This morning AMD is introducing their Ryzen PRO processors for business and commercial desktop PCs. The new lineup of CPUs includes the Ryzen 3 PRO, Ryzen 5 PRO and Ryzen 7 PRO families with four, six, or eight cores running at various frequencies. A superset to the standard Ryzen chips, the PRO chips have the same feature set as other Ryzen devices, but also offer enhanced security, 24 months availability, a longer warranty and promise to feature better chip quality.
I guess it makes sense from a marketing perspective, but I’m not a fan of segmentation like this – it just makes an already complicated market even more complicated.
The link is broken.
More choice is good. There will never be one CPU to fit all budgets and workloads, so variety can allow people to build the exact machine to cater for their needs.
Huh? What is this supposed to mean? Does it mean that regular AMD Ryzens will overheat due to leaking or go poof after a couple of years, or does it mean that the “Pro” Ryzens will last a century?
This “promise” might hurt the the regular AMD Ryzen line because AMD are sending the wrong message.
Edited 2017-06-30 09:11 UTC
These chips effectively have extended warrenty/support contracts on their “PRO” chips. There is also a guarentee that chip will be made for at least 2 years.
Often how the extended warrenty is acheived is not to allow overclocking.
This is where you must chose what you value. Is a slower ship that will last longer more important to you or one that is overclockable?
I can guarentee you Dell will want to know they would have a supply of the same chip for the life of the model. Especially in buisness models, they want those things stable as any issues cost them in support call outs
Probably just that the chips have another kind of binning (sorting of processor chips according to physical variation). For instance the chips may tolerate higher temperatures than normal desktop chips, have a reduced clock frequency compared to what the chip can actually do (further from the margins – more tolerance) etc.
This is exactly it. The chips have tighter tolerances then the regular Ryzen chips, everything works on these chips, or they don’t tolerate overclocking that much. It’s the same difference between an i7 and the higher end Xeon E3 chips.
It’s really a winning situation. AMD moves chips, and people get what they want. I’d be interested in the SME and SEV security features, but I’m not interested in overclocking. This saves me some bucks by not having to buy a full Epyc kit, which I would have to do otherwise. However, someone else may be the opposite, so this is a nice blend of features.
Just suppose that AMD realized that they have a winner, the Ryzen family of CPU’s, but the announced retail prices were set too low.
By providing a different CPU ID, with perhaps an additional instruction, you can legally add $100 or whatever to the existing CPU price.
I am not a gamer, and I would want the 7100 for regular business — home use. If the new cpu is really better, then I will consider going for it.
That is my only question.
It’ll be up to the board manufacturers.
And it’s worth noting that the existing Ryzen CPU’s (at least the Ryzen 7’s, possibly 5’s and 3’s, but not sure about those) already support ECC, you just can’t find any motherboards that support it. Hopefully Ryzen PRO will change that in reasonably short order and the OEM’s won’t be jerks and make boards that only support ECC with processors that identify as PRO models.
I am fairly sure that my ECC is working. I have an ASUS Prime X370-PRO with a Ryzen 1700X and ECC RAM installed. The dmesg reports:
[ 5.874157] EDAC MC: Ver: 3.0.0
[ 5.877166] EDAC amd64: Node 0: DRAM ECC enabled.
[ 5.877167] EDAC amd64: F17h detected (node 0).
[ 5.877203] EDAC MC: UMC0 chip selects:
[ 5.877204] EDAC amd64: MC: 0: 0MB 1: 0MB
[ 5.877204] EDAC amd64: MC: 2: 16383MB 3: 16383MB
[ 5.877205] EDAC amd64: MC: 4: 0MB 5: 0MB
[ 5.877205] EDAC amd64: MC: 6: 0MB 7: 0MB
[ 5.877207] EDAC MC: UMC1 chip selects:
[ 5.877208] EDAC amd64: MC: 0: 0MB 1: 0MB
[ 5.877209] EDAC amd64: MC: 2: 16383MB 3: 16383MB
[ 5.877210] EDAC amd64: MC: 4: 0MB 5: 0MB
[ 5.877210] EDAC amd64: MC: 6: 0MB 7: 0MB
[ 5.877211] EDAC amd64: using x8 syndromes.
[ 5.877211] EDAC amd64: MCT channel count: 2
The UEFI does not have options to select ECC Scrub or continue / halt on error, but there are supposed to be ways to set that from within Linux, although I haven’t yet.
Ah, looks like scrubbing was already on.
# cat /sys/devices/system/edac/mc/mc0/sdram_scrub_rate
390720
Edited 2017-07-01 02:39 UTC
Other than on servers, is ECC ram really necessary?
I am running 64gb machine with an i7, and other than my work HP Z820 workstation has more threads the machines are more or the less the same in reliability.
But I am a developer
Edited 2017-07-01 09:16 UTC
Now that you can have ECC without the Intel tax, why not use it?
The RAM is a little more expensive but much more reliable.
The Rizen 1600 could be underpowered to 25W, according to the tables.
[Could easily turn off the ventilator, if big dice].
Edited 2017-06-30 15:50 UTC
I wouldn’t count on that, because even my 18w AMD E350 needs one running. But I recognize this is good news since the 1600 might offer more performance for a quasi similar TDP.
Maybe if you had a fairly large CPU heatsink and a case fan or three running.
I had a dual core Bobcat in a NAS box for several years and it didn’t need a fan running at 800 MHz but as soon as it spooled up to 1.6 GHz it would overheat without a fan. I know because I had to replace the fan last year.
And that Bobcat was about 8 W maximum, if you weren’t running the GPU.
HUGE Thermaltake aluminum carved block, mirror finish cooper interface [clippers so strong though would break the MB], black edition CPU, The highest clocked ones have the better chances to heavily under-volt. Silent beauty. Needed a very small cooper cooler on the South Bridge. One of my early always-on.
Typing from a E-350, by the way
Edited 2017-07-03 20:38 UTC
That beast still working, I believe. On SSD storage.
It makes sense from a consumer and manufacturing perspective.
Instead of sitting on a bunch of tweener chips, they go ahead and release them under the “Pro” banner. Epyc is a package of 4 8-core chips, and not all of the 8 core chips are going to meet the standards needed to be included in an Epyc chip.
Thus people who need, or would like, the features of the server chips can get them at a lower cost, which helps AMD in the server space.
IC manufacturers have done this in one fashion or another since forever. Remember when Motorola sold the 68040 with a broken FPU? They called it the 68LC040, and added the caveat to the hardware manual that executing an FP instruction would do unknown operations for an unspecified period of time. They also sold the 68040 with a broken MMU as the 68EC040 with the same warning about MMU instructions. Some people deliberately bought EC chips and then tested to see how bad the break was, returning and getting another until they found one that worked “well enough” for their purposes.
I don’t actually remember that chip. It was before my time.
We’re on the same page. Binning and nerfing is something that has gone on for a long time, and it just part of the game.
Like i7 laptop chips. They’re not that different from i5 chips, but they’re more power efficient and clock higher.
I will probably be buying Ryzen 1800X when there are better motherboards on the market as my i7 machine just isn’t up to the task anymore as my main work machine.
Epyc is there to go toe to toe against Xeon i9. AMD have really got Intel on the back foo as the i9 series looks like rebadged Xeons.
In gaming benchmarks Intel i7 processors are still top dog. But developers like myself who don’t need a Xeon rig but does need more Umph the new AMD lineup is pretty interested tbh.