Overclocking Sandbox: Tbred B DLT3C 1700+ and Beyond

[Cuda]

Have you considered the Buffalo 3700 memory modules which use the Winbond 5ns BH-5 chips? Better chance to get to 220+ MHz FSB.

I have the Buffalo 3200 Winbond CH-5 chips and I run at 220 MHz FSB with no problems at all.

 

[hitechjb1]

What happens to programs running in CPU with smaller and bigger L2 cache

That really depends on a particular program and OS, and how the code was generated by the compiler In general, the Bartron 512 KB cache has a performance advantage of 0 to 10+ % over the 256 KB 1700+/1800+, tested on a wide range of programs, assuming the CPU frequency is the same.

Look at it this way, assuming a 1700+/1800+ can be overclocked 100-120 MHz (~4-5%) faster than a Barton.

1. For small programs (code + data) whose size is small enough, say 100 KB, such that both code + data can be fit into the 256 KB (and also 512 KB) cache (the OS needs also some space in the cache), then the 1700+/1800+ would win. When there is a cache miss, i.e. code or data not found in cache, it takes many cycles to get them from memory and to fill the cache. Because running that small program, there is no need to access the main memory for both CPU, the Barton has no advantage over the 1700+/1800+. Since the 1700+/1800+ takes shorter time each cycle, so they win the Barton.

2. For programs (code + data) whose size is larger than 256 KB but smaller than 512 KB, say 300 KB, definitely the Barton will win. Since once a while, the 1700+/1800+ would find code or data missing in cache and have to get them from main memory (i.e. it would take many more cycle to complete the same program). While the Barton can keep running from the cache and without needing additional cycles to get data from the main memory.

3. For programs (code + data) whose size is larger than 512 KB, say big prgrams w/ big matrices. Then both cases, there are continuous cache misses, and continuous memory access. On the average, the Barton would access the main memory less, hence it would "waste" less cycle compared to the 1700+/1800+ which has a shorter time per cycle. Then who wins depends on

- The cycle "wasted" for the 1700+/1800+ memory to access compared to its gain in shorter cycle time.

- How the code was generated by the compiler and how persistent the code + data are in the cache. This is program and application specific and dependent. If the code + data can persist in the cache over a longer period of time, i.e. less frequent cache miss, this would help the 1700+/1800+, and so Barton would win less. But if the code + data are less persistent, and there are lots of cache miss, the 1700+/1800+ would have to get them from main memory much more often than the Barton, and under this extreme condition, the Barton would win by a high %.

So in general, for big program with huge amount of data, and are not persistent in cache, the Barton would win. By how much depends on the amount of code + data, the frequency of cache misses, and how persistent the code + data are.

So I hope we can see that it is really application specific. If the Barton can be clocked to same speed, definitely it would win. But since from many applications seen, Barton has 0 to 10+% advantage, and the 1700+/1800+ can be clocked about 4-5% faster (100-120 MHz), it is about a tie for both.

Related link:

Comparing Tbred B 1700+/1800+ DLT3C and Barton 2500+ overclocking performance, power and temperature (page 14)

 

[hitechjb1]

Does that suggest that if you put a 7mm spacer under the sf2 to decrease the sink's flow resistance to be about equal to the tornado that the sf2 will subtantially close your temperature gap between the sf2 and the tornado? At low speeds having the fan right up against the sink would seem to help create pressure which is good, but at high speeds the proximity of the blades to the sink would create flow resistance which is bad.

btw the above area calculations do not appear to take into account the increased size of the tornado fins due to the 38mm fan height, but are two dimensional only. Shouldn't the blade height figure in somehow?

flow = area (velocity)

I thought
velocity = constant (RPM)
so RPM would take care of that. The cosntant captures the blade shape or geometry for a given fan.

But you brought up a good point about the depth of the blade, or the angle of the blade, I have to think about it.


Would this do it:


velocity = (formFactor) (blade_height) (RPM)

formFactor ~ 50%, which accounts for certain % (~ 50% in this case) of the volume of air is swept out by the blades in a revolution. For exact number of the formFactor, one has to study the shape and geometry of the blade and perform measurement and/or simulation.


But if you look at both fans carefully, the blade_height are rought the same. The 38 mm of the Tornado does not mean greater blade_height. So roughly, the formFactor of both fans are the same.

So the previous CFM estimate can be extened to

flow = formFactor (pi) (OD + ID) (OD - ID) (blade_height) (RPM)

Since the Tornado 80mm and TT SFII have roughly the same blade_height, the last estimate of 11% shoule be OK.


The 80x38 mm Tornado creates a deeper duct for the air flow, but not a deeper blade. The deeper duct impacts on the pressure part, but not on the CFM, I think.

