory clock and timings used in the intel machine, with the ones of the amd machine.
So far it gets slower!

I'd disagree.
 
Realworldtech has a very good summary of the latest Conroe info.

Conroe definitely expanded Intel's fp bottom line, and basically tripled fp width compared to P4.

It can issue a 128-bit FADD and an 128-bit FMUL/FDIV per cycle, or 3 SSE ops.

AMD definitely can't match Conroe on SSE or 128-bit math. The summary didn't mention which units did x87, so there's the chance that AMD can hold onto a scant lead in that dead end if Intel seriously deprecated x87.

If the SSE units don't have much emphasis on x87, K8 might not drop so horribly behind as it would with SSE. Considering how much a pain x87 is performancewise, this is a pale hope.

On the integer side, Conroe definitely is far more robust than P4 and Yonah, and is 64-bt to boot. K8 is relatively matched with Conroe on integer unit counts, but the reservation station capacity and reorder buffer are larger on Conroe.

The front end is also very improved, with some real nifty tricks borrowed from P-M.
In addition, Conroe has an advantage in that it can reorder memory accesses, and is better at it than the P4 was. K8 can't do any of that.

I am very impressed with Conroe. AMD definitely has cause to worry, since Intel is going to very easily muscle in the market gains made in desktop and servers.

K8L won't erase the fact that its architecture is getting a little long in the tooth.
 
I tested the diff between PC4000 and PC3200 on my dual core Opteron @ 2.5. It was a whopping 5% in Divx 6.1.1 (multithreaded) + a simple AVISynth script. I had my latencies about as tight as possible, a lot tighter than DDR2 typically is. So, don't expect much people.

A64 can not hit 3.0 GHz without significant extra voltage (=heat) usually. So, unless AMD is willing to do that, you aren't gonna see 3.0 GHz on 90nm for certain. I've been following the o.c. scene for about a month and a half now, since I got this dual core, and I'd say the top air cooling is usually 2.8GHz or so for Opteron 165/170. FX60 seems to hit 3.0 GHz a bit more often, obviously cuz it's the cream of the crop. But 2.8 GHz really seems to be where the need for cooling and volts spikes up. 65nm might get them a bit more, but I doubt it will do much. Few hundred MHz I'd imagine.

Conroe sounds like what P3 was to the K6. :)
 
Last edited by a moderator:
chavvdarrr said:
ok, bet accepted, AM2 is still ~ 3months away. Anyone else? :)

Let me be clear, I meant: get the gap from a 20% wich is today to a 10-15%. Not get 5-10% faster. If you use some super-duper ddr2 800 with timings 2.0 2 2 2 It will surely be at least 5% faster, but so would be the Intel platform.
Still today ddr2, mostly because of it's timings gets the platform 1% slower.
 
DudeMiester said:
Well given how overclockable AMD chips generally are, and the fact they are moving production to their new fab 35, which surely has an improved process, I suspect they may have a good amount of room to jump their clocks up. Plus, arn't they doubling the FPUs in their monster socket F Opteron line with the 4 memory channels? There's nothing stopping them from doing that on the desktops too.


how many chips are already overclockable that are already at 2.6 or 2.8GHz? Its the 1.8-2.2 GHz chips that have been very kind, especially the 939 opteron series. Any higher then 2.8 though for stable operation and most people will be looking into water or phase cooling. Few hit 3 on air. This is not a step AMD would want to take, what they should be doing for the short term is try to undercut intel in the price/performance ratio for low and mid end processors, which will be very difficult to do. Not many people will find the FX processors and higher ends appealing at hundreds when intel can do the same work as them at hundreds of MHz less. A 2.2 conroe is on par with a ~2.8AMD, and that is going to cause serious headaches.
 
chavvdarrr said:
So, do you claim that updated BIOS decreased K8's score ? :oops:
Or someone expected 30% boost from updated BIOS ? :)

The latter. The numbers remained mostly the same after the bios update.

Ok, what about a bet ?
I say that K8 CPu running on AM2 + DDR2-800 will be 5 to 10% faster than K8 CPu with same frequency and cache sizes running on s.939 with DDR400

Agree or disagree ?

I expect between a 0-5% increase going to DDR2 and 1-5% from a doubling of the cache.
 
Back
Top