TBred B 2100+ still overclocks better than the TBred B 1.5V 1700+ DLT3C ?

[hitechjb1]

I import the CPU database of this forum of about 800 TBred 1700+ and 2100+ CPU listed below, into a spreadsheet.

These are some findings from the data for further discussioins:

1. The TBred B 1.6V 2100+ still the winner, max CPU frequency averaged about 110 MHz over the TBred B 1.5V 1700+ DLT3C.

2. The TBred B 1.5V 1700+ DLT3C, max CPU frequency about 160 Mhz over the TBred B 1.6V 1700+ DUT3C.

It shows also how well the 1.5V 1700+ DLT3C overclocks compared to the 1.6V DUT3C, and the 1.6V 2100+, all are Tbred B.

CPU data used:
1. 335 TBred B 1700+ (which include DLT3C and DUT3C)
2. 55 Tbred B 1700+ DLT3C only (This is from the 1700+ DLT3C poll put up by {PMS}fishy)
3. 156 Tbred A 1700+ (this is just for counter check the numbers)
4. 237 Tbred B 2100+

These are the results:

TBred B 2100+
Mean CPU frequency = 2338 MHz
Std Dev = 174 MHz

TBred B 1700+ DLT3C only
Mean CPU frequency = 2226 MHz
Std Dev = 236 MHz

TBred B 1700+ (all DLT3C + DUT3C)
Mean CPU frequency = 2147 MHz
Std Dev = 195 MHz

TBred A 1700+
Mean CPU frequency = 1877 MHz
Std Dev = 167 MHz

Remarks:

1. In the poll for the 1.5V 1700 DLT3C, the sample of 55 is not as large as the 2100+ and 1700+ (DLT3C+DUT3C), and so may be skewed a bit. Also, there may be some errors of DUT3C being entered there and so lowered its oc frequency a bit. So I think the difference between the 1.5V 1700+ and 2100+ should be a little bit smaller.

2. Both air cool and liquid cool CPU data are used.

3. It remains to be seen whether the 1.5 V DLT3C can be clocked higher than 2.5 GHz if more setup are being worked on with liquid cooling (since the DLT3C is still new, no many aggressive oc yet). That will help the DLT3C to bring the oc frequency average higher. I heard there are DLT3C oc to 3GHz (not in the poll data used).

 

[hitechjb1]

From the number through the database (see last post), for max CPU frequency:

1. The TBred B 1.6V 2100+ still the winner, max CPU frequency averaged about 110 MHz over the TBred B 1.5V 1700+ DLT3C.

2. The TBred B 1.5V 1700+ DLT3C, max CPU frequency about 160 Mhz over the TBred B 1.6V 1700+ DUT3C.


If we define overclock ability as

oc = (max_CPU_frequency - rated_CPU_frequency) / rated_CPU_frequency,

then the 1.5V 1700+ DLT3C is for sure the winner.

1700+ rated frequency = 1.47 GHz
2100+ rated frequency = 1.73 GHz

average oc for 1700+ DLT3C = (2.23 - 1.47) / 1.47 = 52%
average oc for 2100+ = (2.34 - 1.73) / 1.73 = 35%

 

[Dirty_Punk]

If we define overclock ability as

oc = (max_CPU_frequency - rated_CPU_frequency) / rated_CPU_frequency,

then the 1.5V 1700+ DLT3C is for sure the winner.

1700+ rated frequency = 1.47 GHz
2100+ rated frequency = 1.73 GHz

average oc for 1700+ DLT3C = (2.23 - 1.47) / 1.47 = 52%
average oc for 2100+ = (2.34 - 1.73) / 1.73 = 35%

yes, but i think that the most of people want the cpu that can go highter if they use a nforce 2 mobo

they want the cpu that have better cost/performance if they use their old mobo (kt266a, kt333 or else)

howether from my test (only 8 cpu 2100+ for now) from a 2100+ you can expect about 2166mhz at default vcore on air (with horrible AL heatsink)
from my test with 8 2100+
2 can do approx 2250mhz 1.6v
1 can do approx 2300mhz 1.6v
3 can do apporx 2166mhz 1.6v
1 can do 2150 1.6v
1 can do only 2100 1.6v

5 are 0306, 1 are 0251, 2 are 0302

never find yet a 2100+ that can do less than 2100mhz 1.6v, but you see some 1700+ that can do 2000mhz but only at 1.8v...

ah, i haven't a good mobo for o/c, i use the MSI KT4-V with vcore that go from 1.55 ro 1.62 when you set 1.6v, so isn't the mobo that do the difference in my test

 

[maxima88]

If we define overclock ability as

oc = (max_CPU_frequency - rated_CPU_frequency) / rated_CPU_frequency,

then the 1.5V 1700+ DLT3C is for sure the winner.

1700+ rated frequency = 1.47 GHz
2100+ rated frequency = 1.73 GHz

average oc for 1700+ DLT3C = (2.23 - 1.47) / 1.47 = 52%
average oc for 2100+ = (2.34 - 1.73) / 1.73 = 35%

I'm sure I don't speak for myself when I say that what really matters to an overclocker is how fast you can get your rig to go.
Whether it's 2.3ghz or 2.8ghz, it's all about speed and how much faster your pc can run compared to your fellow overclcokers.

