|
I mainly use PCPartsPicker to get a general idea of system cost and put together a parts list. It's less important to buy from the retailers they link to.
|
#
?
Oct 28, 2022 23:04
|
|
|
|
| # ? May 30, 2024 03:12 |
|
|
DerBauer's got a video up where he delidded a 7600x and found it with two CCDs, one inactive. So there might be some interesting binning decisions happening. He also just straight yanks a heatsink off a delidded chip during operation to test the thermal shutoff capability, and it apparently lived, after sitting at 110 C for a few seconds and shutting itself down. The video seems like it was initially intended mostly as a product ad for his delidding tool, which he is now confident won't kill your processor. I'll give him credit for again being pretty up front about his earlier prototype killing a 7700x, although I dunno that a sample size of 13 or 14 safely delidded processors would leave me feeling very comfortable opening up my own hypothetical 7900 or 7950. eviltastic fucked around with this message at 00:01 on Oct 29, 2022 |
|
#
?
Oct 28, 2022 23:55
|
|
|
one thing i learned from ifixit (the hard way) is that having great tools and materials only helps so much if the actual repair guide is mediocre derbauer's delidding tool kinda feels the same way
|
|
#
?
Oct 28, 2022 23:59
|
|
|
eviltastic posted:DerBauer's got a video up where he delidded a 7600x and found it with two CCDs, one inactive. So there might be some interesting binning decisions happening. Bizarre that the 7700X only had 1 core chiplet, so somethin funky goin on. Like you pointed out, he also confirmed that the 2nd die really is a dummy, so this isn't a Ryzen 3 3100 situation. AMD has used dummy dies before in first and 2nd gen Threadripper/EPYC due to mounting concerns, but they were consistently there as far as I know. Weird!
|
|
#
?
Oct 29, 2022 00:12
|
|
|
mdxi posted:Oh man. Tiger Direct was already barrel-scrapingly bad in the days when Gateway 2000 was a low-budget direct operation whose entire marketing budget was several pages of ads in Computer Shopper (which I've just learned, somehow struggled on until 2009 -- I don't think I ever saw an issue after about 1998). Ironically in the end Tiger Direct survived and dropped the rebate game and pivoted into a legit b2b-facing tech store, then went out of business anyway due to their reputation. I remember them imploding in 2014-2015 and they had some legit deals as they were starting to cash out and face to bankruptcy. They recently popped back up into business again but no idea if it's the same owner or what. My first new case since my childhood antec clone was an NZXT S340 (the og not the elite - with the acrylic side panel) with a couple orange fans (cougar?) that TD shipped for $60. That's the cost-reduced version of the H440 (I think?), but it was still the single biggest leap in build quality I've ever done, vs chinesium 2000s mid-tower stuff. NZXT actually had thoughtful design and good build quality vs just pressed sheet metal inside etc, and only 60 dollars. The fans are a nice throw-in even if you value them at $0, that's TD dumping their inventory. Today a Fractal is nicer and more flexible (and solid panel is usually an option) but they're also proportionately much more expensive, $60 for the S340 was a deal, and even at non-dealz pricing it was very no-nonsense, like street price was well sub-$100. I also made the greatest mistake of my enthusiast life, I got in on a clearance (but new) 295x2 for $600 from TigerDirect in 2015. Changed my mind, returned to sender unopened, refund done no problems. 
Paul MaudDib fucked around with this message at 20:24 on Oct 29, 2022 |
|
#
?
Oct 29, 2022 04:35
|
|
|
I've bought a bunch of stuff over the years from outfits like Best Buy, B&H photo, Micro Center, Tiger Direct I even bought the Ampere A6000 from scan.co.uk due to it launching there earlier I can't recall the last time I had an issue beyond courier delays Lol just reminded that I managed to get a PS4 at launch from Amazon and had it shipped to Tokyo because I was on an extended business trip there at the time. Good times shrike82 fucked around with this message at 04:43 on Oct 29, 2022 |
|
#
?
