Memory QVL tables |
Post Reply |
Author | |||||||||||||||||||||||
PetrolHead
Groupie Joined: 07 Oct 2015 Status: Offline Points: 403 |
Post Options
Thanks(0)
Posted: 10 Aug 2017 at 5:19am |
||||||||||||||||||||||
Based on what I've read, the CPU design is what is really limiting the memory speed, so we won't likely see very high memory clockspeeds for the current Ryzen CPUs even with coming AGESA updates. We'll have to wait and see what 1.0.0.7 brings. Improved compatibility will in any case be very welcome.
|
|||||||||||||||||||||||
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit
|
|||||||||||||||||||||||
parsec
Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996 |
Post Options
Thanks(0)
|
||||||||||||||||||||||
I see what you mean about the memory's rated speed vs the Speed column. Mainly if not all of them are in the G.SKILL entries (these tables are blurring my vision... ) We can see that the highest speed listed in the Memory column is 3200, while some of the G.SKILL models your are referring to are rated above 3200, up to 4266. I'm sure you know Ryzen begins reaching its memory speed limit, depending on the memory model, at ~2933, and getting to speeds above that is increasingly difficult. Most memory models above 3200 simply won't reach those speeds with Ryzen systems, which is why those models have 2933 in the Speed column. You can find examples of a few Ryzen systems that can run memory above 3200, but they are all on more expensive boards, normally with extensive manual tweaking of the memory timings and voltage. All Intel systems cannot simply run memory at the very high OC speeds too. We also have the factor of the AGESA updates from AMD attempting to improve memory speeds and compatibility. So things change but what is certain about an OC memory speed is that they are not a given and guaranteed. These memory tables are kept conservative in their maximum speeds, since the higher memory speeds are just not plug and play. |
|||||||||||||||||||||||
PetrolHead
Groupie Joined: 07 Oct 2015 Status: Offline Points: 403 |
Post Options
Thanks(0)
|
||||||||||||||||||||||
Thank you for the clarifications, parsec.
No, I meant 3200 MHz, since that is the rated speed for that particular G.Skill memory (F4-3200C16D-16GTZB). G.Skill states that the speed of 3200 MHz is tested, but like I said in the text you bolded, this does not necessarily apply to AMD platforms. In fact, I'm certain it only applies to Intel platforms, since G.Skill doesn't list any AM4 mainboards in its QVL for this memory. The G.Skill memory is actually the reason I found the speed column odd, since the speed column values do not generally match the rated speeds of the G.Skill modules and these happen to be the modules I've mostly been looking at. For example:
For other memory manufacturers the speed column value matches the rated speeds of the modules, which is why I assumed these are tested memory speeds and which is why I thought it's odd they list memory speeds 3000 MHz and 2800 MHz in the speed column. It seems, though, that it's a mix of both. Maybe only the overclock values are tested?
You are of course right that an overclock might make the 2667 speed unachievable. However, the note says nothing about the CPU being overclocked. Maybe each note is supposed to build upon the previous note, but if this is the case, it would be clearer to explicitly state this. |
|||||||||||||||||||||||
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit
|
|||||||||||||||||||||||
parsec
Moderator Group Joined: 04 May 2015 Location: USA Status: Offline Points: 4996 |
Post Options
Thanks(0)
|
||||||||||||||||||||||
As you wrote, the maximum specified Ryzen 7 and 5 processor speed is 2667. The 'v' in the OC column simply means the memory speed is an over clock. It does NOT mean the memory has been verified to OC to that speed on any system. No memory or processor over clock speed can be guaranteed. (1.) The v is missing in the Panram 2800 or any other 2800 speed memory because 2800 speed memory will be down-clocked to 2666, which is no longer a memory OC. That spec is actually no longer correct after the AGESA 1.0.0.6 update. 