I have read all 29 pages and am confused about which Amd64 processor is the best for overclocking. I am just putting together my system and have purchased the Msi Neo2 Platinum MB and 2-512mg of pc3200 pny memory. Now I'm ready to get the processor. I want to be able to overclock so I'm looking at 3000+, 3200+, 3400+, 3500+. 3700+. And then we get to the core issue. Which is better for gaming and overclocking - Venice, Clawhammer, Winchester or Newcasle? My main question is "Which is the best Speed and core" for what I am looking to do? I see alot of people are using the 3000+ Venice but with the price not being much different than the 3200+, 3400+ and 3700+ I'm confused as to which I should I go with. Does one speed and/or core overclock better than another? I would appreciate any feedback I get. Thanks in advance.
Help 'un-confuse' a newbie
- Thread starter Stonewall
- Start date
3000+ Venice
3700+ San Diego
Basically the best two options for overclocking and gaming. Generally the lower is the better it overclocks.
3700+ San Diego
Basically the best two options for overclocking and gaming. Generally the lower is the better it overclocks.
Overall I'd say the San diegos seem to overclock better than venices. People normally get 3000+ venices because they are dirt cheap (ok, maybe not dirt cheap but cheap compared to other A64s) and overclock pretty well. Clawhammer and newcastle are old 130nm cores, they perform worse than venice and san diegos clock for clock and seem to overclock less while using a lot more power. Winchester is AMD's 90nm test cpu and it isnt that great of an overclocker.
I'd personally go for a 3000+ 'cause it's so much more fun to get all you can out of a 140 dollar cpu and the max overclock potential is only a little worse than most 3700+. The 3700+ does perform a bit better in games due to its 1MB cache and I've only heard of one dud 3700+ which only reached 2.6GHz (that doesnt mean you are assured 2.6GHz, though) while there are plenty of dud venices (which hit like 2.55 tops).
Of course, take my clock speed references as just that, references. Generally, the higher the default clock, the greater the chance of getting a high overclocker since the cpu is validated to work at higher clock speeds at stock volts.
I'd personally go for a 3000+ 'cause it's so much more fun to get all you can out of a 140 dollar cpu and the max overclock potential is only a little worse than most 3700+. The 3700+ does perform a bit better in games due to its 1MB cache and I've only heard of one dud 3700+ which only reached 2.6GHz (that doesnt mean you are assured 2.6GHz, though) while there are plenty of dud venices (which hit like 2.55 tops).
Of course, take my clock speed references as just that, references. Generally, the higher the default clock, the greater the chance of getting a high overclocker since the cpu is validated to work at higher clock speeds at stock volts.
Thanks for the quick replies! One more question then. Am I better off -speed wise- to get a 3000+ venice and overclock it than to get a 3200+, 3400+, 3500+, etc. and NOT overclock it and keep it stock? Hope this makes since to anyone willing to reply. Thanks again.
The 3000+ Venice at 3200+/3500+ speeds may be a bit faster, due to the higher memory speeds and higher FSB.
Basically, as far as overclocking cores go, San Diego and Venice > Winchester > Clawhammer and Newcastle.
The 3000+ Venice and 3700+ San Diego are the best (and most popular) OCing options.
Basically, as far as overclocking cores go, San Diego and Venice > Winchester > Clawhammer and Newcastle.
The 3000+ Venice and 3700+ San Diego are the best (and most popular) OCing options.
Okay I've got this now but (sorry I lied, just one more thing), this 3000+ Venice should be Revision 3, correct? How do I know I am getting this particular chip? Again a big thanks for all the help.
I would say most of the 3000+ cpu's were purchased in May and June, since they were first available in late April and the 1MB L2 cache models were not yet available. I purchased my 3000+ because of the low cost, it was my first overclock attempt, it really is fast enough for my needs, and I can easily purchase a more expensive dual core next year after prices will hopefully decrease.
I am not familiar with a Revision 3 AMD cpu since my 3000+ Venice is a Revision E. Is this a typo or am I missing something? It should not matter since the Venice and San Diego cpu's are clearly labeled as such by NewEgg and should be clearly labeled by other vendors also.
When looking at a set of similar processors with different multipliers it is important to realize that AMD tests/inspects the cpu's and sets the higher multipliers for the better quality cpu's. If there is a larger market for 3500+ than for 3000+, then there will be few high quality 3000+ cpu's released. If the market is larger for the 3000+ than for the 3500+, then many high quality cpu's will be locked at the lower multiplier for the 3000+ and shipped to meet demand.
