Is 6MB L2 cache much better than 3MB on a Core 2 Duo?

[TurtleCrusher]

I bought a really cheap laptop and wanted to upgrade the CPU. The specs are as followed:

Dell Studio 1737
Core 2 Duo T6400 2Ghz 2MB L2 35W
4GB (2x2GB)DDR2 PC6400
Radeon 3650 256mb
17" 1440x900

I bought a P9500, which is a 2.53Ghz 6MB L2 25w CPU for $14. It was defective so I purchased a P8700 for $7, which is the same 2.53Ghz 25w CPU, but only has 3MB L2 cache.

The P8700 is a solid upgrade from a 2Ghz T6400, and so much cooler. I just know that a P9500 has the most cache supported on this platform for a reasonable price. Is it worth it to upgrade from a P8700 to P9500 for just 3MB more of L2 cache?

Edit: I'm using this with CPU demanding games like Starcraft 2 and Source engine games. Just trying to get the most out of this laptop as it only cost me $30.

 

[ninaholic37]

Anand says you might get 0-8% speed increase:
http://www.anandtech.com/show/2464/12

 

[krumme]

Depends if you are on border fps so to speak. There is just a minor difference in games for the cache sizes but its there.

I think its like 5 ears ago when i upgraded a laptop from a 65nm 1.5 ghz core2 to 2.5 ghz 45nm 6mb cache penryn to play tf2. Cost me about 400usd. Lol. Was totally worth it.

That 6mb cache penryn was very fine mobile cpu. Imo nearly the same as the dual core i5 32nm i got that succeded it in the next laptop.

I say try the difference

 

[SPBHM]

6 vs 3 vs 2 MB

http://www.madshrimps.be/articles/a...0-Best-bang-for-the-buck-Cache-test-revisited

I wouldn't worry about it, since your GPU is the Radeon 3650.

 

[TurtleCrusher]

Anand says you might get 0-8% speed increase:
http://www.anandtech.com/show/2464/12

Didn't find this. Thank you. :thumbsup:

Depends if you are on border fps so to speak. There is just a minor difference in games for the cache sizes but its there.

I think its like 5 ears ago when i upgraded a laptop from a 65nm 1.5 ghz core2 to 2.5 ghz 45nm 6mb cache penryn to play tf2. Cost me about 400usd. Lol. Was totally worth it.

That 6mb cache penryn was very fine mobile cpu. Imo nearly the same as the dual core i5 32nm i got that succeded it in the next laptop.

I say try the difference

I think I'll shop around next time I have some disposable income. Even though they're so freaking cheap now, Christmas has me broke! I didn't anticipate these Core 2 Duo laptop CPUs to be so affordable.

The Dell Studio 1737 is one of the easiest laptops to upgrade the CPU. The T6400 was hot as hell and a little Arctic Silver Ceramique worked wonders. Then the switch to a 25w P series CPU is a godsend for power consumption and never having the fan to kick in unless I'm pushing it. If you have a Core 2 Duo laptop, UPGRADE THE CPU!

 

[TurtleCrusher]

6 vs 3 vs 2 MB

http://www.madshrimps.be/articles/a...0-Best-bang-for-the-buck-Cache-test-revisited

I wouldn't worry about it, since your GPU is the Radeon 3650.

In some games I'm definitely GPU limited, but in Source and Starcraft 2 I didn't see much of a change upping the resolution, so a CPU upgrade was definitely needed. It's about 1000x better than the Intel x3100 in my old Thinkpad T61 that I gave away.

 

[krumme]

Didn't find this. Thank you. :thumbsup:



I think I'll shop around next time I have some disposable income. Even though they're so freaking cheap now, Christmas has me broke! I didn't anticipate these Core 2 Duo laptop CPUs to be so affordable.

The Dell Studio 1737 is one of the easiest laptops to upgrade the CPU. The T6400 was hot as hell and a little Arctic Silver Ceramique worked wonders. Then the switch to a 25w P series CPU is a godsend for power consumption and never having the fan to kick in unless I'm pushing it. If you have a Core 2 Duo laptop, UPGRADE THE CPU!

Ahh if christmas made you broke stay on the 3mb. Freaking good upgrade for 7usd you did there.
Enjoy.

 

[Matt1970]

Quick spec lookup shows your T6400 runs a 800 bus and the P9500 is a 1066 bus. That may not even work in there. Laptop motherboards typically don't have a lot of upgrade room especially with higher bus speeds.

 

[krumme]

Yes. I remember i used the t9300 with 800mhz bus for that reason.
Still crazy it worked under the same tdp.

