Overclocking the CPU using low-frequency RAM
Overclocking the CPU using low-frequency RAM
Hi!
I haven't done overclocking before, but I'm open to learning anything about it. My old PC has an i5 3450 and 1600mhz CL 9 RAM. I've read a lot about CPU performance in PS and LR, and I came across some useful articles from Puget Systems:
Adobe Photoshop CC Multi Core Performance
Adobe Lightroom CC/6 Multi Core Performance
These pieces suggested that PS and LR perform better with 4 cores, 6-8 threads, and a high-frequency CPU. I conducted some research and compiled this parts list:
PCPartPicker part list
/
Price breakdown by merchant
CPU:
Intel Core i7-6700K 4.0GHz Quad-Core Processor
($334.99 @ Amazon)
CPU Cooler:
Deepcool ASSASSIN II 70.1 CFM Fluid Dynamic Bearing CPU Cooler
($69.99 @ Newegg)
Motherboard:
MSI Z170A SLI PLUS ATX LGA1151 Motherboard
($109.99 @ Amazon)
Memory:
G.Skill Ripjaws V Series 32GB (2 x 16GB) DDR4-2133 Memory
($108.00 @ Newegg)
Storage:
Samsung 850 EVO-Series 500GB 2.5" Solid State Drive
($149.45 @ OutletPC)
Storage:
Western Digital Blue 2TB 3.5" 5400RPM Internal Hard Drive
(Purchased For $0.00)
Storage:
Western Digital Green 4TB 3.5" 5900RPM Internal Hard Drive
(Purchased For $0.00)
Video Card:
MSI GeForce GTX 780 3GB TWIN FROZR Video Card
(Purchased For $0.00)
Case:
Corsair Carbide 400C ATX Mid Tower Case
($84.99 @ Micro Center)
Power Supply:
Antec 750W 80+ Gold Certified Fully-Modular ATX Power Supply
($89.99 @ Newegg)
Total: $947.40
Prices include shipping, taxes, and discounts where applicable
Generated by PCPartPicker 2016-04-26 06:27 EDT-0400
*Actually, my 2TB WD blue spins at 7200 RPM, but I couldn't find one at PCPP. That's odd.
Questions:
1. Most overclocking guides recommend high memory speeds, but can I really push my CPU to 4.5–4.8 GHz with 2133 MHz RAM? I've heard that fast memory can help, but with my budget of around $950, I probably won't reach 32GB at a higher speed—maybe 2400 MHz max.
2. The article about Adobe Photoshop CS6 Memory Optimization suggested that high-speed RAM isn't worth the cost and that RAM faster than the CPU's native speed can increase failure risk. What are your thoughts? Should I prioritize cheaper parts, perhaps an SSD (even though I love the 5-year warranty on the 850 EVO) or a faster PSU?)
Thanks in advance.*
The data rate of the DRAM doesn't really influence the CPU OC unless you're using a very high-speed DRAM. Usually not unless you exceed 3200 and have a lot of memory. 2133 should work, but faster options are better if available.
Improved memory speed enhances general performance, though your OC might not be as effective; it doesn't affect the processor's OC. When you're using FSB, memory frequency should be reduced initially since it tends to increase with FSB. Changing the multiplier doesn't alter memory frequency.
There are two approaches to overclocking. The main difference lies in the memory frequency adjustment, and whether changing the multiplier affects it. The second method doesn't alter the memory frequency. Which one is more effective? It's a simple question, but I can help clarify if you need. Any relevant links or details would be useful. Thanks!
I reviewed the article and noticed significant improvements when upgrading from 2 cores to 4, but much smaller gains when moving from 4 to 8 cores. An i7 doesn’t actually have eight real cores—it appears to be eight cores pretending to be more. In reality, its performance is only about 1.3 times faster than an i5. Skylake performs slightly better than Haswell, which is marginally better than your CPU, but the difference remains minimal. In short, the 6700k will be noticeably faster than your 3450, though not a substantial increase. Tasks that normally take two minutes with the 6700 could finish in just three minutes compared to the 3450. If you’re a professional video editor or take it seriously as a hobby, investing in a new computer is wise since time really counts. Otherwise, you can tolerate the extra wait for each task. Regarding memory speed, it has far less impact than processor speed or core count. For instance, my 3570k runs at 2000 MHz CL9 and performs almost the same as my newer 4690k with 1600 MHz CL9 at the same CPU speed. A key factor is the SSD; it greatly enhances responsiveness. My 3570k uses a Z77 motherboard with SATA 3, which fully utilizes SSD speed. But if your motherboard only supports SATA 2, it will halve the SSD performance. If you lack SATA 3 support, upgrading would be beneficial.
If the processor supports unlocking the multiplier, it becomes simpler to optimize. Just increase the multiplier and supply the necessary voltage, ensuring stable operation without causing overheating. The VRM unit on the motherboard must deliver sufficient power (in Watts) to maintain performance; otherwise, power output will decrease and stability will decline. All calculations are based on full CPU load.
For enhanced speed and accuracy, the FSB frequency can be increased so that the memory controller operates at its peak. However, performance may drop if it exceeds its optimal range, which varies per processor. The chipset and its most efficient frequency also play a role. The PCI/PCIe bus speed might be adjusted above its standard 100MHz setting. This makes tuning over FSB more challenging compared to adjusting the multiplier, though with limited improvement. Fine-tuning for maximum performance can take several days, requiring extensive testing to assess gains and stability.
Many different parts are connected using a "Silicone Lottery," meaning even identical components can yield varying results. This makes extreme tuning impractical for everyday use. I mean, you end up going through all that and encounter a game so poorly optimized that even a supercomputer might struggle with it.