Multipliers:
9x for 3000+ (512 KB L2 cache) (current price $146)
10x for 3200+ (512 KB L2 cache) (current price $190)
11x for 3500+ (512 KB L2 cache) (current price $220)
11x for 3700+ (1024 KB L2 cache) (current price $272)
12x for 4000+ (1024 KB L2 cache) (current price $368)
From my experiences overclocking my system, I found that my lowly 9x multiplier required a higher FSB speed to obtain the same total 2600 MHz speed others were obtaining with their 3500+ cpu and its 11x multiplier. For example, to reach 2600MHz, my 3000+ (9x) required an FSB of 290 x9 = 2610MHz, whereas the 3500+ only needs an FSB of 238 x11 = 2618MHz. The default FSB setting is 200MHz and it can theoretically go up to 400MHz. Although, most systems are only able to get up to the low 300MHz range. My Asus A8N-E motherboard was able to get up to 320MHz but it was not stable at that setting.
I assume this lower FSB frequency translates into decreased wear on the motherboard. Another advantage of the decreased FSB speed afforded by the higher multiple with the 3500+ is that the RAM does not require a divider to keep up with the high speed of the FSB.
Memory (RAM) Considerations
RAM is rated for a specific speed:
DDR 400 (PC3200) at 400/2 = 200MHz
DDR 433 (PC3500) at 433/2 = 216MHz
DDR 466 (PC3700) at 466/2 = 233MHz
DDR 500 (PC4000) at 500/2 = 250MHz
DDR 533 (PC4200) at 533/2 = 266MHz
DDR 550 (PC4400) at 550/2 = 275MHz
DDR 566 (PC4500) at 566/2 = 283MHz
DDR 600 (PC4800) at 600/2 = 300MHz
Many DDR 400 RAM sticks are able to reach about 225MHz and are therefore able to communicate with a cpu through the FSB at 225MHz at a 1:1 ratio. Remember the RAM must be able to communicate with the CPU through the FSB (Front Side Bus).
This (1:1) ratio allows:
3000+: 225 x9 = 2025MHz
3200+: 225 x10 = 2250MHz
3500+: 225 x11 = 2475MHz
3700+: 225 x11 = 2475MHz
4000+: 225 x12 = 2700MHz
From this description, it can be seen that if higher overclocks are desired, it will be necessary to either:
1) purchase a cpu with a higher multiple,
2) purchase a set of RAM rated higher, or
3) set a memory divider for the RAM.
It should also be understood that it is usually the RAM that limits the overclock, even though the cpu and the motherboard's FSB can also be the limiting factor.
One of the architectural features of the AMD64 is the on-die memory controller. This allows for faster communication and control of the calls to memory at the speed of the cpu, without the inherent increased latency with a memory controller that is not on the cpu die.
The on-die memory controller provides increased efficiency for using a memory divider, which provides for higher FSB speeds than the memory can keep up with and therefore allowing a higher MHz speed overclock to the processor. A memory divider can be set at:
(RAM:FSB) MHz
(1:1) 400:400 = 6 RAM cycles for every 6 CPU cycles
(5:6) 333:400 = 5 RAM cycles for every 6 CPU cycles
(4:6) 267:400 = 4 RAM cycles for every 6 CPU cycles
(3:6) 200:400 = 3 RAM cycles for every 6 CPU cycles
The memory divider then allows the following higher system overclock speeds:
(5:6) provides 225 x(6/5) = 270MHz FSB or a range of 200-270MHz
3000+: 270 x9 = 2430MHz
3200+: 270 x10 = 2700MHz
3500+: 270 x11 = 2970MHz
3700+: 270 x11 = 2970MHz
4000+: 270 x12 = 3240MHz
(4:6) provides 225 x(6/4) = 337MHz FSB or a range of 200-337MHz
3000+: 337 x9 = 3033MHz
3200+: 337 x10 = 3370MHz
3500+: 337 x11 = 3707MHz
3700+: 337 x11 = 3707MHz
4000+: 337 x12 = 4044MHz
Of course, few motherboards are able to reach an FSB speed of 337MHz and few of these processors are able to overclock faster than 2800MHz but the math is understood.
This describes the relationship between the Front Side Bus (FSB), the CPU, and the memory (RAM). This should also help provide some insight into the individual choice between the different cpu models available.
Purchasing Consideration
It is perfectly acceptable to purchase a 3000+ for $150 and overclock the system to 2400MHz with a (5:6) divider or 2700MHz with a (4:6) divider, if you can get the CPU's speed up that high.
If we consider the known fact that a strongly overclocked AMD64, socket 939, cpu typically reaches a maximum speed of 2700-2800MHz, then we see this speed can be reached with the 4000+ at a (1:1) ratio or by a much cheaper 3200+ at a (5:6) ratio, with typical DDR400 RAM. The (5:6) ratio performs at nearly the same performance as the (1:1) ratio due to the on-die memory controller and the simple fact that not every cpu cycle accesses memory.