 

[Audioaficionado]

I think both the BH5 and CH5 are fine. Just many people in the memory section seems to say BH5 are better, which is a bit strange. I heard Kingston 3000 modules use BH5 and their 3500 modules use CH5 (if BH5 is so good for overclocking, why do they do it like this?).

I have some 512 MB CH5, two can do up to 237 MHz and one only goes to 220 MHz, all under 2.9 V.

So it seems to be just a sampling of luck. But getting the 3700 rated at 231 MHz would take out the guess work.

I have KHX3200A/512 and it has a sticker with Winbond on it.

Any idea which one I have?

 

[johnoh]

The 80x38 mm Tornado creates a deeper duct for the air flow, but not a deeper blade. The deeper duct impacts on the pressure part, but not on the CFM, I think.

That makes sense. And my thought is that the back-pressure caused by the proximity of the sink to the sf2 is causing it to lower its effectiveness (cfm or pressure, I don't don't know which). Even with equal blade heights the tornado blade is farther from the sink so perhaps it does not suffer from the back pressure as much. I'm just looking for an explanation to the mystery of why the tornado is a full 11C lower than the sf2, which is a very big difference.

I call it a mystery because if you raise the sf2 by 7mm and believe the open airflow cfm ratings, these two fans should have temps that are much closer together.

 

[Mustanley]

This fan discussion raises some interesting points. Tonight I plan on experimenting with fan spacers to see if they can have a noticable impact on cooling performance. I'll use a gutted 80mm (25mm thick) fan housing as a spacer below an NMB 80mm fan. As soon as I find an old 92mm fan to gut, I'll test that with a 92mm NMB fan as well. This will be done on top of a SLK947u.

 

[hitechjb1]

This fan discussion raises some interesting points. Tonight I plan on experimenting with fan spacers to see if they can have a noticable impact on cooling performance. I'll use a gutted 80mm (25mm thick) fan housing as a spacer below an NMB 80mm fan. As soon as I find an old 92mm fan to gut, I'll test that with a 92mm NMB fan as well. This will be done on top of a SLK947u.
...

Very good ideas, looking forwards to see the results.

Maybe put some temporary taping between the casing of the two fans to minimize air leakage. If things work, then permanent glue can be used to shield the gaps.

 

[hitechjb1]

Another idea, if one can find
- a 92 mm fan with higher CFM
- a smaller 80 mm casing
- some way to connect the 92 mm fan tapering down to the 80 mm casing (with minimal air leak between them), then to the heat sink (such as SLK-800U/900U/947U).

Then one would have the advantage of high CFM (from the 92 mm) and directing most of the air mainly onto CPU area of the heat sink. This may reverse the low CFM efficiency seen of putting a 92 mm fan directly on the heat sink as reported in previous posts ("92 mm fan performs worse than 80 mm fan").

 

[Mustanley]

Another idea, if one can find
- a 92 mm fan with higher CFM
- a smaller 80 mm casing
- some way to connect the 92 mm fan tapering down to the 80 mm casing (with minimal air leak between them), then to the heat sink (such as SLK-800U/900U/947U).

Then one would have the advantage of high CFM (from the 92 mm) and directing most of the air mainly onto CPU area of the heat sink. This may reverse the low CFM efficiency seen of putting a 92 mm fan directly on the heat sink as reported in previous posts ("92 mm fan performs worse than 80 mm fan").

I can try this too. However, I can't think of an easy method to create a tapered adapter from 92mm->80mm, so I'll probably end up connecting the 80mm fan housing directly to the 92mm fan, and then seal the gaps with tape.
I'll also have to deal with getting the fan clips to secure these different sized monstrosities to the heatsink.

 

[johnoh]

I'm just curious hitechjb1, do you have any seat-of-the-pants findings to compare the sf2 and the tornado in open air? Like do they feel similar in airflow to you out in the open?

 

[hitechjb1]

I'm just curious hitechjb1, do you have any seat-of-the-pants findings to compare the sf2 and the tornado in open air? Like do they feel similar in airflow to you out in the open?

I only have these shown earlier using estimation based on
- fan inner diameter
- fan outer diameter
- blade height
- RPM (spec)
- fan shape factor (assume both are same)
using this equation

flow = formFactor (pi) (OD + ID) (OD - ID) (blade_height) (RPM)

As shown earlier, it seems to agree that the Tornado 80 mm is 11% higher in CFM than the TT SF II in open air, agreeing with the spec numbers of 84 vs 75.7 CFM.

I think the longer duct of the Tornado 80 mm would change the load characteristic when putting over a heat sink, hence a better effective CFM when used w/ a HS, compared to the TT SFII, as seen from the load temperature difference of 8-11 C on a Tbred B 1700+ DLT3C running at 2.57 GHz at 1.95 V, around 15-18 C ambient.

 

[hitechjb1]

I have KHX3200A/512 and it has a sticker with Winbond on it.

