Wednesday, February 21, 2007

Patty sounded scared

Facing the imminent onslaught of K10, Pat Gelsinger's only hope is higher clocks and faster FSB.

The problem is, Intel does not have any edge in the clockspeed arena. AMD's 3GHZ K8 is within the strike range of Conroe and K10 will offer 40% bump in speed.

All Intel had with Conroe was about 10% integer IPC lead. That's why a 3GHZ K8 can frag a 2.67GHZ Conroe E6700: 3/2.67 = 1.12 = 112%. 12% clockspeed lead is enough to beat back 10% IPC lead.

7 Comments:

Blogger Ho Ho said...

Sharikou
"12% clockspeed lead is enough to beat back 10% IPC lead"

Depends on benchmarks

http://techreport.com/reviews/2007q1/athlon64-x2-6000/index.x?pg=13
"Performance-wise, the X2 6000+ is slower overall than the Core 2 Duo E6700, but not by much"

x2 also takes ~60% more power than e6700. For me it starts to resemble Netburst vs Athlon: one uses higher clocks to beat the other, too bad it also meant higher power usage.

11:55 AM, February 21, 2007  
Blogger gnoppi said...

To me, Hector Ruiz in the last two years made the right decisions at the right time.
He put pressure on Intel with performance when Intel wasn't ready with his old P4, forcing it to create the Core solution: at the cost of lot of cache, and lot of dumping of legacy P4s.
Then he let Intel win the peformance battle, but at the cost of piles of P4 crap unwanted, because AMD focused on gaining marketshare.
Now with the K10 they will regain the performance crown WITH the marketshare gain. Hat off to Hector!! IMHO, AMD should stress that K10/Barcelona is NOT a mere core-update to Opteron dual core but a new chip (which is, due to new/improved FPU units, L3 cache,lots of microarchitecture enhancements/improvements).
And remember: Intel's power usage is always TDP (typical) and not at full load; it's always the same Intel's marketing strategy. What does it mean "typical"? For a office desktop user? For a user like me that runs Grid.org computing and 3d modelling 24hrs a day typical is near maximum. Typical is an average rate that doesn't make sense, the Intel quad core reaches 220 watt on full load.
So, 90 watt for AMD means at full load, 90 watt for Intel means an average.

12:09 PM, February 21, 2007  
Blogger Ho Ho said...

gnoppi
"And remember: Intel's power usage is always TDP (typical) and not at full load"

There have been rumours that AMD will start using the same nubmers.

gnoppi
"Intel quad core reaches 220 watt on full load"
Can you show some benchmarks that show similar number? Or was that for the entire box and not only for the CPU?

12:28 PM, February 21, 2007  
Blogger Cenuijmu said...

I disagree that intel does not have the edge in the clockspeed arena at least potentially. The AMD K8 seems to be topping out at about 3.2Ghz where as an Intel X6800 will happily run all day at 3.8Ghz. Mine does over 4Ghz. That is a lot more overhead.

Gnoppi, I doubt Hector Ruiz "let" Intel win the performance battle, do you really think Hector said " Lets have lower performing cpu's so that when Intel reduces the price we have to at least follow suit? No, Hector would have been very glad indeed if Core was not much better than P4 performance wise.

You have a good point about market share being a good springboard for K10, but until K10 AMD will have to bite the bullet on profits.

Finally do you think Hector made a good decision in buying ATI ?

10:29 PM, February 21, 2007  
Blogger Ho Ho said...

cenuijmu
"I disagree that intel does not have the edge in the clockspeed arena at least potentially"

I haven't said they don't have the edge, I just said they aren't using it. They'll release 3Ghz quadcores and 50W quadcores soon. Also 45nm is stil on track for this year, possibly in Q3.

They will also increase FSB up to 1.6GHz and for some reason they rename it to Processor System Bus. That means 400MHz per core, just as much as Sharikou said would be needed for feeding a single core.

2:44 AM, February 22, 2007  
Blogger enumae said...

Sharikou
That's why a 3GHZ K8 can frag a 2.67GHZ Conroe E6700...

Not including Integer, the average margin between the E6700 and the X2 6000+ in benches won by Intel is 17%.

The everage margin between the E6700 and the X2 6000+ in benches won by AMD is 17%.

So looking at the results of benchmarks with more than a few percentage points (while not including Integer), that number being 21, AMD won 6, while Intel won 15.

I do not see a frag here, just results that show how each processor is better in certain applications.

On average the advantage belongs to Intel considering it won more than double the amount of benchmarks while having a 13% clock speed disadvantage.

11:09 AM, February 22, 2007  
Blogger TheKhalif said...

I disagree that intel does not have the edge in the clockspeed arena at least potentially. The AMD K8 seems to be topping out at about 3.2Ghz where as an Intel X6800 will happily run all day at 3.8Ghz. Mine does over 4Ghz. That is a lot more overhead.

Releasing a chip that is warrantied to run at that speed is not the same as releasing a chip that can OC to a certain speed.

1:05 PM, February 22, 2007  

Post a Comment

Links to this post:

Create a Link

<< Home