It cost me 400 hours of tf2 because it ran fine.

Btw after 2.5 year the nv 8600m gpu broke - as all 8 series did at that time i knew - we were baking gpu all over. Lol. But i got it repaired under guarantee because i used a lot of threat. So one of the kids continue to use it for a couple of years.

 

[TurtleCrusher]

Quick spec lookup shows your T6400 runs a 800 bus and the P9500 is a 1066 bus. That may not even work in there. Laptop motherboards typically don't have a lot of upgrade room especially with higher bus speeds.

The Dell Studio 1737 used the x45 chipset and is compatible with 1066fsb CPUs. It just utilizes a memory divider. The only known no-nos with this particular chipset is extreme edition and quads. The P8700 I have in it now is a 1066 fsb CPU and runs at spec according to CPUz.

This chipset even goes as far as enabling quick boost/turbo for single threaded tasks, hitting over 2.6 ghz at times. The multiplier is extremely variable compared to the T6400. The T6400 would always clock at multiplier intervals of whole or half numbers, while the P8700 does any interval (6.2, 7.7, 8.3, etc.) based on load.

 

[TurtleCrusher]

Yes. I remember i used the t9300 with 800mhz bus for that reason.
Still crazy it worked under the same tdp.

It cost me 400 hours of tf2 because it ran fine.

Btw after 2.5 year the nv 8600m gpu broke - as all 8 series did at that time i knew - we were baking gpu all over. Lol. But i got it repaired under guarantee because i used a lot of threat. So one of the kids continue to use it for a couple of years.

I assume the mobility 3650 is similar in performance to the 8600m. The limitation I see on mine is 256mb of ddr2 (dedicated). According to CCC it does support 2GB of hypermemory, so my ddr2 800mhz ram might not be much worse. If I put most settings on low at or near native res, the CPU/GPU is fairly balanced. In starcraft 2 in the intense custom games, I'm still CPU limited but the 530mhz/1MBl2 improvement is very noticeable.

Original half-life engine games are wholly CPU dependant and has made it playable on full servers now.

 

[krumme]

The Dell Studio 1737 used the x45 chipset and is compatible with 1066fsb CPUs. It just utilizes a memory divider. The only known no-nos with this particular chipset is extreme edition and quads. The P8700 I have in it now is a 1066 fsb CPU and runs at spec according to CPUz.

This chipset even goes as far as enabling quick boost/turbo for single threaded tasks, hitting over 2.6 ghz at times. The multiplier is extremely variable compared to the T6400. The T6400 would always clock at multiplier intervals of whole or half numbers, while the P8700 does any interval (6.2, 7.7, 8.3, etc.) based on load.

Okey. Interesting. I couldnt change multiplyer on my mb as i recall. The memory was 800 from the go but bus went from 667 to 800 with the penryn upgrade.

cant remember chipset but i think x45 was newer?

 

[krumme]

I assume the mobility 3650 is similar in performance to the 8600m. The limitation I see on mine is 256mb of ddr2 (dedicated). According to CCC it does support 2GB of hypermemory, so my ddr2 800mhz ram might not be much worse. If I put most settings on low at or near native res, the CPU/GPU is fairly balanced. In starcraft 2 in the intense custom games, I'm still CPU limited but the 530mhz/1MBl2 improvement is very noticeable.

Original half-life engine games are wholly CPU dependant and has made it playable on full servers now.

Isnt starcraft this sony ps3 port? As i recall it demands a lot of single thread cpu power. If its the game i remember we had trouble running with a 2.5-3ghz 2mb core2 penryn desktop for one of the kids.
Upgraded to hw quad because of that bad port.
Yes i understand why you ask then. Lol. Would be curious about the extra cache here to. Hmm.

 

[krumme]

Ps3 is 512kB l2 cache xb360 is 1MB...but ofcource the coding is completely different and all the work done in the support cores in the ps3 will have to be done in a single core as that made the porting easier.

 

[TurtleCrusher]

Diablo 3 was made for the consoles as well as PC, but Starcraft 2 has been PC only.

The game does require single threaded performance and single threaded performance only. The odd thing is my desktop plays this same game at a much higher FPS and I've got a CPU that's labeled as bad at single threaded performance. I was thinking it might just be all the available cache allocated for a single thread (2MB L2 + 8MB L3) that makes it good when things get extremely intense.

While the game is now playable with just a 530mhz/1MB added L2 cache bump, I am just wondering if 3MB more L2 might do the trick in making it smoother. My GPU load while the most intense moments in a custom map is 3-5%, showing the CPU is struggling to do all the work.

