Raedon Memory for Ryzen?

[trents]

Did AMD get out of the RAM business? A year or so ago they were marketing their own brand of memory but I only see one 4gb stick of Raedon memory for sale on Egg. Must not have been profitable.

The reason I bring this up is because of all the issues with Ryzen and high speed memory compatibility. In researching this I discovered that AMD representatives are saying part of the problem is caused by RAM manufacturers still gearing their products to work optimally with Intel CPUs (secondary and tertiary timings I think).

It occurs to me that AMD could fix that issue by getting back into the RAM manufacturing business and turning out products that are more compatible with Ryzen CPUs. I suppose, however, with the shortage of ICs right now it isn't a good time.

 

[Bluefalcon13]

AMD never made memory. The contracted patriot to produce AMD branded memory. Basically normal patriot memory with an AMD sticker.

 

[rescuetoaster]

I'm not entirely sure they rolled out a DDR4 option - or if that was ever considered for that matter. The SSD and RAM lines were just rebrands anyways.

I think it does make sense to sell ram optimized for Ryzen, but I'd be concerned that the internet community at large would see it as intentionally designing Ryzen to only work with AMD brand components (even though we know this clearly isn't reality).

 

[trents]

Yes, they did roll out some in the DDR4 platform: https://www.amazon.com/AMD-Radeon-Memory-DDR4-R744G2133U1S/dp/B0195XZQEC

One issue I see here is that the economics of turning out a line of RAM optimized for Ryzen would hinge on selling a high volume of Ryzen processors and capturing a significant portion of the CPU market in order to make it pay. Kind of a catch 22 in some ways.

 

[wingman99]

Intel has the majority of the RAM market for there CPUs. AMD needs to conform to the memory out in the market now.

 

[trents]

Intel has the majority of the RAM market for there CPUs. AMD needs to conform to the memory out in the market now.

Undoubtedly, this is reality. So this would mean retooling the Ryzen IMC I assume to conform better to existing RAM configurations.

 

[wingman99]

Undoubtedly, this is reality. So this would mean retooling the Ryzen IMC I assume to conform better to existing RAM configurations.

I don't know why AMD would have to retool they designed and tested with JEDEC DDR4 and XMP.

 

[trents]

I reckon then they should have done more extensive testing.

This whole memory compatibility issue has put them behind schedule on Vega I think.

 

[Mr.Scott]

Question:
Why should AMD have to conform to their competitions specs? That's ludicrous business practice.
What AMD needs to do is 'grease' the memory manufacturers a little. You know, like Intel does.

 

[wingman99]

AMD does not make memory standers, JEDEC and intel XMP make the memory standards. Intel and AMD don't make memory they qualify what is on the market like the rest of us.

 

[trents]

But do the XMP and JEDEC standards specify secondary and tertiary timings?

 

[Mr.Scott]

AMD does not make memory standers, JEDEC and intel XMP make the memory standards. Intel and AMD don't make memory they qualify what is on the market like the rest of us.

You misunderstood me.
Memory manufacturers use Intel as their test standard.
They don't have to. They could use anybody else, or everybody else.
They do it because Intel kicks them back, just like they do with the OEM PC companys.
They use their clout and large budget to 'influence', thus helping maintain the stranglehold they have on the market. It's borderline racketeering.
It's the same thing as 'optimizing' benchmarks for Intel processors. Same reason. Intel kicks them back.

 

[wingman99]

But do the XMP and JEDEC standards specify secondary and tertiary timings?

Some of the tertiary timings, have a read in the links.

JEDEC and Intel XMP set the Serial presence detect specification. Extreme Memory Profile (XMP) A similar, Intel-developed JEDEC SPD extension for DDR4 SDRAM DIMMs. This uses bytes 176–255, which are unallocated by JEDEC, to encode higher-performance memory timings. LINK: https://en.wikipedia.org/wiki/Serial_presence_detect#XMP

JEDEC memory standards
JEDEC standards and publications are designed to serve the public interest through eliminating misunderstandings between manufacturers and purchasers, facilitating interchangeability and improvement of products, and assisting the purchaser in selecting and obtaining with minimum delay the proper product for use by those other than JEDEC members, whether the standard is to be used either domestically or internationally.
LINK: https://en.wikipedia.org/wiki/JEDEC_memory_standards

 

[bassnut]

I can tell you the new Agesa code does make a difference ..... my Tiachi woould not boot my Team ram above 2133. I just updated and it is running now at 2966 16 18 18 T1

 

[Alaric]

Some of the tertiary timings, have a read in the links.

JEDEC and Intel XMP set the Serial presence detect specification. Extreme Memory Profile (XMP) A similar, Intel-developed JEDEC SPD extension for DDR4 SDRAM DIMMs. This uses bytes 176–255, which are unallocated by JEDEC, to encode higher-performance memory timings. LINK: https://en.wikipedia.org/wiki/Serial_presence_detect#XMP

JEDEC memory standards
JEDEC standards and publications are designed to serve the public interest through eliminating misunderstandings between manufacturers and purchasers, facilitating interchangeability and improvement of products, and assisting the purchaser in selecting and obtaining with minimum delay the proper product for use by those other than JEDEC members, whether the standard is to be used either domestically or internationally.
LINK: https://en.wikipedia.org/wiki/JEDEC_memory_standards

Good intentions are abundant in the world, almost as plentiful as graft and corruption. Compilers were supposed to be neutral, too. Intel's market dominance and deep pockets reach far and wide.

 

[Mr.Scott]

Exactly.

 

[wingman99]

Good intentions are abundant in the world, almost as plentiful as graft and corruption. Compilers were supposed to be neutral, too. Intel's market dominance and deep pockets reach far and wide.

AMD and intel don't have a problem with all memory brands running DDR4 2133 stock, there built for it. AMD and Intel have trouble overclocking memory.

 

[trents]

AMD and intel don't have a problem with all memory brands running DDR4 2133 stock, there built for it. AMD and Intel have trouble overclocking memory.

Not sure what you are saying here. So if I buy RAM rated for XMP 3000 that it's overclocked when running at that frequency and those timings since it's running faster than 2133? Are you saying all DDR4 RAM sold at higher than 2133 is overclocked?

 

[wingman99]

Not sure what you are saying here. So if I buy RAM rated for XMP 3000 that it's overclocked when running at that frequency and those timings since it's running faster than 2133? Are you saying all DDR4 RAM sold at higher than 2133 is overclocked?

This is from Gigabyte

Support for DDR4 3200(O.C.) / 2933(O.C.) / 2667* / 2400 / 2133 MHz memory modules
* Support for higher than DDR4 2667 MHz may vary by CPU. https://www.gigabyte.com/Motherboard/GA-AX370-Gaming-K5-rev-10#sp

For AM4 anything over 2667 is overclocking (O.C.) memory.

 

[trents]

I've always understood that to mean overclocked from the board's point of reference, not from the RAM's perspective. I could be wrong but it seems to me that different motherboards will insert that "(O.C.)" note at different frequencies.

Besides, it doesn't stack up to what you said earlier in post #17 where you state that "Ram sold at higher than 2133 is overclocked." In the gigabyte motherboard example you give there are two steps after 2133 that don't have "(O.C.)."

For instance: https://www.newegg.com/Product/Product.aspx?Item=N82E16813132927

It's a Z270 board but even declares DDR4 2133 to be an "O.C." frequency.

Here's another: https://www.newegg.com/Product/Product.aspx?Item=N82E16813157746 that doesn't use append the "O.C." note until you get to 2800.

When I hear the term "overclock" I think of pushing something faster than the speed it was sold to run at.