Oct 29, 2022 04:39
|
|
|
Tiger Direct always felt spammy to me, like a lovely version of Fry’s or Microcenter. I did buy a couple smaller things from them in 2013-14 when they had solid deals, namely my G502 mouse, a Bluray driver, and an external HDD that I shucked. The rebate process was scammy af though. RIP Fry’s, long life Microcenter.
|
|
#
?
Oct 29, 2022 18:38
|
|
|
kliras posted:one thing i learned from ifixit (the hard way) is that having great tools and materials only helps so much if the actual repair guide is mediocre pretty much no electronic product are worth repairing on your own now unless you do it all the time
|
|
#
?
Oct 30, 2022 04:41
|
|
|
Cygni posted:Bizarre that the 7700X only had 1 core chiplet, so somethin funky goin on. Like you pointed out, he also confirmed that the 2nd die really is a dummy, so this isn't a Ryzen 3 3100 situation. AMD has used dummy dies before in first and 2nd gen Threadripper/EPYC due to mounting concerns, but they were consistently there as far as I know. Weird! They also did this with ryzen 5000. My 5600x has 2 ccds with only one of them working. I don’t think these are dummys, just disabled. Maybe the chiplets were fine on their own, but got damaged when they assembled the package?
|
|
#
?
Oct 30, 2022 13:10
|
|
|
Tiger Direct had a physical store where I lived at the time that was good initially but quickly went down hill. The last time I ever went there I got my parts, went to the register and found out that everything was 20-30% more than the listed price on the website. I had all of the sku’s pulled up and asked them to price match the site. I was told that they had nothing to do with the tiger direct website and don’t price match them. Ladies and gentlemen, I present to you the physical storefront that had NOTHING to do with Tiger Direct Dot Com and therefore will not price match:  That was the day I started using Newegg.
|
|
#
?
Oct 30, 2022 15:17
|
|
|
Palladium posted:pretty much no electronic product are worth repairing on your own now unless you do it all the time Steam Deck is a notable exception to this, but that’s probably because Valve worked with iFixit on the guides and provide them with official parts.
|
|
#
?
Oct 30, 2022 16:47
|
|
|
the battery is a huge pain in the rear end to replace on the deck so 
|
|
#
?
Oct 30, 2022 16:48
|
|
|
spckr posted:They also did this with ryzen 5000. My 5600x has 2 ccds with only one of them working. I don’t think these are dummys, just disabled. Maybe the chiplets were fine on their own, but got damaged when they assembled the package? drat i knew two die 5600Xs existed but i thought they were ultra rare, because there shouldn't really be many/any actual failures by that point. The dies have been repeatedly checked by that point and the packaging process itself is an automated, low risk/high volume sort of operation. Especially without things like fan out bridges. So if it really is another actual die in the 7600Xs, we may be looking at a pure binning/stack engineering reasons for the decision.
|
|
#
?
Oct 30, 2022 21:31
|
|
|
For the 7600x I guess there may have been a defect found in their packaging solution or the new IO die that only became apparent late. A 2CCD 5600x one sure sounds like a 5900x that they rebadged.
|
|
#
?
Oct 30, 2022 22:26
|
|
|
you really need more than n=1 sample size to have any idea what's going on with the dual-CCD 5600Xs. one possibility is that it's a "structural blank" like threadripper had (although iirc they were later found to be actual full dies anyway). In that case it would be present in every single 5600X because it's part of the design. if it's a packaging failure then presumably the number of failures would be relatively low. You bin before you package, that's why AMD knows what's a high-bin CCD and what's a low-bin CCD, so they know in advance, so you only have the handful of chips that actually fail as you're putting them together, which is very low these days, probably <1%. if you're seeing it on lots of chips, but not all chips, then it's due to sales. AMD is struggling to sell 7900X's and 7950X's with Zen4 3D still obviously coming in a few months, and they don't want to drop the prices and ruin their margins (which will eventually percolate across the whole market, and perhaps even ruin their intended pricing structure for Zen4D and force them to launch much cheaper than they wanted to to have the new SKUs make pricing sense). The only Zen4 SKU that is selling any appreciable numbers right now is 7600X. So what do you do? You keep prices high and turn 7950Xs and 7900X's into 7600Xs and ship them. If that's the case it may not last either, they cut Zen4 production and if they are turning existing Zen4 chips into lower SKUs they'll eventually burn through their stockpile and go back to just single-CCD packages once they resume production. This is sort of the German Tank Problem, or QC sampling more generally. It's hard to tell what's going on from n=1 but statistically speaking the odds of Der8auer just randomly happening into a chip with the the 1% of package failures is pretty low. If you're finding it repeatedly in small samples, then (just like with the NVIDIA power cables) odds are the % is relatively large in the general public. Paul MaudDib fucked around with this message at 23:16 on Oct 30, 2022 |
|
#
?