Any idea which one I have?

Not 100% sure, I heard the Kingston HyperX KHX3200A is BH5, same for its Hyper-X 3000. But the HyperX 3500 is CH5. I don't know why. But then it shows BH5 is not better than CH5 in nature. I don't know much more about their difference in overclocking, except a luck of draw or getting the 3700 which is officially rated higher, at 231 MHz, against the "not so good" one in 3000, 3200 and 3500.

 

[johnoh]

I don't know hitech, I can't yet seem to rationalize 11C difference. Surely giving sf2 another 7mm of duct won't reduce its temps by 11C???

Maybe something about the blade shape causes the sf2 to suffer at high rpm? I wonder if with a very slow fan a large blade surface area helps, but with high rpms that larger surface ara creates flow resistance due to excess turbulance? Or impeding air intake from behind? This could be the case given that the tornado was designed only for very high speeds, but the sf2 was meant to cover a wide range of speeds.

 

[Audioaficionado]

Not 100% sure, I heard the Kingston HyperX KHX3200A is BH5, same for its Hyper-X 3000. But the HyperX 3500 is CH5. I don't know why. But then it shows BH5 is not better than CH5 in nature. I don't know much more about their difference in overclocking, except a luck of draw or getting the 3700 which is officially rated higher, at 231 MHz, against the "not so good" one in 3000, 3200 and 3500.

I'm thinking it's six of one or a half dozen of the other.

As I lookd at the specs I noticed the faster rated RAM had to relax its CAS and other timings to meet rated speed. Heck you could also just relax the slower rated stuff to match timings and reach the same speed. So it might be a numbers game to some extent. That's why I got PC3200 as it had a good price/performance rating and I hope will let me get past 230. I'll see soon enough when I try to OC next week.

 

[hitechjb1]

I don't know hitech, I can't yet seem to rationalize 11C difference. Surely giving sf2 another 7mm of duct won't reduce its temps by 11C???

Maybe something about the blade shape causes the sf2 to suffer at high rpm? I wonder if with a very slow fan a large blade surface area helps, but with high rpms that larger surface ara creates flow resistance due to excess turbulance? Or impeding air intake from behind? This could be the case given that the tornado was designed only for very high speeds, but the sf2 was meant to cover a wide range of speeds.

Once the air leaves the fan, through some short air duct (in the case of Tornado 80 mm or the TT SFII mod with some air duct mod), into the fins of the heat sink, all the CPU sees is flow of air shaped by the fins with certain flow rate (CFM). So the air flow (rate) is the determining factor in relationship to the CPU, since the boundary conditions between CPU and the heat sink are the same.

For two fans with the same or about the same CFM in open air, the differentiating factor between them would be the interface between the fan per se and the top of the heat sink.

As you pointed out, the Tornado fan may be shaped optimally at top speed, and the TT SFII is for a wide range of rpm, I don't know how different that would impact the flow resistance in determining the actual, effective CFM after fan and heat sink are put together. From the blade size, the two fans look about the same. And CFM is estimated or specified as only 11% different.

So the main different between the two fans is the duct interface between the fan and the top of heat sink. So the interesting thing is to see how adding some kind of duct to the TT SFII would help.

Mustanley is doing some testing using NMB fans, will see how it turns out. One could try similar things on a TT SFII to see whether it helps to reduce the CPU load temperature, from an extreme of 8-11 C difference compared to the Tornado 80 mm (whose CFM is only 11% more), to something lower.

 

[Mustanley]

The two fans I used in this test were made by NMB, an 80mmx25mm 25CFM and a 92mmx25mm 48CFM model.
Ambient temp was 27C during testing.
testbed: 2500+@2310(220x10.5) 1.75V, NF7-S, slk947u, Ceramique TIM.

92mm fan-> cpu:49C, board:29C
92mm w/80mm fan housing reducer-> cpu:51C, board:31C
80mm w/80mm fan housing spacer-> cpu:55C, board:35C
80mm fan-> cpu:52C, board:35C

I wasn't very impressed with my 92mm->80mm reducer, since it was not tapered. This likely more than offset any enhancement to the pressure created by the 80x25mm spacer. My next project is to fabricate a few spacers and tapered reducers of various sizes to test. I also want to use some higher performance fans. The 80mm NMB is very weak and not adequately equipped to push air through the slk947u.
Another thing that struck me was the 92mm fan overhang lowering the board temperature. Board temps rose when the air column was reduced to only 80mm, either via the 80mm fan or the 92mm->80mm reducer.

 

[hitechjb1]

From the test, since 80 mm was not tested, it did not compare 80 mm and 92 mm directly. You may also need higher CFM fans and better seals for air leakage .

