Sign in with
Sign up | Sign in
Your question

Where come the 2% increase clock for clock in 2.8

Last response: in CPUs
Share
August 27, 2002 5:47:10 AM

Aceshardware anandtech harware.fr have test show that C1 is faster clock for clock.Most of the test run faster wich cab be margin of erorr.

So we get many test that run faster for a unknow reason like why socket 478 running 1% faster for a unknow reason and why C1 stepping in willi have also increase the speed by 1%.I get to the conclusion that intel dont release all the information.

There a rumor of a L1 TLB increase.TBL entry increase is a mistake from CPUID so that a fasle rumor but that is according to intel official.



At the end i have speak with a horny lady<P ID="edit"><FONT SIZE=-1><EM>Edited by juin on 08/27/02 01:35 PM.</EM></FONT></P>

More about : increase clock clock

August 27, 2002 4:03:22 PM

I'm not sure what you are saying.

Some motherboards have slight differences in bus clock than other boards (brands). Just for example, I once saw an article about Athlon motherboards where some had a bus clock at 133 MHz while others were at the true 133.333 MHz while still others cheated at 134 MHz and even 135 Mhz. Don't know if the same kind of thing is happening with P4 mobos. Is this what you mean?

<b>I have so many cookies I now have a FAT problem!</b>
August 27, 2002 5:29:52 PM

No i am have 3 review they all say that the new stepping is faster clock for clock due to some improvement wich intel dont disclose.

At the end i have speak with a horny lady
Related resources
August 27, 2002 5:54:30 PM

OK, I understand now. The translation of your second message was clearer than the first.

I wasn't aware of an speed improvements to the Northwood B due to a new stepping. Thanks for pointing it out. If hear of an explanation I will let you know.

<b>I have so many cookies I now have a FAT problem!</b>
August 27, 2002 6:26:46 PM

I don't think it's a false rumour. Intel probably does not want to make an "official" announcement, so that's why they're denying it. Anand has documented the TLB increase as well, same with HardOCP, and I don't think that they're ALL wrong, or that it's a bug in WPCUID. I think that they did double the TLB maybe in preparation to hyperthreading which will appear on the 3.06G NW.

Also, the TLB usually makes a dramatic difference when large amounts of memory is required, so this also could be a big help on the "Gallatin" Xeon, which is coming out in a few months.

The point is, Intel doesn't ALWAYS tell the truth (especially when it comes to x86-64 extensions).

There IS a chance that it's a bug in WPCUID, and all these sites are wrong. But I'm doubtful about it.
- - - - -
Tejas - "supposedly" the last P4 revision
[successor to Prescott]<P ID="edit"><FONT SIZE=-1><EM>Edited by Dark_Archonis on 08/27/02 02:29 PM.</EM></FONT></P>
August 27, 2002 6:46:22 PM

The way I perceive, it can be true, but also it can be that the new bus and RAM has finally unlocked its true potential. Think about it, you could give an Athlon at 650MHZ a 266MHZ bus and mem, though the difference in performance will grow weak. In time though there'll be a speed that finally uses that efficiently, thus the performance suddenly jumps higher.
That's just my theory.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!<P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 08/27/02 04:10 PM.</EM></FONT></P>
August 27, 2002 6:47:33 PM

For some reason, naming a server/workstation chip to something that looks and sounds so close to gelatin just seems wrong...

Look ma, it's as fast and wobbly as Jello! ;) 

It reminds me of trying to sell a Nova in Mexico...

<pre><A HREF="http://www.nuklearpower.com/comic/186.htm" target="_new"><font color=red>It's all relative...</font color=red></A></pre><p>
August 27, 2002 8:02:13 PM

Ya, the codename is pretty lame. Usually, Intel has some interesting codenames.

Eden, that's an interesting theory. Come to think about it, it could be the case. But, that doesn't explain the anomaly that Ace's discovered. After underclocking the 2.8 to 2.5, he tested it against the "old" 2.5, and the "new" 2.5 showed a 2-3% performance improvement.

- - - - -
Montecito - successor to Madison, and one monster of a CPU.
August 27, 2002 8:09:14 PM

Well there is also the margin of error factor here as well, ya can't use a 3% +/- only! (sorry if I sound persistant against the P4's performance here!)

But yeah it could be improved TLBs, though it's pretty much the improvement amount that the Palomino got as well, so it's not the main place to update. Still a nice thing, though it'd been cool if the 2.53GHZ or lower were using it. More competition towards AMD. Makes you wonder if the 3 other new CPUs by Intel aren't also using this improvement?