Oct 30, 2022 23:11
|
|
|
Finally got my 7700X and 4090 build together, went with the MSI 650M Mortar Wi-Fi and some ddr5 6000 CL32. Been playing with the negative curve optimizer and PPT a bit. Also noticed my 4090 was scoring near the bottom of every synthetic benchmark, usually 20-30 percent slower. Finally figure out with GPU-Z that it was stuck at PCIE 16x1.1 and nothing I did in windows or the bios fixed. Thankfully MSI posted a new bios in the last couple of days and updating to that fixed the issue, scores seem to be normal for a stock card now. So definitely some teething issues to be worked out. Oh and boot times are definitely longer than previous builds. There’s about 15 seconds of black screen but once the Msi logo appears it gets into windows quickly. Also the iGPU crashes when trying to watch twitch with hardware acceleration is on thigh YouTube was fine. B-Mac fucked around with this message at 00:50 on Oct 31, 2022 |
|
#
?
Oct 31, 2022 00:33
|
|
|
B-Mac posted:Finally got my 7700X and 4090 build together, went with the MSI 650M Mortar Wi-Fi and some ddr5 6000 CL32. Been playing with the negative curve optimizer and PPT a bit. A lot of motherboards were having issues with 4090's and have been rolling out BIOS updates to fix that. The boot time being longer I suspect will go away with more BIOS updates. My guess is it is re-training on the memory on each boot.
|
|
#
?
Oct 31, 2022 02:44
|
|
|
I've been looking some more at mainboards. The decently only usable PCIe slot (the 4x one) is sure decently close to the PEG one on the B650 mainboards, making it a hindrance with thicc GPU in regards to airflow. I don't know, all designs are weird so far (what's with the five hundred M.2 slots, anyway). --edit: Nevermind. Probably fake. Combat Pretzel fucked around with this message at 11:15 on Oct 31, 2022 |
|
#
?
Oct 31, 2022 10:50
|
|
|
tpu did a new megabenchmark comparing 5800X to 5800X3D in 1080p with the 4090 as the bottleneck killer https://twitter.com/TechPowerUp/status/1587127732600799236
|
|
#
?
Oct 31, 2022 18:59
|
|
|
Combat Pretzel posted:I've been looking some more at mainboards. The decently only usable PCIe slot (the 4x one) is sure decently close to the PEG one on the B650 mainboards, making it a hindrance with thicc GPU in regards to airflow. I don't know, all designs are weird so far (what's with the five hundred M.2 slots, anyway). yeah I really don't understand this. 100% of the time when i've bought a board without enough PCI slots I've regretted it at some point when i've wanted to add some kind of connectivity or functionality. meanwhile i've only ever used one of the m.2 slots. maybe i would at some point use 2, but the amount on these boards is absolutely ridiculous I guess at this point they assume everyone buying one of these boards is going to be using some gigantic 3-4 slot GPU which makes half of the lower slots pointless, but that's not me!
|
|
#
?
Oct 31, 2022 19:45
|
|
|
kliras posted:tpu did a new megabenchmark comparing 5800X to 5800X3D in 1080p with the 4090 as the bottleneck killer shoutout to gta5 for somehow performing exactly the same
|
|
#
?
Oct 31, 2022 19:46
|
|
|
repiv posted:shoutout to gta5 for somehow performing exactly the same
|
|
#
?
