|
Is there a point to a 16 phase board like the upcoming MSI MEG, if you don't actively overclock? Everyone's saying that the CPU tries to get the maximum out of itself on its own, would anything beyond the usual 8 phases of the other crop of boards make a difference? The thing is going to run a 2950X, and likely a 3950X going forward, if they bump the core count per CCX.
|
#
?
Aug 6, 2018 21:20
|
|
|
|
| # ? May 17, 2024 17:06 |
|
|
TheJeffers posted:XFR 2 is more about boosting sustained performance in multi-threaded workloads when you have high-quality cooling on the chip and/or favorable ambient temperatures. What exactly is the distinction between Boost, XFR1, XFR2, and PBO? It seems like all of them more or less fall into the category that Intel would define as "turbo". Not even Wikichip seems to have a good distinction of what they are.
|
|
#
?
Aug 6, 2018 21:31
|
|
|
Cygni posted:The Threadripper embargo is "unboxing only" to drive pre-orders for a August 13th release. So even though people have them in hand, they aren't allowed to post numbers before retail availability. Trash, AMD. We are having pre-orders for chips that we aren't allowing anyone to benchmark 
|
|
#
?
Aug 6, 2018 21:37
|
|
|
XFR2 is about the boost clocks OCing to meet the thermal headroom of the cooler. An 2000X chip with XFR2 is going to achieve higher boosts on a 280mm AIO or open loop than it would a 120mm AIO, which will get higher boost than a skimpy air cooler. PBO is about core use VS power draw, and can up vcore into unsupported out-of-spec “don’t come crying to us” levels. Your average user can get basically all the OC headroom they used to have to endlessly test for simply by enabling XFR2 and throwing on a big cooler. Your average user shouldn’t touch PBO. Most mainstream reviews of 2000X chips with X470/B450 (both an X chip and a 400 board are necessary for XFR2) go “yeah, there’s like no joy in DIY OC for most people now because this thing is so good at hitting the sweet spot for your cooler and going no further.” Craptacular! fucked around with this message at 21:41 on Aug 6, 2018 |
|
#
?
Aug 6, 2018 21:38
|
|
|
Combat Pretzel posted:Is there a point to a 16 phase board like the upcoming MSI MEG, if you don't actively overclock? Everyone's saying that the CPU tries to get the maximum out of itself on its own, would anything beyond the usual 8 phases of the other crop of boards make a difference? The thing is going to run a 2950X, and likely a 3950X going forward, if they bump the core count per CCX. More real phases = better voltage stability. Voltage stability is always nice whether you're overclocking or not, it's good for component lifespan as well as overclocking. Extreme overclockers need voltage stability because they're also overvolting the CPU to an inch of it's life. So the voltage needs to stay rock-solid when it's already way out of spec. For just XFR or other basic overclocking that's not a problem, the CPU draws spec volts, the spikes and droops stay in spec, and the capacitors don't get overwhelmed by twice as much power draw as they were designed for. More VRM parts = still worth something, as it distributes load across more parts and means mosfets are operating more efficiently. This has become a bigger issue with the popularity of AIO watercooling since the VRM heatsinks may not get much airflow. That MSI MEG is using a controller with 8 phases, and IDK whether the phases are really doubled or just stacked 2 per phase. Regardless, if you're not planning to do extreme overclocking it's way more than sufficient. That thing was designed with the new WX 24 & 32 core parts in mind that might draw 300w if you have a huge water cooler. If you're just going to use a 2950X and are positive that you will be sticking with a hypothetical 3950 or whatever other sub-200W part* in the future, you can totally get away with one of the "first wave" x399 boards. The MSI MEG's other gimmick is NVMe slots if you want a ton of those. *edit: by "official TDP" numbers Klyith fucked around with this message at 01:03 on Aug 7, 2018 |
|
#
?