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 27, 2002 8:17:25 PM

Most of Intel's codenames are based on things in Oregon (Intel has a big presence over here). Like the first P2 core was "Klamath". The second P2 core was "Deschutes". The .13u P3's were "Tualitin". And the first P4 core was "Willamaitte".

Knowledge is the key to understanding
August 27, 2002 11:18:10 PM

I have able to find some information on intel website all P4 have 128 entries it CPUID that make a mistake now with the C1 stepping it appear to solve this issue.

There 6 minor change giving 0 to 5 % increase manly in Scienmark and others FPU test.

At the end i have speak with a horny lady
August 27, 2002 11:19:31 PM

I have able to find some information on intel website all P4 have 128 entries it CPUID that make a mistake now with the C1 stepping it appear to solve this issue.

There 6 minor change giving 0 to 5 % increase manly in Scienmark and others FPU test.Anandtech as receive word from intel like normal they dont give much information.

At the end i have speak with a horny lady
August 27, 2002 11:21:46 PM

all cpu over 2.5 will have C1 stepping but only the 2.56 will be buyable now with the 62% price cut.

At the end i have speak with a horny lady
August 27, 2002 11:33:06 PM

And so that means what, that all new C1 steppings like 2.53GHZ C1 should have an IPC increase, is that what you're saying?

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 28, 2002 12:44:16 AM

If you read all those reviews again, most of them (especially Anand's) have been updated confirming that there was no TLB increase.
My best guess as to the increase in IPC would be the thermal throttling. The newer stepping reduced heat and optimized critical paths. This means that at the higher speeds, the P4 would be throttling less. Since it is such a small increase, it may be that the P4 was not throttling a lot.
This is just a wild guess though. I really have no way of confirming this. The only real way would be if someone were able to disable the thermal throttling on the P4's. Maybe someone would take a watercooling or some other extreme cooling setup that would guarantee a P4 won't throttle and cool both P4's, and test the two?

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 28, 2002 3:15:58 AM

Woah woah woah slow down buddy, if I got you right, P4s have been throttling all this time? What the crap? That could seriously mean Intel has problems maintaining their P4s at optimum!
I don't beleive this unless there is proof, throttling would mean that they are NOT being good at keeping the processors up and running, and that could be very bad if they were put in render racks.

Then again you said your best guess, so I'll leave it as that, and once there are facts, we'll put that as a claim!

I still personally beleive the FSB and Memory have finally attained the peak power, where the CPU is finally using it all optimally.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 28, 2002 4:05:13 AM

Well yeah, that's what a heat speader does. It reduces the chips clock speed when the CPU's temp gets to a certain point so they don't burn themselves up (unlike AMD's chips).

Knowledge is the key to understanding
August 28, 2002 5:34:31 AM

No that false all P4 have 128 entries all of them.

At the end i have speak with a horny lady
August 28, 2002 5:38:05 AM

that for extreme use (like in case of a FAN defection) i do SETI all the time and my score dont do down that trolling and poor information from low-end reviewer or you.

At the end i have speak with a horny lady
August 28, 2002 5:50:23 AM

N50 : Incorrect data may be returned when page tables are in Write Combining (WC) memory space
N52 : Processor issues inconsistent transaction size attributes for locked operation
N58 : CR2 may be incorrect or an incorrect page fault error code may be pushed onto stack after execution of an LSS instruction
N59 : BPM[5:3]# VIL does not meet specification
N62 : L2 cache may contain stale data in the Exclusive state
N65 : Processor does not flag #GP on non-zero write to certain MSRs



At the end i have speak with a horny lady
August 28, 2002 2:10:56 PM

Thermal throttling doesn't reduce the performance significantly if the temperature drops immediately below a certain critical point. Unless your CPU continues to rise in temperature even when it throttles down to one level, it'll continue to throttle and that's when you'll see a significant performance decrease. Realistically, at the higher clockspeeds with good cooling, you should not see much of a difference as it'll be throttling very little and only for a short time. Although this may explain the 2% IPC difference for the new stepping.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.<P ID="edit"><FONT SIZE=-1><EM>Edited by imgod2u on 08/28/02 01:40 PM.</EM></FONT></P>
August 28, 2002 5:40:34 PM

SETI use allwayse the same part of the cpu at 100% according to WIN 2K.SO after 12 hour it will be allwayse in throttles.YOu can do a sisoft benchmark during your sessesion and you will get the same score.I keep conclusion.

