The new pricing on PC parts and Macs has me not buying anything for a while. The day after the Mac pricing hit, I decided to try doing some tuning on my Windows PC as I had been going back and forth on upgrading it and the price increases have dissuaded me there. There are no 9900X bundles at Microcenter anymore. I had also run tests on the other 9900X system and there's no performance benefit for my main program compared to Lunar Lake even though it has much higher single-core and multicore performance.
My Windows desktop has an ASUS TUF Gaming Z490 WiFi and an i7-10700 with 128 GB of RAM. It is hard to give up that RAM in this environment and I've looked at all of the potential ways to try to reuse it and nothing makes financial sense. So I went into the BIOS and on the main screen, it has an option called AI Overclocking.
It was set on Normal and I hit the right arrow and the option was Fast Tuning. I hit it again and the option was Extreme. I assume that there are options for undervolting too.
So I set it to Fast Tuning, saved and rebooted. Geekbench single-core increased by 1.7% and multicore increased by 10.7%. Overall responsiveness was a lot better too. This reminds me of going to The Apple Store and playing with the 2020 i5 iMacs and noting how responsive they were compared to my i7-10700 and I think that was due to them having K variants of the CPU.
I kept an eye on fan speeds, CPU package power and thermals for a day. Fan speeds stayed under 1,000 RPM (it's a huge case with large, slow fans and the CPU cooler is a huge chunk of metal with dual fans). Idle power before the change could be as low as 1.2 watts. It gets down to about 7 watts with the faster settings now. The 9900X build idle power usage is about 21 watts so I'm not concerned about higher package power usage for a noticeable performance improvement.
The GTX 1050 ti used 35-36 watts normally and I asked Deep Seek if I could improve power consumption and performance with a spare GTX 1660 Ti. I put the 1050 Ti in it thinking that it used less power because it's a much less powerful GPU. It turns out that the GTX 1660 Ti uses 12 watts for my workload.
So these two changes resulted in a net 17 watt decrease in power consumption.
Thermals are fine and it runs most of the time around 33-34 degrees with spikes to 43 degrees. CPU Package power high that I've noted is 83 watts but that's rare. The i7-10700 has a TDP of 65 watts and I think that the motherboard was limiting power at that level. It appears that the improved responsiveness is due to spending less time in low-power idle states and making more power available for multicore use.
I also put in an old Gen 3 NVMe SSD in the system which I will move my virtual machine to to reduce contention with foreground stuff. These old Gen 3 NVMe SSDs actually have some wear on them as the two I have are at 96% and 99% health. Nice to get a good amount of use out of these.
I ran a virtual machine file copy from guest to host and copy speed went from about 600 MBps to 900 MBps. The best I can get on my Mac Studio is about 110 MBps but that's due to VMWare, not the hardware. VMWare has a lot of catching up to do on ARM compared to x86. The 110 MBps limit is due to network speed; VMWare doesn't support internal copy between guest and host yet.
From what I can tell, a lot of hobbyists are looking at what I've looked at. Upgrading to newer systems, upgrading while reusing DDR4, tuning, etc. It is a difficult environment because prices of just about everything have increased as the prices of GPUs, DRAM and SSDs has gone up.
I'm working more on tuning my environment for performance and efficiency over buying new hardware given the current market realities.