Aug 7, 2018 00:52
|
|
|
Combat Pretzel posted:Boost clock ain't all-core on Ryzen, right? Paul MaudDib posted:What exactly is the distinction between Boost, XFR1, XFR2, and PBO? It seems like all of them more or less fall into the category that Intel would define as "turbo". I'm still kind of confused by it myself, but I can tell you how the 1x00 chips behave vs. how the 2x00 chips behave. I had a 1700, and still have a 1600. Their base clocks were 3.0GHz and 3.2GHz, respectively. Given sufficient thermal overhead (which they had), both ran at their "all cores" boost clocks of 3.2 and 3.4GHz, 100% of the time, while being at 100% utilization. And while there would be variance in speed between cores, it was usually <1MHz. The 2x00 chips don't have an all-cores boost clock; just base and max. I upgraded the 1700 to a 2700X last week, and its behavior has been...unexpected. For the first 24 hours it ran right at its base clock of 3.7GHz. Like plus or minus 25MHz (and sometimes it was minus). I was checking so obsessively because I also switched from air cooling to water cooling, and I was really interested to see how effective that was. Over the past 4 days, the clocks have slowly been drifting upward. First I noticed that it was reliably above 3.7GHz. Then it was almost 3.8GHz. Then right around 3.8GHz. And about 20 minutes ago I noticed it cracking 3.9GHz for the first time. Per-core speeds vary far more considerably than in a 1x00 chip as well. code:mdxi fucked around with this message at 03:24 on Aug 7, 2018 |
|
#
?
Aug 7, 2018 03:22
|
|
|
Klyith posted:More real phases = better voltage stability. With the existing ones, it's always something. Either not 100% clear it supports ECC (due to lack of user testing), there being a Passmark difference (which varies from 24K-28K) or being plain Gigabyte.
|
|
#
?
Aug 7, 2018 04:10
|
|
|
mdxi posted:I'm still kind of confused by it myself, but I can tell you how the 1x00 chips behave vs. how the 2x00 chips behave. So it's learning how best to manage it's own resources and power to achieve maximum boost clocks on it's own? I thought this was kind of established but not sure.
|
|
#
?
Aug 7, 2018 08:45
|
|
|
Is it the paste bedding in, or differences in ambient temps?
|
|
#
?
Aug 7, 2018 08:55
|
|
|
EmpyreanFlux posted:So it's learning how best to manage it's own resources and power to achieve maximum boost clocks on it's own? I thought this was kind of established but not sure. Gonna be hell for the benchmarkers if they gotta wait weeks for the CPU to figure out where it can settle. It'd also be hilarious if a short heat wave spikes the ambient temperature long enough to reset the resting point so you have to wait for clocks to climb again.
|
|
#
?
Aug 7, 2018 09:14
|
|
|
Color me extremely skeptical. My anecdote is my 2600x on a giant air cooler immediately saw, and continues to see 4250 MHz in single and double core activity. Sits around 4000mhz well running all out, dropping roughly linearly as temperatures increase.
|
|
#
?
Aug 7, 2018 10:22
|
|
|
GRINDCORE MEGGIDO posted:Is it the paste bedding in, or differences in ambient temps? Gotta be one of those, or something in the water cooler like particulate being washed off the block microfins. Some environmental factor outside the chip anyways. XFR2 is smart but it's not learning over time smart.
|
|
#
?
Aug 7, 2018 14:23
|
|
|
EmpyreanFlux posted:So it's learning how best to manage it's own resources and power to achieve maximum boost clocks on it's own?
|
|
#
?
Aug 7, 2018 14:33
|
|
|
Klyith posted:Gotta be one of those, or something in the water cooler like particulate being washed off the block microfins. Some environmental factor outside the chip anyways. I must be combining this behavior with what I guess is the smart micro op cache function? lllllllllllllllllll posted:This one takes weeks, the next gen will do this at a geometric rate. We as humanity 100% deserve whatever the collective of 3990WXs do to us
|
|
#
?
Aug 7, 2018 15:01
|
|
|
I acknowledge that my clock checking has not been automated, or regular; it's just been me wondering "hmmm, what's it up to now?" It's an anecdote, and checking right now shows clocks around 3.77GHz, so it is fluctuating. I tend to agree with the idea that it's small environmental factors. The temperature in my apartment is decently stable, but not perfectly so. The thing I'm surprised by is that there's so much thermal overhead available with this water cooler, but boost clocks on the 2700X have been relatively low. Under air cooling the 1700 was happily chugging along at all-cores boost speeds while running at 68C. The 2700X meanwhile, appears to be much more conservative with boost clocks when running at 55-57C.
|
|
#
?
