Simple inquiry demanding sophisticated response.
Simple inquiry demanding sophisticated response.
Over the last six months, I've been consistently overclocking my system with good results. I'm confident in my approach, but there are aspects that still puzzle me. Topics like Northbridge, HT Link, and the impact of lowering RAM speed on performance are areas I'm still exploring. Right now, I'm focusing on the Northbridge and hyper threading, though I find RAM speed confusing. I buy RAM with a low timing rating intentionally to avoid complications. My main concern is how RAM speed influences CPU performance. In overclocking, the idea often seems to be that higher is better. But I'm discovering this isn't always true. Stability remains the top priority. Ryzen also features "infinity fabric," which I understand is a trademarked design of its CCX chips. I'm still trying to grasp the details but am making progress. Many people are confused about how RAM speed affects performance in unexpected ways. My main question is: for purely gaming tasks, should I aim for a higher maximum clock speed on the CPU, or should I stick with increasing bus frequency? As you know, boosting all frequencies together—Clock speed, bus frequency, and RAM speed—can improve communication between components, which seems beneficial. But I'm wondering if this is the right path. Is it better to increase clock speeds without adjusting other frequencies, or is there another way? Thank you for your time and insights. If you can help clarify any of the other points I mentioned, I'd really appreciate it!
I can address the Ryzen question. All current Ryzen CPUs (Ryzen 7 and Ryzen 5) comes from the same die... they're effectively the same physical CPU just with cheaper models having some resources disabled. These CPUs have two "Core Complexes" (CCXs), with each CCX having 4 cores and 8MB Lvl3 Cache. The "infinity fabric" is the interconnect between these two core complexes. While it is fast, it is relatively slow and high latency. This means that cores on the same CCX can share more data at a lower latency than cores from the opposite CCX.
When Ryzen was reviewed it was shown to have very capable single threaded performance, but for some reason that strong single threaded performance didn't translate to games as we'd expect it to. One of the theories explaining this was that games need high bandwidth between cores and on Ryzen, CPU cores could be left underutilised because they were regularly waiting on data from a core on the opposite CCX.
One solution to this is through game optimisation, getting game developers to code their games such that threads that are depending on one another are assigned to cores on the same CCX. These cores can communicate with oneanother with higher bandwidth and lower latency, improving performance.
The other solution/mediation is in making the infinity fabric run faster. AMD confirmed that the performance of the infinity fabric is tied to RAM speed, so faster ram = faster infinity fabric, which explains why game performance particularly scales so well on Ryzen CPUs with faster RAM.
In terms of your larger question, I don't claim to be an expert on OCing, but remember that "faster is better" is only true when the component you're tweaking wasn't fast enough to begin with. There's no need to make PCIe slots run faster, for example, because a x16 lane PCIe 3.0 slot is plenty fast enough for a GPU in a gaming machine. My understanding is (and again - I'm no expert so someone will correct me if I'm wrong) that raising the base clock doesn't really net performance gains, which makes me question the merits of doing so. I'd also say that from my (somewhat limited) experience, BCLK overclocking is more difficult to stress test and find stability and results in much less graceful failures than simply tweaking a CPU multiplier and core voltage. I was completely stable in Prime95 but would hard-lock or BSD crash in a graphics benchmark with my BCLK OC. A multiplier OC on the CPU, on the otherhand, would simply result in a notification on Prime95 that a core had failed. To fix that, 99% of the time a have a simple choice between raising the voltage and lowering the frequency... simple!
I'm glad you found the solution helpful. I managed to upgrade the bus clock speed from 200 to 220, achieving a stable performance of 4400mhz in the core, 2840mhz on the HT link, 2400 on the NB, and 1720mhz for RAM. After applying a 23x multiplier, the core clock reached 4840mhz, with a voltage set at 1.525v. This is considered high, but I'm using a custom waterloop setup with a 480mm radius and maintaining temperatures around 47°C. The stability was confirmed using Prime95 in blend for several hours without any throttling or failures. Overall, I've noticed significant performance improvements with this configuration. However, I'm learning that workstations and gaming rigs have different needs—higher clock speeds and IPCs are crucial for games. As seen with the 7700k, this setup works well. Still, I wonder if simply raising the core clock to 5.0ghz without adjusting other parameters would yield better results compared to my current setup. Also, thank you for explaining Ryzen architecture in simple terms. PS: My RAM is GSKILL Sniper at 1866mhz, but the system won't boot at that speed and will only run at a maximum of 1600 (8.0x) with XMP if boosted, defaulting to 1333mhz.
I appreciate your response. I managed to clock the bus faster. The processor I'm using is an AMD FX 8350. I raised the bus speed from 200 to 220, achieving 4400mhz in the core, 2840mhz on the HT link, 2400 on the NB, and 1720mhz for RAM. Then I applied a 23x multiplier which pushed the core clock up to 4840mhz. I set the voltage to 1.525v, which is considered high, but I'm using a custom waterloop with a 480mm radius. My temperatures stayed at 47°C. I tested stability with Prime95 in blend for several hours—no throttling or failures. So I’m confident my current overclock works well. Regarding your question, I’ve noticed clear performance improvements with this configuration. Still learning that workhorse and gaming rigs have different needs. Higher clock speeds and higher IPCs are what games require. As shown by the 7700k, that’s why I’m curious if simply raising my core to 5.0 GHz would be better than my current setup with all other parts upgraded. Also thanks for explaining Ryzen architecture in simple terms. P.S.: My RAM is GSKILL Sniper at 1866mhz. But the computer won’t boot at that speed; it will only run at a maximum of 1600 (8.0x) if I enable XMP, and defaults to 1333mhz.