Oh, I forgot to mention. This PC is going to be my working PC. I'll mainly use it for Photoshop, Lightroom, and Illustrator / Corel Draw, some light 3D interior modelling, and weekend gaming on single 1920x1200p monitor. Going to pair it with two 1200p 25" monitor, and I'm planning to get 10-bit monitor + workstation GPU later this year for my primary monitor.
lodders :
I had a look at the article. You get a lot of benefit going from 2 cores to 4 cores, but a lot less benefit going from 4 to 8 cores.
An i7 does not have 8 real cores, it has 4 cores pretending to be 8 cores. Actual performance is 1.3x faster than an i5.
Well, from these article at Puget System, lightroom / photoshop doesn't work well with multiple CPUs and higher frequency is more helping than multiple cores. I'm glad to see that, because my current budget doesn't allow me to get anything better than 6700K. Also, later this year I'm going to use 10-bit monitor for better result in printed media, and I need to save more to get a workstation card.
lodders :
Skylake is a bit faster than Haswell, which is a bit faster than your CPU, but we are only talking about a few % difference.
Long story short, the 6700k will be faster than your 3450, but not a huge amount faster. Processing tasks may take 2 minutes with the 6700 instead of 3 minutes with the 3450. If you are doing video editing full time as a profession or as a serious hobby, then buy a new computer, because time is money. But if not, just live with having to wait a minute longer for the task to complete.
Thanks for the information, but video editing isn't part of my job (yet), but I'm currently learning and probably going to do that one day.
And yes, I'm planning to build a new PC - from locked ivy bridge i5 to unlocked skylake i7. The only components I keep from my old PC are the GTX 780 and some HDDs.
lodders :
As for memory speed, it has much less of an effect than processor speed or number of cores.
For example, my 3570k has 2000 MHz CL9 memory, and at the same CPU clock speed it has almost identical performance to my newer 4690k with 1600 Mhz CL9 memory
One big difference is an SSD, which makes a PC a lot more responsive. My 3570k is on a Z77 motherboard with SATA 3, which fully exploits the speed of an SSD. However, if your motherboard is SATA 2, it will limit SSD to half speed. If you don't have SATA 3, then that is another good reason to upgrade.
Ya, that's my main reason for asking this. I don't see the real benefit of getting higher speed memory, at least from these internet articles and benchmarks. But I think I need to overclock my CPU for best possible result in lightroom and photoshop, and since I never do OC before, I need to find out if a low speed memory affects CPU overclocking capability. If I can get higher CPU speed with higher frequency memory, then I'd consider myself to get cheaper case, SSD and PSU (maybe a 650W bronze?) to get higher memory frequency. But if it's possible to get the same result with lower speed memory, then I'd stick to my current partlist.
Good to know about the SSD. I'm quite sure 850 EVO is currently one of the fastest SATA III SSD that fits to my budget, and my next motherboard (on my partlist) has SATA III port. My old PC has 128GB sandisk SSD plus, and I had around 20GB of empty space after installing windows and all my apps + some light games. So I guess 500GB is more than enough for all my apps, and I'd have approximately 390-400 GB free space for my main export / import storage. Then I can use my slower 6TB HDD for my secondary / backup storage. I think it's pretty effective that way.
Again, thanks for your input. I appreciate it.
CountMike :
If processor has unlocked multiplier than it's easier to OC. Just raise multiplier and add required voltage to it, enough to run stable but without overheating. VRM module on the MB has to be able to provide enough power (in Watts) otherwise power will drop, with it voltage and stability will suffer. All that is calculated at full CPU load of course.
To get some more real performance (faster calculations) FSB could be raised so memory controller (new processors have it built in) runs to it's highest performance. Performance can suffer if you go over it's most sweet spot which may be different for each processor, there's no set value for all of them. There's also a question of chipset and it's most efficient frequency. PCI/PCIe buss may also raise from it's set value that is mostly 100MHz. That's why it's more difficult to OC over FSB than over multiplier but with limited gain. Tuning it up for maximum performance can take days because you have to run tests between every step to see if you are gaining or loosing performance and stability.
Tradesman1 :
The data rate of the DRAM doesn't really affect the CPU OC unless you have a very high data rate DRAM. Generally not unless you get up over 3200 and with a lot of DRAM. 2133 will be fine, but if you can get faster go for it.
😉
Wow, thanks guys for all these information. The 6700K does have unlocked multiplier.
So does it like... Once I get the highest stable speed by raising voltage and multiplier, I can go higher by raising the FSB clock after that?
If it does, then higher memory clock is always a better choice for OC, isn't it?
What if I only want a highest possible cpu speed with my current components, and I don't want (or simply can't afford) any memory higher than 2133mhz?
CountMike :
Many different components coupled by "Silicone Lottery", meaning that even two identical components may give different results, make that extreme tuning impractical for casual use.
I mean, you go thru all that and there comes a game that's so unoptimized that even a super computer may have troubles with it.
I've read about the "silicon lottery" on multiple OC'ing guide, and some says that the earlier CPU batch - mostly those from malaysia almost always worse for OC compared to costa rica CPU, but some says it depends on luck too. No wonder why it's called 'lottery'.
Also, pardon my english, I'm not from US, but I was using USD to make it easier.
On Intel it is preferred to use the Multiplier to OC, and leave the BCLK at 100MHz for 100% functionality. I don't see any reason to change it. It will be easier to hit higher clocks with lower speed RAM, but 2133 is easy for your CPU either way. IMO using XMP, and keeping VCORE below 1.3v is the way to go.