RAM Memory Upgrade Consideration
Lets see what can be obtained by purchasing a set of memory rated at the higher speed of DDR500 (250MHz). For 2x512MB, this RAM currently has a mid-range price of $150, $160, or $180 with a low at $145 and a high at $230.
(1:1) ratio provides 250MHz x(1/1) = 250MHz FSB or a range of 200-250MHz
(5:6) ratio provides 250MHz x(6/5) = 300MHz FSB or a range of 200-300MHz
(4:6) ratio provides 250MHz x(6/4) = 375MHz FSB or a range of 200-375MHz
It can be seen that the (4:6) ratio is not useful since few motherboards can support an FSB above 300MHz.
(1:1) or (6:6)
3000+: 250 x9 = 2250MHz
3200+: 250 x10 = 2500MHz
3500+: 250 x11 = 2750MHz
3700+: 250 x11 = 2750MHz
4000+: 250 x12 = 3000MHz
(5:6)
3000+: 300 x9 = 2700MHz
3200+: 300 x10 = 3000MHz
3500+: 300 x11 = 3300MHz
3700+: 300 x11 = 3300MHz
4000+: 300 x12 = 3600MHz
The DDR500 memory can support:
1) (1:1) ratio with a 3200+ at 2500MHz,
2) (1:1) ratio with a 3500+ at 2750MHz, or
3) (5:6) ratio with a 3000+ at 2700MHz.
It can be seen that the more expensive DDR500 memory allows the 3200+ to switch from the (5:6) ratio to the (1:1) ratio. The performance increase with the higher memory ratio is minimal as shown by Zebo in his Official 939 Memory Matrix and as I have observed through my own benchmarking when making the decision to run with a (5:6) memory divider ratio at 2400MHz or with a (4:6) memory divider ratio at 2600MHz. (I chose the 2400MHz with the (5:6) since the performance was not much greater with the (4:6) memory divider ratio and I hope for less wear and tear with the lower FSB speed at 268MHz.)
Other considerations
One further consideration includes the larger 1024KB (1MB) L2 cache in the 3700+ and the 4000+.
Another consideration includes the different memory (RAM) latency ratings, such as CAS 2, 2.5, or 3.
Considering current prices:
DDR400 RAM, 2x512MB ($90-$250, mid-range $130)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
1) 3000+ ($146), 9x, (5:6), 1800MHz, 2400MHz, $276,
2) 3000+ ($146), 9x, (4:6), 1800MHz, 2700MHz, $276,
3) 3200+ ($190), 10x, (5:6), 2000MHz, 2700MHz, $310, ***(Kensai's choice)***
4) 3700+ ($272), 11x, (1:1), 2200MHz, 2475MHz, $402, (RickUK's choice)
5) 3700+ ($272), 11x, (5:6), 2200MHz, 2970MHz, $402, (RickUK's choice)
6) 4000+ ($368), 12x, (1:1), 2400MHz, 2700MHz, $500.
DDR500 RAM, 2x512MB ($160, $180)
(There is no need to purchase memory faster than DDR400 if you will not overclock)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
1) 3000+ ($146), 9x, (5:6), 1800MHz, 2700MHz, $326,
2) 3500+ ($220), 11x, (1:1), 2200MHz, 2750MHz, $400,
3) 3700+ ($272), 11x, (1:1), 2200MHz, 2750MHz, $452. (RickUK's choice)++
Edit: X2 Dual Core addition
DDR400 RAM, 2x512MB ($90-$250, mid-range $130)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
0) 3800+ X2 ($374), 10x, (1:1), 2000MHz, 2250MHz, $504, (Bona Fide's choice)
1) 3800+ X2 ($374), 10x, (5:6), 2000MHz, 2700MHz, $504, (Bona Fide's choice)
2) 4000+ X2 ($???), 10x, (1:1), 2000MHz, 2250MHz, $???,
3) 4000+ X2 ($???), 10x, (5:6), 2000MHz, 2700MHz, $???, (1MB L2 cache) (no practical sense in going faster)
4) 4200+ X2 ($473), 11x, (1:1), 2200MHz, 2475MHz, $603,
5) 4200+ X2 ($473), 11x, (5:6), 2200MHz, 2970MHz, $603,
6) 4400+ X2 ($545), 11x, (1:1), 2200MHz, 2475MHz, $675,
7) 4400+ X2 ($545), 11x, (5:6), 2200MHz, 2970MHz, $675,
8) 4600+ X2 ($690), 12x, (1:1), 2400MHz, 2700MHz, $820,
9) 4600+ X2 ($690), 12x, (5:6), 2400MHz, 3240MHz, $820. (few people have achieved speeds > 2800MHz)
DDR500 RAM, 2x512MB ($160, $180)
(There is no need to purchase memory faster than DDR400 if you will not overclock)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
0) 3800+ X2 ($374), 10x, (1:1), 2000MHz, 2500MHz, $554,
1) 3800+ X2 ($374), 10x, (5:6), 2000MHz, 3000MHz, $554,
2) 4000+ X2 ($???), 10x, (1:1), 2000MHz, 2500MHz, $???,