Aug 7, 2018 15:48
|
|
|
Craptacular! posted:XFR2 is about the boost clocks OCing to meet the thermal headroom of the cooler. An 2000X chip with XFR2 is going to achieve higher boosts on a 280mm AIO or open loop than it would a 120mm AIO, which will get higher boost than a skimpy air cooler. loving finally OC is a way to get people to pay for a warranty but not have to pay out. It always should have been automated based on sensor feedback and AMD finally did it.
|
|
#
?
Aug 7, 2018 15:57
|
|
|
Cygni posted:The Threadripper embargo is "unboxing only" to drive pre-orders for a August 13th release. So even though people have them in hand, they aren't allowed to post numbers before retail availability. Trash, AMD. The x399 Zenith Extreme goes against AMD's guidelines by placing the first PCIE slot close to the CPU socket. AFAIK, the Zenith Extreme is the only threadripper motherboard that has CPU-fan clearance issues.
|
|
#
?
Aug 7, 2018 18:01
|
|
|
At least Threadripper has the lanes to give you a second x16 slot if the first one is blocked! 
|
|
#
?
Aug 7, 2018 19:00
|
|
|
Malcolm XML posted:loving finally
|
|
#
?
Aug 7, 2018 20:05
|
|
|
A ton of Threadripper slides leaked. The one that AMD probably wants people to see most: https://cdn.videocardz.com/1/2018/08/AMD-Ryzen-Threadripper-2000-1-2.jpg edit: Bugger, just checked it on my phone, and no timg. Changing to url. SwissArmyDruid fucked around with this message at 02:58 on Aug 8, 2018 |
|
#
?
Aug 8, 2018 01:49
|
|
|
 
|
|
#
?
Aug 8, 2018 01:59
|
|
|
should have gone with "render fast die young" imo
|
|
#
?
Aug 8, 2018 02:10
|
|
|
If I'm going to stick a TR2 in one of the existing mainboards, pre-updated BIOS, will it boot but just not support all the boost features and poo poo, or not boot at all? I'm wondering because the X399 Taichi I'm looking at might require two updates in a row (something something bridge BIOS). While it can flash a BIOS without a CPU installed, needing a "bridge BIOS" kind of implies that it needs to boot it at least once to do its bridging poo poo.
|
|
#
?
Aug 8, 2018 21:07
|
|
|
I assume it'll be ok after all the updates as long as you don't try the 32c version.
|
|
#
?
Aug 9, 2018 01:52
|
|
|
Combat Pretzel posted:If I'm going to stick a TR2 in one of the existing mainboards, pre-updated BIOS, will it boot but just not support all the boost features and poo poo, or not boot at all? I'm wondering because the X399 Taichi I'm looking at might require two updates in a row (something something bridge BIOS). While it can flash a BIOS without a CPU installed, needing a "bridge BIOS" kind of implies that it needs to boot it at least once to do its bridging poo poo. [edit] the VRM cooler on that board sucks. I'd probably try and cool it with an aftermarket thing. 11 Phases is super dreamy though... if it is really 11. redeyes fucked around with this message at 02:02 on Aug 9, 2018 |
|
#
?
Aug 9, 2018 01:52
|
|
|
R3 2300X and R5 2500X specs leaked: https://www.anandtech.com/show/13178/lenovo-posts-specs-of-amds-ryzen-3-2300x-ryzen-5-2500x 
|
|
#
?
Aug 9, 2018 02:38
|
|
|
SwissArmyDruid posted:R3 2300X and R5 2500X specs leaked: With the 2600 dropping to $160 on sales I doubt anyone is gonna put those two high in the care meter.
|
|
#
?
Aug 9, 2018 03:07
|
|
|
I really want to try that 32core and Zephyr. I'm also glad marketing ran spellcheck on the slides  
|
|
#
?