 

[SPBHM]

Diablo 3 was made for the consoles as well as PC, but Starcraft 2 has been PC only.

The game does require single threaded performance and single threaded performance only. The odd thing is my desktop plays this same game at a much higher FPS and I've got a CPU that's labeled as bad at single threaded performance. I was thinking it might just be all the available cache allocated for a single thread (2MB L2 + 8MB L3) that makes it good when things get extremely intense.

While the game is now playable with just a 530mhz/1MB added L2 cache bump, I am just wondering if 3MB more L2 might do the trick in making it smoother. My GPU load while the most intense moments in a custom map is 3-5%, showing the CPU is struggling to do all the work.

FX is bad for ST compared to desktop Intel CPUs like Sandy Bridge-Haswell, compared to Core 2 Duo, specially at 2.5GHz, the 4GHz FX is good for single thread...

you are focusing on cache more than you should in my opinion, yes, it can help in this game, but you are still talking about a difference that could be difficult to notice,

all 45nm "Core 2"
e5300 16,8FPS (2.60GHz FSB 200 2MB l2)
e7300 18,6FPS (2.66GHz FSB 266 3MB l2)
e8200 22,1FPS (2.66GHz FSB 333 6MB l2)




3% GPU load sounds wrong, unless the game is frozen or something,
also you might be overestimating your GPU, the 3650 only had 120sps from the HD 3000 era (3850 had 320) and the 500MHz DDR2 makes the memory bandwidth really low ... even the 8600M GT mentioned earlier would have a clear advantage, or current Intel IGPs.

 

[ninaholic37]

all 45nm "Core 2"
e5300 16,8FPS (2.60GHz FSB 200 2MB l2)
e7300 18,6FPS (2.66GHz FSB 266 3MB l2)
e8200 22,1FPS (2.66GHz FSB 333 6MB l2)

Hmmm, but if the bus speeds are all different, how do you know the bottleneck isn't the bus speed? Normalized it could mean:

e5300 16,8FPS/200*333 = 28.0 FPS/2.60*2.66 = 28.6 FPS
e7300 18,6FPS/266*333 = 23.3 FPS
e8200 22,1FPS/333*333 = 22.1 FPS

Worst case scenario.

 

[TurtleCrusher]

FX is bad for ST compared to desktop Intel CPUs like Sandy Bridge-Haswell, compared to Core 2 Duo, specially at 2.5GHz, the 4GHz FX is good for single thread...

you are focusing on cache more than you should in my opinion, yes, it can help in this game, but you are still talking about a difference that could be difficult to notice,

img snip


all 45nm "Core 2"
e5300 16,8FPS (2.60GHz FSB 200 2MB l2)
e7300 18,6FPS (2.66GHz FSB 266 3MB l2)
e8200 22,1FPS (2.66GHz FSB 333 6MB l2)




3% GPU load sounds wrong, unless the game is frozen or something,
also you might be overestimating your GPU, the 3650 only had 120sps from the HD 3000 era (3850 had 320) and the 500MHz DDR2 makes the memory bandwidth really low ... even the 8600M GT mentioned earlier would have a clear advantage, or current Intel IGPs.

When in custom matches like Nexus Wars in SC2 (not SP or regular online matches) have several hundred units to a thousand at times on any given spot, the FPS drops to 2-6fps near the end. It's far more intense than any single player or typical online match. The graphics are already on low, and the GPU only has to render 2-6fps, and the utilized Vram is only ~175mb. The CPU is pegged at 60% or so, with the game balanced onto two cores. Regular SC2 matches are a breeze on low and practically never hit sub-60fps. This does also reflect nearly the same performance issues I had with my old Q8200 @ 2.8Ghz in custom matches. That CPU only had 2MB of cache for a single thread, but was at 400FSB (1600Mhz) with the overclock.

I probably should've specified custom SC2 matches, because they are super demanding. I just find it nuts with the game maxed on my desktop (even CPU-specific settings) I get 15-25fps near the end of games, with small dips to 8-10 when I watch the action unfold directly. Even then my GPU isn't taxed at all with 4xAA.

 

[SPBHM]

Hmmm, but if the bus speeds are all different, how do you know the bottleneck isn't the bus speed? Normalized it could mean:

e5300 16,8FPS/200*333 = 28.0 FPS/2.60*2.66 = 28.6 FPS
e7300 18,6FPS/266*333 = 23.3 FPS
e8200 22,1FPS/333*333 = 22.1 FPS

Worst case scenario.

the fsb impact is negligible for most cases, and as you can see this test mostly confirms it, unless you think the impact of the l2 is even lower than what I think it is.