At the end i have speak with a horny lady
August 28, 2002 6:38:10 PM

Dude, I don't get you, now you're saying the higher the speed, the more the throttling, the better the performance boost? You're using reverse analogy!

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 28, 2002 8:43:50 PM

It was a typo. I've fixed it. The more it throttles, the less speed you'll get. Also, we're not exactly sure where the diode for the thermal manager is. It may not be anywhere near the part of the CPU that Seti uses. We also don't know whether even at 100%, the CPU is throttling at all. There is no direct way to know unless you have either:

1. some way to disable throttling and test 2 CPU's
2. supercool a CPU vs regular cooling and see if it makes a difference.

Remember, throttling depends on temperature, not heat generation. The two aren't neccessarily the same thing.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 28, 2002 9:14:07 PM

Yeah but if the P4s throttle all the time as you seemed to think, that would make them very low quality silicons. That means Intel can't hold a core stable at a temp to perform as it should!

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 28, 2002 10:15:56 PM

Throttling is to make sure the temp never goes too high. No one can guarantee the temperature of a CPU. The previous P3's just kept getting hotter until a critical point. Intel added this in the P4 to make sure it never gets past even the lowest of safe points. As for "constantly", I wouldn't say constantly. But frequently enough to cause a 2% difference I guess. If what I think is true. Intel's documents say the thermal manager's diode is placed in the hottest place on the CPU and will start throttling at 55C. That's pretty low for the hottest place on the CPU. However, since the performance difference apparantly isn't that much anyway, I would say it doesn't happen a lot either. Throttling has no bearing on stability. Your system will run just as stable, so I don't know where you're getting at. Would you rather let the CPU just run at high temps instead of attempt to cool down?

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 28, 2002 11:48:30 PM

If 55ºC is throttling, Wilamettes would be performing worse than they were. 55ºC is not uncommon, and if that is a throttle temp, then the silicon Intel deals with must be weak. AMD CPUs don't lock at these temps, well rarely will you stumble on a chip like that. The rest will lock up over 70 or just before the fry temp of 90.

But 55ºC to me would indicate that their chips are not capable of handling higher temps without locking up, thus being unstable at high temps. Are you saying that? I hope you're wrong, because IMO that truly says alot about Intel's chip quality. Not so much as how good throttling is, but as to why it was put in the first place. Is it just to prevent chip frying, or is Intel the only one aware the P4 silicon would not handle high temps?

I would be damn happy to have thermal protection like the P4, but if it's used like this, then that's more wreckless handling than prevention. I still wonder if that is true, then how the heck do render farms survive and perform.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 29, 2002 12:10:58 AM

If you have a good cooler, air or otherwise, then it should not throttle. My friends p4 is at 38 C, and I don't think that it throttles. Plus, most p4's don't reach 55 C unless they are OC'ed or weren't installed properly. But I doubt that's what's causing this "anomaly" or performance difference at aces, as well as HarcOCP.

- - - - -
Montecito - successor to Madison, and one monster of a CPU.
August 29, 2002 3:22:41 PM

Quote:
If 55ºC is throttling, Wilamettes would be performing worse than they were. 55ºC is not uncommon, and if that is a throttle temp, then the silicon Intel deals with must be weak. AMD CPUs don't lock at these temps, well rarely will you stumble on a chip like that. The rest will lock up over 70 or just before the fry temp of 90.


At 55C, the chip throttles at its lowest level. Such a level would mean running a 2.0 GHz chip at say 1.9 GHz or something similar to that. If the temperature still doesn't decrease, it goes to the next level until it does. It doesn't "lock up", it purposely decreases the clockrate to lower temps. This insures that a CPU does not overheat or even gets to a point where it impacts the CPU in any way. Judging from the P3, it prolly could go higher, but Intel wants to make sure it doesn't. Who knows.

Quote:
But 55ºC to me would indicate that their chips are not capable of handling higher temps without locking up, thus being unstable at high temps. Are you saying that? I hope you're wrong, because IMO that truly says alot about Intel's chip quality. Not so much as how good throttling is, but as to why it was put in the first place. Is it just to prevent chip frying, or is Intel the only one aware the P4 silicon would not handle high temps?


Umm, they don't "lock up". Have you seen how throttling works? The chip is scaled down to a lower frequency. 55C is quite high in terms of temperature for Silicon. Above 60, electron gates start wearing down more and more. The reason it's at 55 is to give the throttling circuitry a buffer zone so that it never reaches 60. The maximum operational temperature for Silicon is 90C. The CPU should get nowhere near that.