Regarding to overhang and cooling CPU surrounding, from the test results of the 4 fans:

Originally posted by hitechjb1
The following four poplular fans were tested on overclocking
- a Tbred B 1700+ DLT3C WPMW 0310
- to 2.57 GHz (223 x 11.5) @ 1.95 V (Prime95 stable)
- NF7-S rev 2.0 motherboard (Vdd 1.7 V)
- 2x512 MB 3500 memory @ 6-3-3-2 (Vdimm 2.9 V)
- SLK-947U
...

fan.................................CPU frequency…......load temp (CPU/sys)……..fan speed reported
Vantec 92 mm Tornado.......2.57 GHz.….............45 / 16 C………....…....………4821 rpm
Vantec 80 mm Tornado………2.57 GHz.….............37-40 / 15 C…........…………5444 rpm
TT SF II 80 mm…........……….2.57 GHz.….............48 / 18 C…………….......…....4561 rpm
Enermax 92 mm……….....……2.34 GHz (1.875V)...48.5 / 24 C…....................2377 rpm

It shows that when the CFM is high enough (Tornado 80/92mm or TT SFII), even it is a 80 mm fan, the surrounding area around the CPU (such as voltage regulator and north bridge) can be cooled and benefited from the fan, especially when the heat sink base is similar to that of the SLK-800/900/947 whose copper base is smaller so air can be passed out from the area not covered by the base. I have been using high CFM CPU fans to cool the surrounding area.

When the rpm (or CFM) is lower such as the Enermax 92 mm, even it is 92 mm, the surrounding of the CPU is not cooled effectively as the other three, unless extra fans are directed into that area.



Edit:
1. Your test showed 92 mm NMB fan performs better than 80 mm NMB fan by 3 C CPU temperature.

2. My test showed 92 mm Tornado fan performs worse than 80 mm Tornado fan by 5-8 C with same CPU, voltage and frequency.

Inconclusive, which is better between 80 and 92 mm fans ???

 

[Mustanley]

I edited the previous post with info for the single 80mm fan.

You may also need higher CFM fans and better seals for air leakage

Both 80mm adapters/spacers I used were completely sealed, one being screwed together and sealed with tape, and the other bound with plastic packaging stretch wrap (and taped).

Still want to try some higher cfm fans and better designed adapters. Like these from www.1coolpc.com ...

 

[hitechjb1]

Comparing Tornado 80mm, Tornado 92mm and TT SFII (second test data)

*Update: add Tornado 92 mm fan

Another data point for comparing a Vantec Tornado 80mm / 92mm and a TT SFII, the tests were done under similar ambinent conditions (within 1 hour):

CPU at 2.57 GHz (223 x 11.5), 1.90 V
- Tbred B 1700+ DLT3C JIXIB 0332 MPMW
- NF7-S rev 2.0 motherboard (Vdd 1.7 V)
- 2x512 MB 3500 memory @ 6-3-3-2 (Vdimm 2.9 V)
- SLK-947U

Light load temperature 15 / 35 C (both cases).

Tornado 80 mm........2.57 GHz........40.5 / 17 C........5625 rpm........84 CFM (spec)
TT SFII...................2.57 GHz........50 / 23 C..........4688 rpm........75.7 CFM (spec)
Tornado 92 mm........2.57 GHz........44.5 / 16 C........4821 rpm........119 CFM (spec)

There is a difference of 9.5 C (+- 1 C) between the two fans. We also see the secondary effect (a difference of 6 C) of the Tornado 80 mm on temperature around the CPU where a system ambient sensor is located.

In a previous test (refer to earlier post), between Tornado 80 mm and TT SFII, a difference of 8-11 C was found using another CPU (1700+ JIUHB WPMW 0310 at 2.57 GHz 1.95 V) on the same motherboard.

This test confirms the previous measurment between the Tornado 80 mm and TT SFII, i.e. 9.5 C +- 1C temperature difference for a 11% difference in CFM (fan CFM specified on open air).

What is interesting is that the Tornado 80 mm (84 CFM) performs better than the Tornado 92 mm (119 CFM) by 4 C. This confirms the same result as in the first test (5 C difference). The open air CFM of the 92 mm is 42% higher than the 80 mm. Due to the larger fan size and smaller heat sink interface boundary, the blow is not fully concentrating on the heat sink central area. The 92 mm fan static pressure created in combination with the heat sink (flow resistance) is not as high as the 80 mm fan, resulting in smaller effective CFM over the CPU copper base area.

The larger overhang of the 92 mm cools the CPU surrounding area a bit better, but is not significant (little or by at most 1 C system ambient), provided that a 80 mm fan with enough CFM such as TT SFII or better is used.

This second test confirms the first test conclusion: a Vantec Tornado 80 mm performs better than a Tornado 92 mm fan, by as much as 4-5 C on CPU temperature at same voltage and frequency.

... Screen shots follow ...

 

[hitechjb1]

Vantec Tornado 80 mm