3) 4000+ X2 ($???), 10x, (5:6), 2000MHz, 3000MHz, $??, (no practical sense in going faster)
4) 4200+ X2 ($473), 11x, (1:1), 2200MHz, 2750MHz, $653,
5) 4200+ X2 ($473), 11x, (5:6), 2200MHz, 3300MHz, $653, (impossible to reach)
6) 4400+ X2 ($545), 11x, (1:1), 2200MHz, 2750MHz, $725,
7) 4400+ X2 ($545), 11x, (5:6), 2200MHz, 3300MHz, $725, (impossible to reach)
8) 4600+ X2 ($690), 12x, (1:1), 2400MHz, 3000MHz, $870,
9) 4600+ X2 ($690), 12x, (5:6), 2400MHz, 3600MHz, $870. (impossible to reach)
It is doubtful that memory prices will decrease significantly while it can be expected for cpu prices to decrease. It may be worthwile to purchase DDR500 RAM now with a lower range processor and then upgrade that processor in a year or two. But then I usually just purchase a whole new system and give my system to a family member when it comes time to upgrade.
I am not familiar with a Revision 3 AMD cpu since my 3000+ Venice is a Revision E. Is this a typo or am I missing something? It should not matter since the Venice and San Diego cpu's are clearly labeled as such by NewEgg and should be clearly labeled by other vendors also.
When looking at a set of similar processors with different multipliers it is important to realize that AMD tests/inspects the cpu's and sets the higher multipliers for the better quality cpu's. If there is a larger market for 3500+ than for 3000+, then there will be few high quality 3000+ cpu's released. If the market is larger for the 3000+ than for the 3500+, then many high quality cpu's will be locked at the lower multiplier for the 3000+ and shipped to meet demand.
Multipliers:
9x for 3000+ (512 KB L2 cache) (current price $146)
10x for 3200+ (512 KB L2 cache) (current price $190)
11x for 3500+ (512 KB L2 cache) (current price $220)
11x for 3700+ (1024 KB L2 cache) (current price $272)
12x for 4000+ (1024 KB L2 cache) (current price $368)
From my experiences overclocking my system, I found that my lowly 9x multiplier required a higher FSB speed to obtain the same total 2600 MHz speed others were obtaining with their 3500+ cpu and its 11x multiplier. For example, to reach 2600MHz, my 3000+ (9x) required an FSB of 290 x9 = 2610MHz, whereas the 3500+ only needs an FSB of 238 x11 = 2618MHz. The default FSB setting is 200MHz and it can theoretically go up to 400MHz. Although, most systems are only able to get up to the low 300MHz range. My Asus A8N-E motherboard was able to get up to 320MHz but it was not stable at that setting.
I assume this lower FSB frequency translates into decreased wear on the motherboard. Another advantage of the decreased FSB speed afforded by the higher multiple with the 3500+ is that the RAM does not require a divider to keep up with the high speed of the FSB.
Memory (RAM) Considerations
RAM is rated for a specific speed:
DDR 400 (PC3200) at 400/2 = 200MHz
DDR 433 (PC3500) at 433/2 = 216MHz
DDR 466 (PC3700) at 466/2 = 233MHz
DDR 500 (PC4000) at 500/2 = 250MHz
DDR 533 (PC4200) at 533/2 = 266MHz
DDR 550 (PC4400) at 550/2 = 275MHz
DDR 566 (PC4500) at 566/2 = 283MHz
DDR 600 (PC4800) at 600/2 = 300MHz
Many DDR 400 RAM sticks are able to reach about 225MHz and are therefore able to communicate with a cpu through the FSB at 225MHz at a 1:1 ratio. Remember the RAM must be able to communicate with the CPU through the FSB (Front Side Bus).
This (1:1) ratio allows:
3000+: 225 x9 = 2025MHz
3200+: 225 x10 = 2250MHz
3500+: 225 x11 = 2475MHz
3700+: 225 x11 = 2475MHz
4000+: 225 x12 = 2700MHz
From this description, it can be seen that if higher overclocks are desired, it will be necessary to either:
1) purchase a cpu with a higher multiple,
2) purchase a set of RAM rated higher, or
3) set a memory divider for the RAM.
It should also be understood that it is usually the RAM that limits the overclock, even though the cpu and the motherboard's FSB can also be the limiting factor.