I've used C2Ds for years, apart from the impact on memory clock, if you managed to keep all the rest the same, the difference from 800, 1066 and 1333 was almost nothing, even with core 2 quads.

you can see here all the 3 configurations of l2 running with the same FSB
http://www.madshrimps.be/articles/a...the-buck-Cache-test-revisited/4#axzz3MaFmgMgh

but no starcraft 2.

When in custom matches like Nexus Wars in SC2 (not SP or regular online maatches) have several hundred units to a thousand at times on any given spot, the FPS drops to 2-6fps near the end. It's far more intense than any single player or typical online match. The graphics are already on low, and the GPU only has to render 2-6fps, and the utilized Vram is only ~175mb. The CPU is pegged at 60% or so, with the game balanced onto two cores. Regular SC2 matches are a breeze on low and practically never hit sub-60fps. This does also reflect nearly the same performance issues I had with my old Q8200 @ 2.8Ghz in custom matches. That CPU only had 2MB of cache for a single thread, but was at 400FSB (1600Mhz) with the overclock.

I probably should've specified custom SC2 matches, because they are super demanding. I just find it nuts with the game maxed on my desktop (even CPU-specific settings) I get 15-25fps near the end of games, with small dips to 8-10 when I watch the action unfold directly. Even then my GPU isn't taxed at all with 4xAA.

in that case I would think upgrading the CPU to something with almost the same performance is not worth it.

as you can see the difference from the cache is not enough to make 6FPS turn into anything good, it would probably just increase to 7FPS...

 

[fuzzymath10]

The Core 2 CPUs were never that great for SC2. Back in 2010-2011 I had a Q8200 + HD 5750 and the moment I replaced the Q8200 with an i5 2400 all my issues with dips in frame rates went away and I instantly became GPU bound.

 

[TurtleCrusher]

Okey. Interesting. I couldnt change multiplyer on my mb as i recall. The memory was 800 from the go but bus went from 667 to 800 with the penryn upgrade.

cant remember chipset but i think x45 was newer?

I believe the x45 chipset was the last for Core 2 Duo. Dell didn't allow those extreme edition/quads via BIOS restriction. I had the x35 chipset (santa rosa IIRC) with the T7500, but to disassemble of the T61 would've been extremely difficult without running the risk of breaking something. The Dell Studio is just unscrew the bottom and you're already at the heat-sink.

 

[VirtualLarry]

This chipset even goes as far as enabling quick boost/turbo for single threaded tasks, hitting over 2.6 ghz at times. The multiplier is extremely variable compared to the T6400. The T6400 would always clock at multiplier intervals of whole or half numbers, while the P8700 does any interval (6.2, 7.7, 8.3, etc.) based on load.

I've never heard this. The actual CPU only supports whole or half multis, Intel never produced any Core2-era CPUs with continuously-variable multis. Personally, I think that you are simply mistaken. No Core2-era CPU or chipset supported the notion of "Turbo" back in those days either.

 

[fuzzymath10]

No Core2-era CPU or chipset supported the notion of "Turbo" back in those days either.

http://en.wikipedia.org/wiki/Intel_Dynamic_Acceleration

I have a laptop with a T8300 tops out at 2.4GHz, but can reach 2.6GHz with one extra bin (200x13 vs 200x12) under certain conditions where just one core is loaded up.

 

[TurtleCrusher]

I've never heard this. The actual CPU only supports whole or half multis, Intel never produced any Core2-era CPUs with continuously-variable multis. Personally, I think that you are simply mistaken. No Core2-era CPU or chipset supported the notion of "Turbo" back in those days either.

Intel Dynamic Acceleration technology (IDA) has been supported on mobile since the Santa Rosa platform launch. One core hits the extended multiplier while the other hits its C3 state. Some people have modified their BIOS to run both cores at IDA speeds.

In regards to the variable multiplier, I'll get those screenshots when I get back on the laptop. This is the Penryn lower-voltage series of CPUs I'm talking about here, specifically tuned for lower power consumption.

 

[VirtualLarry]

http://en.wikipedia.org/wiki/Intel_Dynamic_Acceleration

I have a laptop with a T8300 tops out at 2.4GHz, but can reach 2.6GHz with one extra bin (200x13 vs 200x12) under certain conditions where just one core is loaded up.

Huh, interesting. Most of my experience with 775 was with Desktops, and I don't recall ever seeing that technology for their desktop CPUs.