low ram write speed
low ram write speed
poor memory write is typical on FX CPUs due to slow write cache (L1/L2). Overclocking the northbridge won’t help; it only raises RAM speed to 2GHz or higher. Boosting your CPU will provide a much larger RAM boost. You might experiment with memory/bank interleaving—there are likely several options, so choose the one that offers the best performance.
The memory write performance is generally poor on FX CPUs due to slow write caches (L1/L2). Overclocking the northbridge won’t help; it only raises the RAM speed limit to 2GHz or higher. CPU overclocking will actually boost RAM performance significantly. You might experiment with memory/bank interleaving—there are likely some settings that improve results. It would be nice if there were more options. Thank you! I’m really eager to overclock my CPU since I have a 4+1 power phase MB. Lowering the timings to 1333MHz could improve how the RISC handles RAM, but it might reduce memory copy and read speeds while slightly increasing write performance. I already have bank and channel interleaving enabled, so I’ll test those. Is there a way to enable AMP? I don’t mind, but I currently get 30%-40% lower scores on my 1060 6GB CPU compared to others—7600 points versus over 10000 on a 3GB model.
Mircea Benga:
InfoSponge16 shared some insights suggesting the issue might lie with the board or the RAM. He mentioned that incorrect memory types could prevent the board from handling all four sticks at 1600 speed. It seems important to check the chip configuration in the RAM. If it's not compatible, running two sticks instead of four might be necessary. The board may also only support four 2GB sticks at full speed.
He provided details about his experience with BIOS memory: each stick is rated at 1600Mhz, and in ASRock tuner it shows a good performance (11215 MB/s read, 11482 Mb/s write), though the memory write speed was lower (5710 Mb/s) and latency was around 63.3 ns.
He noted uncertainty about whether the memory controller is on the board, which could affect thermal performance. Running the memory faster might help but could also increase heat. Checking the board's ventilation and considering alternative cooling methods would be wise.
hyperthreading is for storage, nb is for ram transfer limit
as for half write speed then read...theres nothing u can do about it, cpu was designed this way
u can check other userbenchmark scores, click on your cpu in there that will give u page with other users test for this cpu, open any, and check theirs ram speed, no matter which frequency/timings they use, its about same result u have
kerberos_20 :
hyperthreading is for storage, nb is for ram transfer limit
as for half write speed then read...theres nothing u can do about it, cpu was designed this way
u can check other userbenchmark scores, click on your cpu in there that will give u page with other users test for this cpu, open any, and check theirs ram speed, no matter which frequency/timings they use, its about same result u have
I remember the name, its not hyperthreading. Its the Hyper transport bus. Which is the connection between the CPU and the sb. Storage and PCI/PCIe lanes are connected to the CPU through the Hypertransport bus. You aren't wrong about the way some programs report the speed of the memory. However, he still higher latency by a hair, his graphics seem to be quite low compared to others. Or that is what he has said.
Because he has issues with his memory latency, and graphics, I suspect an inner case ambient temp issue.
InfoSponge16 provided insights on storage and RAM transfer limits, mentioning hyperthreading is related to storage rather than RAM speed. They discussed checking benchmark scores for other users and noted discrepancies in reported memory speeds. The person also referenced the Hyper transport bus as a connection between CPU, storage, and PCI/PCIe lanes, suggesting potential causes for latency issues.
InfoSponge16 shared insights on performance factors. Hyperthreading applies to storage, not RAM transfer limits. At half speed, read operations remain constrained; the CPU architecture was intentional. Review other user benchmarks by selecting your CPU—this will display tests from other users for similar models. Open any benchmark and compare their reported RAM speeds across different frequencies or timings; results should align closely with yours.
I recall the terminology, but it’s not about hyperthreading. It refers to the Hyper transport bus, which links the CPU to the motherboard, storage, and PCI/PCIe lanes. You’re correct that some programs may misrepresent memory speeds due to slight latency differences. Given his memory-related concerns and graphics performance, I suspect an internal ambient temperature problem.
I don’t think his graphics card is malfunctioning, as he hasn’t specified the model. There are three variants of a 6GB card—6pin, 8pin, and DDR5x. My previous benchmark used a 6GB GTX 1060 with a 6pin PCIe slot, achieving lower voltages for overclocking and saving about 6 watts via an external fan controller.
Here’s the latency for an FX 5950 paired with 1866 RAM: 63ms. His GTX 1070 runs at 65.1%, while his 6GB card reaches 68.8%, matching its expected performance. He likely uses a WindForce 6pin PCIe edition with a 150W TDP, which should help reduce latency slightly when overclocking the CPU and RAM.
According to your observations, his card is functioning properly. I didn't invest much effort in gathering his answers. I also lack direct experience with that particular card. His memory latency is recorded at 63ns, and on the user benchmark it indicates underperformance. Another test showed 60ns with excellent results. It's puzzling that both AMD systems report 3ns as 58.3ns, while my results are slightly quicker. Someone mentioned his system is comparable to others, with a 970 versus your 990fx, which might explain the minor variation.
userbenchmark ignores ns details, it just displays benchmark results for the same hardware on different systems. As long as you see yourself in the lowest to highest settings, it's okay. If you notice yourself outside those ranges, that might be incorrect.
If someone from that era could get intel, the same ram sticks would run much better.
The discussion revolves around storage and memory performance, noting that hyperthreading applies to storage but not RAM transfer limits. At lower speeds, the CPU's design plays a role. Comparing benchmark results shows the card performs well relative to others, with some minor latency concerns. The user suspects environmental factors affecting performance. A specific benchmark is shared for reference.