Quote:
I would be damn happy to have thermal protection like the P4, but if it's used like this, then that's more wreckless handling than prevention. I still wonder if that is true, then how the heck do render farms survive and perform.


It still performs great. Look at the benchmarks. If throttling were that big a hit on performance then it should run like sh!t now shouldn't it? It's a gradual, self-adjusting process and unless the temperature becomes too high, it won't throttle down to a point where you'll even notice it. Think about it, asside from in benchmarks, where else are you gonna notice the difference between 2.0 GHz and 1.9 GHz for 1.5 seconds before it goes back to 2.0 GHz?

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 29, 2002 9:09:11 PM

Wouldn't notice.

But as I said, lock up as in if it had without thermal protection throttling, it woulda locked up at around it. That is only IF that was the case, that they did it because P4 silicon with no protection can lock at rather low temps of 60 and 70, compared to 90.
Personally I'd put it at 65ºC, the P4's reader is very fast to respond, since it is IN the die, and since it can communicate as fast as a simple 20 cycles in a second, you'd assume even at 65ºC and rising it'd have time to throttle.

Although unnoticeable at first, it does make benchmarking rather unreliable or not exact. The end result is like what THG does, not use WinXP because it uses different memory management.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!
August 30, 2002 4:39:45 AM

Well, the P4 has gone over that temperature. Just look at when Tom took off the heatsink from the HD. The throttling eventually brought the temperature down but it went well over 55C. I should also stress that the response time for the die material isn't instantaneous. Meaning if you begin to throttle, it wouldn't neccessarily instantaneously reduce heat. It may take 1 second or 1 ms but that could be the difference between the throttling circuitry capable of working or not. Intel feels it is best to keep the CPU below 60C and frankly, I agree with them. 60C is nowhere to be going with a .13 micron, 55 million transistor MPU.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 30, 2002 6:06:35 AM

Do you have any link or you try to provve something.

Any p4 never throttle that a safety mesure like do you air bad blow each time you brake no?

At the end i have speak with a horny lady
August 30, 2002 2:01:31 PM

Prove what? That the CPU shouldn't go over 60C? Or the P4's throttling. I would think the throttling part would've been widespread knowledge by now.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 30, 2002 3:47:47 PM

Forget the P4, I just wish I knew what juin's sig meant.

Just because you're not paranoid, doesn't mean they're not watching you.
August 30, 2002 3:51:43 PM

P4 can go way over 60%

widespread

that dont mean it true

At the end i have speak with a horny lady
August 30, 2002 5:15:17 PM

Can and should are two completely different things. The Athlon could probably operate at 80C, doesn't mean it should. When the thermal diode reads 90C, there could be many other parts that approach 90C. Silicon's maximum operational temperature is 90C. So you figure it out.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
August 30, 2002 7:37:38 PM

Just because it's always throttling, that doesn't mean that it's low quality silicon. It simply means that Intel is raising it's expectations when it comes to how cool their CPU's should run. That's interesting, I always though that the p4 starts throttling at about 60 C, not 55. Ideally, your CPU shouldn'r run any higher than about 45 C, because that's the lowest of safe points. I mean this is really low. Starting at about 45 C, for every 10 degrees <i>higher</i> your CPU is running (than that safe point), the component life of the CPU is reduced by half. Now, this information is still semi-valid for P4's and XP's, even if they are using better materials, or packaging. So, if your CPU is running at about 65 C, then the component life is reduced by 200%. But, according to Intel white papers, the Northwood can run fully stable at up to 66 C. Also, I'm not exactly sure whether this rule is still fully valid for the XP. I know that this was valid for the P3 and Tbird. But either way, I agree that it (p4) should start throttling at least at 55 C. And if it is true that it throttles at this temperature, there is no notciable performance difference at all.

Imgod2u, I would be interested in actually seeing some solid proof about your "throttling theory", because I have never heard about this anywhere before. Also, about that 2% performance discrepency, I think I have the answer. Head over to the "the truth about the performance of the 2.8 P4" topic that I posted to read more.

- - -
All good things must come to an end … so they can be replaced by better things!
September 1, 2002 1:16:45 AM

If you're refering to the actual temperature in which throttling first starts, I got the idea from here:

<A HREF="http://forum.oc-forums.com/vb/showthread.php?s=&threadi..." target="_new">On the OC Forums</A>

If you're refering to something else, specify what you want proof of.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.
!