Aug 9, 2018 03:15
|
|
|
Palladium posted:With the 2600 dropping to $160 on sales I doubt anyone is gonna put those two high in the care meter. For terrible enthusiasts like us, sure. On the global scale though, when you factor in much more price sensitive segments like gaming cafes, emerging markets, and prebuilts, the ~$100 market is incredibly important for AMD and has been since the K6 era.
|
|
#
?
Aug 9, 2018 03:27
|
|
|
I wonder how the 2200g will stack up against the 2300X since they are both 4 core 4 threads but the 2200g is normally $99. I guess maybe the 2300X could sit below that MSRP of $125 and maybe the slightly higher clocks will be worth it. 2300X - $110 2500X - $135 2600 - $160 That pricing makes a bit more sense to me, like others said with the 2600 at $160 a $150 2500X doesn't make a lot of sense.
|
|
#
?
Aug 9, 2018 03:35
|
|
|
redeyes posted:I'm pretty sure you can do the BIOS flashback without even having a CPU installed in the thing. I've done it on other Taichi boards, but they were Intel ones. redeyes posted:[edit] the VRM cooler on that board sucks. I'd probably try and cool it with an aftermarket thing. 11 Phases is super dreamy though... if it is really 11. Combat Pretzel fucked around with this message at 04:07 on Aug 9, 2018 |
|
#
?
Aug 9, 2018 04:04
|
|
|
As far as I can tell the bridge BIOS thing just means you flash that 'bridge' BIOS first. Let the board power up and throw whatever boot codes. Shut down and then flash the latest BIOS. I can't find anyone talking about needing a CPU to do it. I do know after BIOS flashback boot with no CPU, the board does a 2 or 3 reboot sequence and then idles at a no CPU code. It might be writing UEFI variable stuff during this time. At least that is my theory.
redeyes fucked around with this message at 04:43 on Aug 9, 2018 |
|
#
?
Aug 9, 2018 04:40
|
|
|
I guess I'll take my chances with the ASRock then. I prefer explicit than implicit ECC support.
|
|
#
?
Aug 9, 2018 13:35
|
|
|
Combat Pretzel posted:I guess I'll take my chances with the ASRock then. I prefer explicit than implicit ECC support. You could always email support and ask directly. They have decent support.
|
|
#
?
Aug 9, 2018 15:45
|
|
|
Combat Pretzel posted:I guess I'll take my chances with the ASRock then. I prefer explicit than implicit ECC support. Asrock Taichi x399 supports ECC.
|
|
#
?
Aug 9, 2018 17:14
|
|
|
Yeah I know. The taking chances was more about the CPU-less double flashing. The Taichi seems to be the only board explicitely supporting it, i.e. mentioning it as big bullet point on the product page and there's actually memory scrubbing and data poisoning options relating to ECC in the BIOS, whereas with the other boards, you're at the mercy of there being the necessary traces and the OEM not forcefully disabling it.
|
|
#
?
Aug 9, 2018 17:59
|
|
|
Methylethylaldehyde posted:Asrock Taichi x399 supports ECC. Does the X399M too? Same/similar BIOS? I know the full-ATX boards make more sense for a processor like that, but Asrock's mini-boards are just so cool 
Paul MaudDib fucked around with this message at 19:10 on Aug 9, 2018 |
|
#
?
Aug 9, 2018 18:40
|
|
|
Microcenter has 1950Xs down to sub $600. If you are buying a many core CPU yourself and not on the corporate dime, thats a drat good price and probably a better value than a 2950X at $900.
|
|
#
?
Aug 10, 2018 17:07
|
|
|
x86 CPUs: The Intel/AMD Thread It's been discussed before--and if memory serves me right--with positive responses. Nobody pulled the trigger. If there are no serious objections, at 00:00 UTC Sunday 12 August 2018 (8 PM Saturday EST), AMD will acquire Intel (or something similarly improbable and official sounding) and I'll petition a mod to close the separate CPU threads.
|
|
#
?
Aug 10, 2018 20:21
|
|
|
|
| # ? May 17, 2024 17:06 |
|
|
What would be gained with such a move?
|
|
#
?
Aug 10, 2018 20:22
|
|