Oct 31, 2022 20:14
|
|
|
Generic Monk posted:yeah I really don't understand this. 100% of the time when i've bought a board without enough PCI slots I've regretted it at some point when i've wanted to add some kind of connectivity or functionality. meanwhile i've only ever used one of the m.2 slots. maybe i would at some point use 2, but the amount on these boards is absolutely ridiculous Yeah i think the combination of GPUs blocking the slots anyway and the vast majority of users never using the extra PCIe slots these days means they are more of an afterthought. Users, including me, LOVE M.2 slots though. The problem with only having 1-2 M.2 slots is people i built for always wanted to "save" them for some future big fast SSD thats not on the market yet. I haven't heard that same complaint now that nearly every board has 3+. If you do need the PCIe slots, Asus has a number of cheaper boards with lots of slots, like the Z690-P or B650M-A. If you want a higher end board though, you are likely going to have to look at boards like the ProArts that have different, more equal bifurcation balances between the slots, or green board workstation products that fill the niche.
|
|
#
?
Oct 31, 2022 20:45
|
|
|
kliras posted:tpu did a new megabenchmark comparing 5800X to 5800X3D in 1080p with the 4090 as the bottleneck killer 
|
|
#
?
Oct 31, 2022 20:45
|
|
|
Josh Lyman posted:They used different motherboard, RAM, and RAM speed/timings. Why wouldn’t you use the same text bench just swapping out the CPU? i thought the whole point was to stop doing that with this
|
|
#
?
Oct 31, 2022 20:51
|
|
|
Josh Lyman posted:They used different motherboard, RAM, and RAM speed/timings. Why wouldn’t you use the same text bench just swapping out the CPU?
|
|
#
?
Oct 31, 2022 22:11
|
|
|
https://www.bhphotovideo.com/c/product/1696096-REG/amd_100_100000651wof_ryzen_7_5800x3d_3_4.html 5800X3D for $329, AMD seemingly trying to clear shelves quickly
|
|
#
?
Oct 31, 2022 22:37
|
|
|
5800X3D is down nearly £100 from launch in some UK stores, I might legit end up grabbing that, a GPU and power supply instead of a whole new build if it gets much lower.
|
|
#
?
Oct 31, 2022 23:05
|
|
|
https://youtu.be/I7-2ArdYvfA Hardware Unboxed compared the performance of various CPUs, the 4090 and DDR5 or fast DDR4 last week. Fewer games were tested but at least there was consistency in the setup.
|
|
#
?
Nov 1, 2022 00:09
|
|
|
HUB talked about some of the DDR5 memory weirdness that most reviews had on their Q&A show too. Apparently the early Zen4 bios were extremely picky about memory and they straight up couldn’t boot the DDR5 kits Intel supplied with 13th gen at the XMP/EXPO timings on Zen4 and vice-versa with the Zen4 kits on Intel. You could manually set some of the primary timings, but it was still dodgy. Hence all the reviewers using different ram settings between the two platforms. I think Steve said he would be revisiting it soon with the post launch bios for both platforms to see where speeds and computability are now.
|
|
#
?
Nov 1, 2022 00:31
|
|
|
Cygni posted:HUB talked about some of the DDR5 memory weirdness that most reviews had on their Q&A show too. Apparently the early Zen4 bios were extremely picky about memory and they straight up couldn’t boot the DDR5 kits Intel supplied with 13th gen at the XMP/EXPO timings on Zen4 and vice-versa with the Zen4 kits on Intel. You could manually set some of the primary timings, but it was still dodgy. Hence all the reviewers using different ram settings between the two platforms. I think Steve said he would be revisiting it soon with the post launch bios for both platforms to see where speeds and computability are now. Some reviews tested Raptor Lake with DDR5-6800 while Zen 4 was tested with 6000, and TechPowerUp even tested RL with DDR5-7400 lmao. I understand the idea behind "use the fastest memory available to show the CPUs' potential," but it's difficult to tell how much of an advantage Raptor Lake has in some of these reviews due to the faster CPU, and how much is due to the faster memory. And realistically, I don't think many people should be buying anything faster than DDR5-6000 right now anyway due to the high price of 6400 and above. Dr. Video Games 0031 fucked around with this message at 01:45 on Nov 1, 2022 |
|
#
?