One of the architectural features of the AMD64 is the on-die memory controller. This allows for faster communication and control of the calls to memory at the speed of the cpu, without the inherent increased latency with a memory controller that is not on the cpu die.
The on-die memory controller provides increased efficiency for using a memory divider, which provides for higher FSB speeds than the memory can keep up with and therefore allowing a higher MHz speed overclock to the processor. A memory divider can be set at:
(RAM:FSB) MHz
(1:1) 400:400 = 6 RAM cycles for every 6 CPU cycles
(5:6) 333:400 = 5 RAM cycles for every 6 CPU cycles
(4:6) 267:400 = 4 RAM cycles for every 6 CPU cycles
(3:6) 200:400 = 3 RAM cycles for every 6 CPU cycles
The memory divider then allows the following higher system overclock speeds:
(5:6) provides 225 x(6/5) = 270MHz FSB or a range of 200-270MHz
3000+: 270 x9 = 2430MHz
3200+: 270 x10 = 2700MHz
3500+: 270 x11 = 2970MHz
3700+: 270 x11 = 2970MHz
4000+: 270 x12 = 3240MHz
(4:6) provides 225 x(6/4) = 337MHz FSB or a range of 200-337MHz
3000+: 337 x9 = 3033MHz
3200+: 337 x10 = 3370MHz
3500+: 337 x11 = 3707MHz
3700+: 337 x11 = 3707MHz
4000+: 337 x12 = 4044MHz
Of course, few motherboards are able to reach an FSB speed of 337MHz and few of these processors are able to overclock faster than 2800MHz but the math is understood.
This describes the relationship between the Front Side Bus (FSB), the CPU, and the memory (RAM). This should also help provide some insight into the individual choice between the different cpu models available.
Purchasing Consideration
It is perfectly acceptable to purchase a 3000+ for $150 and overclock the system to 2400MHz with a (5:6) divider or 2700MHz with a (4:6) divider, if you can get the CPU's speed up that high.
If we consider the known fact that a strongly overclocked AMD64, socket 939, cpu typically reaches a maximum speed of 2700-2800MHz, then we see this speed can be reached with the 4000+ at a (1:1) ratio or by a much cheaper 3200+ at a (5:6) ratio, with typical DDR400 RAM. The (5:6) ratio performs at nearly the same performance as the (1:1) ratio due to the on-die memory controller and the simple fact that not every cpu cycle accesses memory.
RAM Memory Upgrade Consideration
Lets see what can be obtained by purchasing a set of memory rated at the higher speed of DDR500 (250MHz). For 2x512MB, this RAM currently has a mid-range price of $150, $160, or $180 with a low at $145 and a high at $230.
(1:1) ratio provides 250MHz x(1/1) = 250MHz FSB or a range of 200-250MHz
(5:6) ratio provides 250MHz x(6/5) = 300MHz FSB or a range of 200-300MHz
(4:6) ratio provides 250MHz x(6/4) = 375MHz FSB or a range of 200-375MHz
It can be seen that the (4:6) ratio is not useful since few motherboards can support an FSB above 300MHz.
(1:1) or (6:6)
3000+: 250 x9 = 2250MHz
3200+: 250 x10 = 2500MHz
3500+: 250 x11 = 2750MHz
3700+: 250 x11 = 2750MHz
4000+: 250 x12 = 3000MHz
(5:6)
3000+: 300 x9 = 2700MHz
3200+: 300 x10 = 3000MHz
3500+: 300 x11 = 3300MHz
3700+: 300 x11 = 3300MHz
4000+: 300 x12 = 3600MHz
The DDR500 memory can support:
1) (1:1) ratio with a 3200+ at 2500MHz,
2) (1:1) ratio with a 3500+ at 2750MHz, or
3) (5:6) ratio with a 3000+ at 2700MHz.
It can be seen that the more expensive DDR500 memory allows the 3200+ to switch from the (5:6) ratio to the (1:1) ratio. The performance increase with the higher memory ratio is minimal as shown by Zebo in his Official 939 Memory Matrix and as I have observed through my own benchmarking when making the decision to run with a (5:6) memory divider ratio at 2400MHz or with a (4:6) memory divider ratio at 2600MHz. (I chose the 2400MHz with the (5:6) since the performance was not much greater with the (4:6) memory divider ratio and I hope for less wear and tear with the lower FSB speed at 268MHz.)
Other considerations
One further consideration includes the larger 1024KB (1MB) L2 cache in the 3700+ and the 4000+.
Another consideration includes the different memory (RAM) latency ratings, such as CAS 2, 2.5, or 3.