2800 memory is now supported by Ryzen, with a UEFI/BIOS with at least the AGESA 1.0.0.6 update. You wrote: " Does the "v" mean the 2993 MHz memory - which according to the manufacturer has been tested to work at 3200 MHz , although not necessarily with AMD platforms - can be overclocked further? To what extent?", which I don't understand. Did you mean 3000 MHz, not 3200 MHz? If so, Ryzen does not support a memory clock speed of 3000 MHz. The closest values are 2934 and 3066. Otherwise I don't see where a 3200 speed memory is listed to operate at 2933. It is impossible to predict how high if at all any memory will OC beyond its specifications, on any system. Normally the faster the memory OC spec is, the less likely it will OC beyond that speed. (2.) A processor over clock can affect the maximum memory OC possible. Meaning a high CPU OC may cause a reduction in the memory OC capability. I assume since 2667 is the maximum stock memory speed (for a stock processor core speed), it was found that a high processor OC may cause some memory models operating at 2667 to become unstable. (3.) That caveat is there to cover memory that has no heat spreaders at all, and/or when using four DIMMs, which causes less air flow between them, and/or for users that set the DRAM voltage above 1.35V. DDR4 memory with basic heat spreaders operating with 1.35V does not run hot, my FlareX 8GB x2 3200 kit (operating at 2933) reports between 28C - 30C. (4.) Ryzen is the first system from AMD that uses DDR4 memory. Intel had three PC platforms before Ryzen was released that supports DDR4 memory, and now has even more. 95% of existing DDR4 memory was designed to work with Intel systems. Intel platforms support DDR4 memory speeds of 2800 and 3000. Ryzen does not support a DDR4 memory speed of 3000 (1500MHz frequency). Also related to your question (2.), the XMP profile data of most DDR4 was designed for Intel platforms. These are most of the memory frequency's that Ryzen currently supports in ascending order. Recall that the memory "speed" of Dual Data Rate memory is twice the memory frequency. |
|||||||||||||||||||||||
PetrolHead
Groupie Joined: 07 Oct 2015 Status: Offline Points: 403 |
Post Options
Thanks(0)
|
||||||||||||||||||||||
So, I've been looking at the AB350M Pro4 memory QVL, as I might be
buying one at some point, and the table and its explanations raise some
questions. The tables are essentially the same for other motherboards, so the questions are not limited to the AB350M Pro4, but I will take my examples from that table. Some of these may be "dumb" questions, but here goes:
1. "Note3: The O.C. mode is not guaranteed. It depends on whole system configuration and other parameters." Fine, but what does the "v" actually mean in the OC column? That an OC has been verified? What sort of an OC was it? Take these rows for example:
The memory that's supposed to work at 2933 MHz has a "v" in the OC column, but the memory with the 2800 MHz doesn't, yet both speeds are overclock speeds, since the (stock) maximum memory speed for any Ryzen processor (as well as the board itself) is 2667 MHz. Does the "v" mean the 2993 MHz memory - which according to the manufacturer has been tested to work at 3200 MHz, although not necessarily with AMD platforms - can be overclocked further? To what extent? 2. "Note4: When running XMP at DDR4 2667MHz or higher, the system's stability depends on the CPU's capabilities." At 2667 MHz? That's the (maximum) stock speed for all currently available Ryzen CPUs. Surely there shouldn't be any question of stability at that speed when running XMP, if the modules themselves can handle it. 3. "Note5: For system stability, use more efficient memory cooling system to support a full memory load (4 DIMMs) when overclocking." Does this mean top-down blowing CPU cooler design? Is it enough if the modules have heatsinks? Shouldn't 4 DIMMs stress the memory controller and not the DIMMs themselves? 4. Notes 6 and 7: "Each RAM whose frequency is 3000 will downgrade to 2933 with AB350M Pro4" and "each RAM whose frequency is 2800 will downgrade to 2666 with AB350M Pro4". Why? The QVL has memory listed for 3000 MHz, such as G.Skill's F4-3000C15Q-16GRR, as well as 2800 MHz, for example Corsair's CMD16GX4M4A2800C16. If these will downgrade, what's the point in listing these memories as having these speeds? If not, what's the point of these notes? Edited by PetrolHead - 09 Aug 2017 at 1:39am |
|||||||||||||||||||||||
Ryzen 5 1500X, ASRock AB350M Pro4, 2x8 GB G.Skill Trident Z 3466CL16, Sapphire Pulse RX Vega56 8G HBM2, Corsair RM550x, Samsung 960 EVO SSD (NVMe) 250GB, Samsung 850 EVO SSD 500 GB, Windows 10 64-bit
|
|||||||||||||||||||||||
Post Reply | |
Tweet
|
Forum Jump | Forum Permissions You cannot post new topics in this forum You cannot reply to topics in this forum You cannot delete your posts in this forum You cannot edit your posts in this forum You cannot create polls in this forum You cannot vote in polls in this forum |