Being able to o/c your cpu by 52% or 35% is secondary. I've seen a Duron o/c by almost a 100% of it's default speed. But who cares if a Duron can run at 1.6ghz compared to default 800mhz. I'd rather have an Athlon XP that can only o/c by only 35% but run at 2.3ghz.

 

[hitechjb1]

From the same data, I also has the max oc frequency for them:

The top 10 absolute overclock frequency are:

2100+: 2910, 2808, 2800, 756, 2750, 2750, 2730, 2702, 2691, 2640

1700+: 3487, 3003, 2710, 2625, 2576, 2550, 2541, 2536, 2512, 2508


Except for two data points for the 1700+ (3487??, 3003), the 2100+ still seems to be consistently ahead by about 100 MHz.

It remains to be seen whether the 1.5 V DLT3C can be clocked higher than (over a bigger sample) if more setup are being worked on with liquid cooling (since the TB B 1700 DLT3C is still new, no many aggressive oc yet as the 2100 which has been around longer).

 

[hitechjb1]

One may ask why single out the 1700 and 2100?

If you look at the CPU database of this forum, these are the distribution:

CPU entry average_max_frequency
1700 493 2054 <-- local peak
1800 215 1999
1900 3 2109 (sample too small)
2000 23 1936
2100 237 2338 <-- local peak
2200 37 2120
2400 97 2355
2600 16 2419
2700 17 2495
2800 5 2724

So the most popular CPU reported in the database are 1700, 1800, 2100. I think they combine pally, TB A and TB B.

Observations:

1. There are two peaks for the 1700 and 2100.

2. Average max frequency tracks according to the rated frequency (or speed) of the CPU I.e. what you get is what you pay for, even for overclocking.

Explanation:

Most of us would believe 1700 and 2100 are better (esp the 1.5V DLT3C on a relative basis) as shown from these numbers (peak in average oc frequency).

It does not conclude that 1700 and 2100 are better overclocking CPU. I think they are mostly bought by overclockers who tends to stress them as much as possible (including better HS, liquid cooling, ...), and so skew the results a bit higher. These two have the largest entry in the database.

 

[forlack]

There are too many variables to simply run a stats study and pull out local peaks. The database has chips that are old as well as new, but any old joe can type in a number and any old joe isnt a serious overclocker. I am not here to say which is better I am just here to say that the database gives a good indication of what a chip can do, but it does not tell you which chip is necessarily better. This goes for the 2100 as well. To do a correct study you need to control as many variables as possible and this means you need to oc the chips yourself (more than one of each). I wont ask you to do this, but my argument is the database isnt the best to pull samples from. I know you know this, but too say that a chip is better based on these stats is not correct. I understand that you are working with what you can and i do appreciate your calculations. They are very intersting.
Well I havent done some stats in a long time so I thought I would.
haha dont call me a hypocrite.
Based on your data using the samples 55 tbred 1700s and 237 2100s. These are the confidence intervals for the 2 chips.
1700(dtl3) 95% t-interval (2274.2, 2401.8)
2100b 95% t-interval (2315.7, 2360.3)
I ran a 2 sample t test with an alternate Hypothesis that the 2100 > 1700.
With the given data the probability that the two samples are equal and the given data was produced is .000731. So basically we can accept that the 2100 has a greater overclock speed than the 1700, based on this data.(which i said is not a great sample)

 

[hitechjb1]

Thanks for the feedback.

This is just an attempt based on the "best" we have to do some relatively objective discussions than subjective arguments. This is nothing compared to what are done in labs like intel, amd, ..., they have control environment, tests, ... to deal with a large, control samples.

Agree that the database data is not confident by any means. Actaully it mix pally, tbread A, tbread B for some CPU range, experimental errors, ... but it still can reveal certain trends, ... within some margin of confidence. I think technical discussions like this based on facts, numbers, ... helps.

Actually, without these numbers, I got the impressions from so many exciting posts about the recent 1700 DLT3C that it is the best oc cpu. That is what prompt me to do the calculation since I am shopping for which CPU to buy, 1700 or 2100 or both.

And the calculation reveals that

1. 2100 still looks better in absolute speed than the TB B 1.5V 1700 (though it is better price/performance)

2. overclocking tracks the rated frequency of CPU, i.e. CPU with higher rated freq clocks higher on the average

And this is pretty obvious after all (what you get is what you pay for).

 

[forlack]

Awesome... I totally agree with you. I mean I have heard many people rant and rave about "a guy" they saw in a form that had some outstanding overclock, when in reality that was just an anomolie. I kind of rambled on, but the database is all that we can really use. Its hard to tell what will yield the best results. Eiether way the 1700 or the 2100 will be a great chip.

 

[OC-Master]

Dont forget the 2100+ has remapped multipliers which allow you to select between 13X~24X. The XP1700+ does not got this..


Since not everyone has nForce2 motherboards .


OC-Master