Considering current prices:
DDR400 RAM, 2x512MB ($90-$250, mid-range $130)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
1) 3000+ ($146), 9x, (5:6), 1800MHz, 2400MHz, $276,
2) 3000+ ($146), 9x, (4:6), 1800MHz, 2700MHz, $276,
3) 3200+ ($190), 10x, (5:6), 2000MHz, 2700MHz, $310, ***(Kensai's choice)***
4) 3700+ ($272), 11x, (1:1), 2200MHz, 2475MHz, $402, (RickUK's choice)
5) 3700+ ($272), 11x, (5:6), 2200MHz, 2970MHz, $402, (RickUK's choice)
6) 4000+ ($368), 12x, (1:1), 2400MHz, 2700MHz, $500.
DDR500 RAM, 2x512MB ($160, $180)
(There is no need to purchase memory faster than DDR400 if you will not overclock)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
1) 3000+ ($146), 9x, (5:6), 1800MHz, 2700MHz, $326,
2) 3500+ ($220), 11x, (1:1), 2200MHz, 2750MHz, $400,
3) 3700+ ($272), 11x, (1:1), 2200MHz, 2750MHz, $452. (RickUK's choice)++
Edit: X2 Dual Core addition
DDR400 RAM, 2x512MB ($90-$250, mid-range $130)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
0) 3800+ X2 ($374), 10x, (1:1), 2000MHz, 2250MHz, $504, (Bona Fide's choice)
1) 3800+ X2 ($374), 10x, (5:6), 2000MHz, 2700MHz, $504, (Bona Fide's choice)
2) 4000+ X2 ($???), 10x, (1:1), 2000MHz, 2250MHz, $???,
3) 4000+ X2 ($???), 10x, (5:6), 2000MHz, 2700MHz, $???, (1MB L2 cache) (no practical sense in going faster)
4) 4200+ X2 ($473), 11x, (1:1), 2200MHz, 2475MHz, $603,
5) 4200+ X2 ($473), 11x, (5:6), 2200MHz, 2970MHz, $603,
6) 4400+ X2 ($545), 11x, (1:1), 2200MHz, 2475MHz, $675,
7) 4400+ X2 ($545), 11x, (5:6), 2200MHz, 2970MHz, $675,
8) 4600+ X2 ($690), 12x, (1:1), 2400MHz, 2700MHz, $820,
9) 4600+ X2 ($690), 12x, (5:6), 2400MHz, 3240MHz, $820. (few people have achieved speeds > 2800MHz)
DDR500 RAM, 2x512MB ($160, $180)
(There is no need to purchase memory faster than DDR400 if you will not overclock)
Model# ($price), Multiplier, memory divider ratio, standard clock speed, max overclocked speed, combined price
0) 3800+ X2 ($374), 10x, (1:1), 2000MHz, 2500MHz, $554,
1) 3800+ X2 ($374), 10x, (5:6), 2000MHz, 3000MHz, $554,
2) 4000+ X2 ($???), 10x, (1:1), 2000MHz, 2500MHz, $???,
3) 4000+ X2 ($???), 10x, (5:6), 2000MHz, 3000MHz, $??, (no practical sense in going faster)
4) 4200+ X2 ($473), 11x, (1:1), 2200MHz, 2750MHz, $653,
5) 4200+ X2 ($473), 11x, (5:6), 2200MHz, 3300MHz, $653, (impossible to reach)
6) 4400+ X2 ($545), 11x, (1:1), 2200MHz, 2750MHz, $725,
7) 4400+ X2 ($545), 11x, (5:6), 2200MHz, 3300MHz, $725, (impossible to reach)
8) 4600+ X2 ($690), 12x, (1:1), 2400MHz, 3000MHz, $870,
9) 4600+ X2 ($690), 12x, (5:6), 2400MHz, 3600MHz, $870. (impossible to reach)
It is doubtful that memory prices will decrease significantly while it can be expected for cpu prices to decrease. It may be worthwile to purchase DDR500 RAM now with a lower range processor and then upgrade that processor in a year or two. But then I usually just purchase a whole new system and give my system to a family member when it comes time to upgrade.
