It really takes the budget out of budget to be honest... With current ram prices it´s not even logical to budget build with new parts because of it.
Thats is really a problem for budget builds, then ill like to see a 2200G with dual DDR4-2400 without the power limits. Ill probably do this myself by friday or so.
Maybe you could wait a little bit and see if W7 drivers are released later?surprised by no win7 drivers. even more so considering people running igpu is more likely to use it as an office computer where its also likely its running 7 and not 10.
and if you are not interested in the gpu part, but still want a slightly updated ryzen architecture you must go win10.
how come? same chipsets.
It really takes the budget out of budget to be honest... With current ram prices it´s not even logical to budget build with new parts because of it.
I assume some sarcasm there.
Since, unlike GPU's, they are limited to one per system. Should keep them safe.
Looking a minute ago on Amazon, the 2400 is already sold out. We don't need no stinkin miners to wipe out supply.
Well then i dont understand how the stock cooler can handle it, unless the OC tests i seen werent using the stock cooler. ill need to double check.
There are several 4Gb ICs available, but none of them can even come close to the characteristics of the B-die 8Gb IC.
Samsung D & E-die is probably the best bet, however new modules based on them are becoming rare. And they're not cheap either, by any means.
Why wouldn't they be able to handle it?
The coolers are rated by AMD for a specific power dissipation, but it doesn't mean that it's the absolute maximum figure it can dissipate by any means.
The only thing what happens if you exceed the actual power dissipation capability of the cooler is increased temperatures or the CPU.
Also neither of the existing Zen designs (Zeppelin or Raven) is sensitive to temperatures. Their rated operating temperatures are up to 100°C
Ill need to how it performs with DDR4-2400 ram and no power limits first
Multi-socket motherboards require more than slapping two or more sockets on the PCB. There needs to be an interconnect between the CPUs themselves. Only AMD can allow such a design to exist.We know that those Vega GPU cores are good for crypto mining, although these APU's don't really have enough of them to make it worth while.
I wouldn't be too surprised if someone did a crazy quad socket "mining" motherboard for these APU's if they ever get powerful enough to mine crypto.
We know that those Vega GPU cores are good for crypto mining, although these APU's don't really have enough of them to make it worth while.
I wouldn't be too surprised if someone did a crazy quad socket "mining" motherboard for these APU's if they ever get powerful enough to mine crypto.
I understand that you are aiming for the best memory- but for buyers of these APUs- best memory is the very cheapest DDR4 2400 that overclocks well, preferably to 3200 Here, for example, russian techtuber is recommending these 'green bare samsung sticks' which he says reliably OC all the way. It would be great to know such memory in 4GB sticks exists.
Two words for you: Google Translate.It will most likely not affect daily usage scenarios but gaming is affected 100%... Check out this Russian guy (don´t know if he used dual or just one channel 8GB 3200)...
During the test, the number of memory channels was not taken into account. Indeed, 8 GB was achieved by 1 module. This leads to such significant differences. Now I re-conducted the test with one module for 16 gigabytes and got the same results as with the module at 8.
It will most likely not affect daily usage scenarios but gaming is affected 100%... Check out this Russian guy (don´t know if he used dual or just one channel 8GB 3200)...
Performance takes a massive kick in the nuts if you ask me...
Friends. Thank you for pointing out the mistake in the memory test, which really does take place. I apologize, the rush, as you see does not lead to anything good. During the test, the number of memory channels was not taken into account. Indeed, 8 GB was achieved by 1 module. This leads to such significant differences. Now I re-conducted the test with one module for 16 gigabytes and got the same results as with the module at 8. The total conclusion about the memory is not correct. Not the volume in this case affected the performance, but the number of channels involved.
Shouldn't Hynix ICs be able to do 2933CL16 these days with updated BIOS?
Two words for you: Google Translate.
If you translate the description, it makes it clear that the 8 GB tests are with a single stick (single channel) which is why it effects the performance so much. Still have yet to see any results showing 8 GB of dual channel memory will have any noticeable effect on performance for the settings the APU will need to use to game on.
Samsung has 4GB D & E-die based OE sticks (green) available, however they are already phased out.
The are not cheap either, 70 - 80€ per module in Europe.
M378A5143DB0-CPB - D-die
M378A5143EB1-CPB - E-die
RAM prices are the biggest issue indeed with these APUs.
Standard 2666MHz 16GB kit can be had for ~150€, however the cheapest B-die based kit (which is both fast and compatible with Ryzens) costs 197€.
Frankly I don't think 6-7GB of RAM (usable after UMA allocation) is even remotely enough with modern standards.
16GB is IMO the bare minimum and personally I've been using 64GB for years.
That's because with the 65W power limit ceiling the APU cannot sustain it's maximum default clocks, when both domains are simultaneously stressed.
HUH so there´s almost no dual channel 8Gb kits available which work with Ryzen? You mean compatible with these APU´s or what?
thats a really bad use of turbos if you drop the clock when there is heatroom, unless they are worried about blowing up motherboards and not heat. i thought XFR would have come to the rescue here.
If 16gb is the bare minimum, how much performance improvement would you see in games going to 32gb that would make it worth double the (very high at the moment) cost?
Are you suggesting that AMD should violate the power limits (i.e. the advertized TDP)?
There is plenty of margin in the temperatures but none in the power limits. That's the only limiting factor for the frequencies at stock.
XFR is just some predetermined extra boost. SenseMI is akin to GPUBoost in behavior, ie. it's temperature sensitive and changes clocks in real-time.I was under the impression that was what XFR/SenseMI does. Boost as long the thermals permit, it really needs to be good thermals in order to drop power usage.
Unless... this only include CPU clocks and no IGP clocks? or CPU clock have more priority over the GPU clock, this is what could be happening at stock, even at base clock the CPU perf should be more than plenty.
You can check that? because if CPU clocks have more priority over the IGP clock at stock that should be reported to AMD because it has to be a mistake.