Nov 1, 2022 00:45
|
|
|
The idea of like for like memory comparisons is usually stupid because usually you would be buying different memory to optimize for a specific platform. With DDR5 price concerns that's less of an issue right now, and maybe within a year or so it will finally change for good if both platforms perform equally well, but for the last four years it's been really stupid seeing reviewers doggedly insist on using only one type of memory to benchmark that either favors one vendor or just gimps both.
|
|
#
?
Nov 1, 2022 00:53
|
|
|
Toms got a direct and straightforward answer from AMD about the whole 2 CCDs on CPUs that only need 1 CCD: https://www.tomshardware.com/news/amd-recycles-dual-chiplet-ryzen-7000-as-ryzen-5-7600x-cpus It's not damage or failure, it's purely that they're fusing off a CCD to adjust production to demand. OTOH this does not necessarily that demand for the high-end processors is devastatingly low. For one thing, AMD says this was done on 5000 series CPUs too. (Probably we didn't notice because far fewer people were delidding those.) But the real way to think about is: if AMD is going to have a shortfall of CPUs that have 1 CCD or 2 CCD due to inexact predictions about what people are going to buy, which way do they choose to be wrong? If they make too many 2 CCD models, they can still sell them as 1 CCD CPUs. They make less profit (or maybe no profit), but that's it. If they go the other way it's actually worse for them. A shortfall of 2 CCD models means they aren't selling as many high-end CPUs as they could be -- and the high-end ones are way more profitable. If someone wants to buy a 7950X and there's none on the shelf, that's a big loss for AMD. Way more than the loss of trashing one CCD to sell an extra 7600X. I'd guess that right after the launch of new models is when their error margin is highest, so we're seeing a lot of them. I'm sure they're using sales data to tighten that spread.
|
|
#
?
Nov 1, 2022 14:08
|
|
|
there have been reports of the same thing happening for 5000 series cpus too yeah, just don't know if it actually got any attention at the time
|
|
#
?
Nov 1, 2022 15:21
|
|
|
lih posted:there have been reports of the same thing happening for 5000 series cpus too yeah, just don't know if it actually got any attention at the time I definitely remember people detecting this at the time - I think it was either igor or der8auer who found it. It didn't really catch any attention besides the idle speculation if you could do a Phenom and reactivate the extra chiplet somehow
|
|
#
?
Nov 1, 2022 15:41
|
|
|
Cygni posted:https://www.bhphotovideo.com/c/product/1696096-REG/amd_100_100000651wof_ryzen_7_5800x3d_3_4.html The optimist in me is hoping this means a 7800X3D is being targetted for Q1
|
|
#
?
Nov 1, 2022 15:51
|
|
|
Lockback posted:The optimist in me is hoping this means a 7800X3D is being targetted for Q1 I wonder if they're talking about a single-CCD 7600X3D internally. Give the gamers what they want!
|
|
#
?
Nov 1, 2022 15:58
|
|
|
Lockback posted:The optimist in me is hoping this means a 7800X3D is being targetted for Q1 The rumor mill is that it'll be announced or presented in some fashion at CES 2023 and be available in either late Q1 or mid-ish Q2.
|
|
#
?
Nov 1, 2022 16:39
|
|
|
Twerk from Home posted:I wonder if they're talking about a single-CCD 7600X3D internally. Give the gamers what they want! Dumb SKU, TBH. Why have all that extra material and not the two extra cores? If you want cheap performance, you get a 13600k. Or a 5800X3D while available. Any 7xxx3D CPU is not going to be a "budget" option. PC LOAD LETTER posted:The rumor mill is that it'll be announced or presented in some fashion at CES 2023 and be available in either late Q1 or mid-ish Q2. This is probably what'll happen, but I hope Intel's 13th gen pressure somehow magically drives the release date up to like January 20.
|
|
#
?
Nov 1, 2022 16:55
|
|
|
|
| # ? May 30, 2024 03:12 |
|
|
Lockback posted:The optimist in me is hoping this means a 7800X3D is being targetted for Q1 the rumour is february/march yeah
|
|
#
?
Nov 1, 2022 22:06
|
|