Birdpup you have done it again! You explain everything instead of giving a "few sentence answer" that explains little. You said " am not familiar with a Revision 3 AMD cpu since my 3000+ Venice is a Revision E. Is this a typo or am I missing something? " I actually got this from something I read here at Anantech but I guess I misunderstood what it was saying. Quote: 3200+ Venice - Ver E6, 3000+ Venice - Ver E3 & E6 (E6 is a Venice Core with updated memory controller), 3500+ Venice ending in "BW" is the E6 stepping chip and 3500+ Venice inding in "BP" is the E3 stepping chip with weeks 15 & 17 the best. The E6's don't overclock well. So to everyone that has tried to help me understand all this, here is the thousand dollar question: What I guess I just need to know if it was the E3 version that I need? Now since I have NEVER overclocked and know absolutely nothing about it or even how to read most of the numbers Birdpup listed am I better of to just get a SD 3700+ and use it as is - straight out of the box. Or should I go ahead with (which) the 3000+ Venice or 3200+ Venice and start learning how to overclock. (Which I wouldn't mind doing). Thanks everyone for all your help
I'm surprised no one's recommended dual-core. In all honesty, dual-cores are the future. There is no sense to buying a new single-core CPU anymore, unless you are strapped for money. If you have the money for a 3700+ San Diego, just go up a little bit to the 3800+ X2. You can overclock it to 2.4-2.6GHz, and you'll get the same performance as having TWO overclocked 3000+ Venices.
Personally, I would go with a 3200+ Venice and overclock it. Don't expect to hit any higher than 2.5-2.6GHz. The key is not to expect it. Of course there's always the possibility that you could hit higher like 2.7-2.8GHz+ but not commonly.
Also. Most members of this board prefer to keep voltages <1.6V on air so you should keep that in mind.
Also keep your temperatures in check. You may get your CPU up to 2.8GHz but it could be running at 70C load which is no good.
Generally, you want <55C load on air.
Also. Most members of this board prefer to keep voltages <1.6V on air so you should keep that in mind.
Also keep your temperatures in check. You may get your CPU up to 2.8GHz but it could be running at 70C load which is no good.
Generally, you want <55C load on air.
I am not familiar with these specifics. It is probably the luck of the draw when the cpu is shipped to you. Some vendors may stock certain production weeks since that may be what they were shipped. It is best to return to the article or thread this data was taken from and see if more information is provided, or even ask for more information concerning how to order the specific E3 weeks 15&17. If you will not overclock, then this information does not matter and any version of the ordered cpu will work just fine.
This is entirely your choice depending on your comfort level and willingness to experiment, read, and learn new material. Zebo's overclocking guide stickied at the top of this category explains the procedure very well in a much simpler manner than I have provided the information here.
It is possible the information I have provided will be difficult to understand for someone who has not yet worked out the details of overclocking and this is probably why few people wish to go to the effort of explaining all the details. Hopefully the comparison data at the end will be useful.
I have added RickUK's, Bona Fide's, and Kensai's choices above.
Quoted for emphasis.
THANK YOU to everyone who has offered their advice to my questions. It's nice to have a place like Anandtech to go to for help. This seems like one of the few sights that really has "up to date" information and help. Thanks again everyone
uOpt
Golden Member
- Oct 19, 2004
- 1,628
- 0
- 0
Just one thing I didn't quite like about the above advice: expentive RAM is almost never worth it from a pure performance standpoint, especially not when you have a cheap CPU. Real-world and gaming benchmarks leave few doubt that DDR400 is pretty much OK and if anything low latency is the way to go for AMD64.
It is important to understand that, on AMD64, you don't have top have memory capable of running at 225MHz if you set the bus to 225MHz. Because there is a divider that allows you to tune down your RAM after bumping up the bus speed.
Another thing that hasn't been mentioned is that the FX chips have a clear advantage, a multiplier that is unlocked so that you can bump it up, not only down. This allows you to overclock the CPU and just the CPU without any fiddling at all with the buses or any other multiplier.
It is important to understand that, on AMD64, you don't have top have memory capable of running at 225MHz if you set the bus to 225MHz. Because there is a divider that allows you to tune down your RAM after bumping up the bus speed.
Another thing that hasn't been mentioned is that the FX chips have a clear advantage, a multiplier that is unlocked so that you can bump it up, not only down. This allows you to overclock the CPU and just the CPU without any fiddling at all with the buses or any other multiplier.
This is true. I did not mention that MHz != performance. MHz only equals clock speed of the system. It may not be worth the extra cost of the more expensive memory to obtain the higher clock speed (MHz) because the actual performance improvement may only be 3-8%. It is far easier to just run the memory with a divider than it is to purchase more expensive memory.
I wanted to work that out though because I had not seen anyone else do it yet.
I wanted to work that out though because I had not seen anyone else do it yet.
So if I go FX then should it be FX 53? I have already gotten PNY PC3200-1GB. I have always used PNY with no problems and that's why I went with it. So then what FX? Thanks.
don't get an fx, its a waste of money
you have 3 choices really
a 3000, the 3700 or the x2 3800
it depends on how much you want to pay, i would get the 3000 if you don't want to spend that much or the 3800 if you want to future-proof yourself
you have 3 choices really
a 3000, the 3700 or the x2 3800
it depends on how much you want to pay, i would get the 3000 if you don't want to spend that much or the 3800 if you want to future-proof yourself
Yikes !!! -- I guess THIS must be the right place for "confuzzled Noobies" ... huh ? [giggles] Rich -- addr this to U cuz you are showing ON and active while I am writing it -- hope that is OK ... YOU and "Birdpup" seem to be the "True Neolithic Gods" of the CPU world and as such, I would like to ask what I think MIGHT be the question of the lips of MILLIONS of not just noobs right now -- A source inside AMD (lives down the beach from me) .. indicated this last w/e that AMD routinely uses X2 - d/core chips for "under-revving" to whatever market is in the most need Ok ? (Good point Bird !!) The issue is THIS for me: with the choice made to use Venice 3000/3200 (not sure yet) for the base proc of my first new system in a VERY long time, my BOEING engineering side is telling me to be VERY careful about determining not THAT AMD is dumming down "TOLEDO"(s) to make their count for available Venice customers, but ...... WHY they are using "fault ridden/defective" d/c's to do it -- Don't get me wrong, I totally understand the bottom line decision to trash 500,000+ chips that show bad L2 caching -- that adds up to like $20,000,000 and even AMD doesn't have pockets THAT deep ! My issue is for MYSELF (and those OTHER customers) who simply wish to know they are buying "un-fooled with" gear, and opt for a "PURE SINGLE CORE Venice 3000/3200 Rev. E6 ... and our need for SOME POSSIBLE WAY .. I repeat SOME POSSIBLE WAY to identify w/o ANY shadow of a doubt ... which E6 Venices are single cores... and which are disabled TOLEDO(s) .. FROM THE OUTSIDE OF THE CHIP !! The indication from the [current employee] is that ... THAT INFORMATION .. is buried in the ID/Serial # of EACH chip .. and a LONG discussion w/ AMD on the fone today found them ...Um ... LESS THAN WILLING .. to follow much of a discussion about this question. That's kinda it -- No one will EVER convince me that I shouldn't break out laughing if an AMD engineer were to tell me when asked to ID ANY chip of theirs ... that "I DUNNO" and they would have to crack it open to look !!! That is so INCREDIBLY beyond the pale of even CHILDISH REASON as to be ....... (fill in what seems right) !! [giggles] ... If YOU .. or ANYONE on this site has this answer and not just suppositional guessing (I mean those are fun to read too) BUT ........... I would love to hear how to do this and KNOW I am getting/buying the chip I am wishing/paying for !! ...
As a new member here, I hope this is NOT a"trash talk" response, but is taken as one of importance -- IF the issue were just OVER-PRODUCTION (for whatever silly reason), and the TOLEDO(s) are just "half alive", I could live w/ that -- But, the issue of "down-revving" chips w/ initial faults (i.e. BROKEN traces, etc) just goes against my grain as a VERY RESPONSIBLE and LONG TIME Materials Engineer @ BOEING. Going back to the drawing board is NOT a death sentence for AMD regardless of whether they believe it or not ! @ BOEING -- trust me -- we live w/ that reality EVERY day ! Any response to this would be WARMLY and APPRECIATIVELY welcomed --------
:: greybunny ::
As a new member here, I hope this is NOT a"trash talk" response, but is taken as one of importance -- IF the issue were just OVER-PRODUCTION (for whatever silly reason), and the TOLEDO(s) are just "half alive", I could live w/ that -- But, the issue of "down-revving" chips w/ initial faults (i.e. BROKEN traces, etc) just goes against my grain as a VERY RESPONSIBLE and LONG TIME Materials Engineer @ BOEING. Going back to the drawing board is NOT a death sentence for AMD regardless of whether they believe it or not ! @ BOEING -- trust me -- we live w/ that reality EVERY day ! Any response to this would be WARMLY and APPRECIATIVELY welcomed --------
:: greybunny ::
SolMiester
Diamond Member
- Dec 19, 2004
- 5,330
- 17
- 76
why create a 3800+ if yields were bad enough that they had to start neutering one of the cores and sell them as single cores
but most of their volume is going to be in the 3000/3200 but that should be coverable from their higher-end single cores
the only way i know to check would be to take off the heatsink (IHS) and check if you see one or two metal boxes
it could explain why the e6's don't seem to be overclocking that well (compared to the e3's)
but most of their volume is going to be in the 3000/3200 but that should be coverable from their higher-end single cores
the only way i know to check would be to take off the heatsink (IHS) and check if you see one or two metal boxes
it could explain why the e6's don't seem to be overclocking that well (compared to the e3's)
Coom-- Thanks for the look-see and thoughts. Just makes me ALMOST start considering looking @ Intel !! (kinda just kidding) [giggles]
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 23K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Intel processors crashing Unreal engine games (and others)- Started by BFG10K
- Replies: 2K
Facebook
Twitter