Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


Server vs Consumer Hardware.. - Page 5
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

Server vs Consumer Hardware..

1235»

Comments

  • emghemgh Member
    edited October 2023

    @host_c said: @emgh , do not get so attached to 1 brand. keep an open-mind for others also.

    They're the only big server brand I've seen to advertise Ryzen's so openly, others might follow, or they might not, but not focusing on that makes absolutely zero sense, as it's my argument.

    That as suppliers start to offer them, they'll start to be used.

  • MaouniqueMaounique Host Rep, Veteran
    edited October 2023

    @emgh said: If Supermicro didn't see a real need for Ryzen's in their clientele, they wouldn't have launched the product.

    Ryzens are packing increasingly powerful video and, at least on OpenGL, they have some considerable power. With few exceptions (and then really impotent ones) the server-grade CPUs the like of Xeons and Epycs do not have any video capabilities at all and it is not hard to understand why.
    However, in today's world, a video card has other applications than, say, FIFA or GTA, servers are not used 90% for virtualization only, there are increasing numbers of cases where sheer compute power is needed and a DC edition of a video card costs so much and offers so little flexibility that a bunch of Ryzens could make up for that at a fraction of the cost compared to the DC-grade stuff.

    Thanked by 1host_c
  • @Maounique I'm not sure I follow

    Most Ryzen's don't either, right? Zero GPU capabilities?

    Also, and GPU heavy task should likely use, well, a GPU. Be it consumer grade if needed.

  • MaouniqueMaounique Host Rep, Veteran
    edited October 2023

    @emgh said: Also, and GPU heavy task should likely use, well, a GPU.

    True, but the integrated GPU in many Ryzens is good enough for many tasks and it comes almost free.
    Of course, it can't compare with a DC-grade GPU, but neither does the price, for one dollar you get much-much more power and you have the flexibility of tons of implementations available on the market to take advantage of that, unlike the heavily specialized world of DC-Grade GPUs.
    In sheer compute power, not only on CPU-specific tasks, but also GPU ones, you definitely get much more from a Ryzen than from the server-grade CPUs and GPUs per each dollar invested.

    Also, when we talk about CPUs, they break much less often than other consumer-grade HW such as SSDs/NVMes and the like, so, all the advantages with little drawback.

  • @Maounique I haven't seen Supermicro mention the G variants more than any other Ryzen CPU so I don't think that graphics is it really, I just think they're superior for some needs, even those of companies

  • @emgh said:
    @Maounique I haven't seen Supermicro mention the G variants more than any other Ryzen CPU so I don't think that graphics is it really, I just think they're superior for some needs, even those of companies

    Intel basically abandoned the E series, I have hundreds of e-2388g systems which are currently the best Intel has to offer. These chips are years old, and are basically just desktop chips with ECC support. An enterprise grade platform using Ryzens could be interesting for workloads that need very high single core performance, since that's a niche not filled by most Xeons and basically all Epycs.

    Thanked by 2emgh host_c
  • @fluffernutter said:

    @emgh said:
    @Maounique I haven't seen Supermicro mention the G variants more than any other Ryzen CPU so I don't think that graphics is it really, I just think they're superior for some needs, even those of companies

    Intel basically abandoned the E series, I have hundreds of e-2388g systems which are currently the best Intel has to offer. These chips are years old, and are basically just desktop chips with ECC support. An enterprise grade platform using Ryzens could be interesting for workloads that need very high single core performance, since that's a niche not filled by most Xeons and basically all Epycs.

    For sure, we're not a big comapany at all, but where I work at we've gone all-in on Ryzen's. We don't need a lot of cores, but our tasks are very database-heavy. The Ryzens, with their single-core performance (OVH watercooled) just shine

  • MaouniqueMaounique Host Rep, Veteran

    @emgh said: I haven't seen Supermicro mention the G variants

    They don't have to, the platform is supported so you can have any kind of a Ryzen, G or not, given support in the socket and BIOS. Do you think they would engineer a bios to block RDNA 2 and the like? I doubt it, that is a place where all hybrid MB producers should look IMO.

  • @Maounique said:

    @emgh said: I haven't seen Supermicro mention the G variants

    They don't have to, the platform is supported so you can have any kind of a Ryzen, G or not, given support in the socket and BIOS. Do you think they would engineer a bios to block RDNA 2 and the like? I doubt it, that is a place where all hybrid MB producers should look IMO.

    Obviously, what I meant is I don't think that's the main goal or purpose because otherwise they'd likely push it more

    I just think there's a GENERAL use case for Ryzens

    The G-variants, of course, could serve some of that

  • MaouniqueMaounique Host Rep, Veteran

    @emgh said: The G-variants, of course, could serve some of that

    Yep, RDNA 2 is powerful and ROCm is promising, I can see a lot of integration possible in the future where both computing engines are used for different parts of the workload.
    Might not be targeted now, but I think it would grow in use case scenarios.

    Thanked by 1emgh
  • davidedavide Member
    edited October 2023

    Am I the only dude with a cronjob that compiles the linux kernel once a week solely to test the hardware?

    When I still used consumer garbage GCC was the only tool that could detect hardware defects. Never one issue was ever found or suspected on this Supermicro.

    NOT ONE!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! B)


    that got emotional

    Thanked by 1host_c
  • host_chost_c Member, Patron Provider

    @davide lol, no, you are not that special.

  • davidedavide Member
    edited October 2023

    @host_c said:
    @davide lol, no, you are not that special.

    But I'm quite sure the archaeologists will never find another record but mine that GCC is a better hardware diagnostics tool than memtest and friends.

  • tentortentor Member, Host Rep

    @davide said:
    But I'm quite sure the archaeologists will never find another record but mine that GCC is a better hardware diagnostics tool than memtest and friends.

    That reminded me how AMD Epyc 1044 day bug was found :smiley:

  • ralfralf Member
    edited October 2023

    @davide said:
    Am I the only dude with a cronjob that compiles the linux kernel once a week solely to test the hardware?

    "I use Arch, BTW "

  • PulsedMediaPulsedMedia Member, Patron Provider

    @Maounique said: I have a RAIS-Redundant Array of Inexpensive Servers with many providers. Arguably, my RAIS is more resilient than any one setup which relies exclusively on one provider, be it Azure or AWS. My costs are also pennies and, while the quality is obviously not the same, the resilience is undoubtedly better.

    Curious on what you run there, and how did you set it up?

    @Maounique said: True, but the integrated GPU in many Ryzens is good enough for many tasks and it comes almost free.

    Intel's iGPU is widely used for transcoding and is surprisingly powerful for that task as well.

  • MaouniqueMaounique Host Rep, Veteran

    @PulsedMedia said: Intel's iGPU is widely used for transcoding and is surprisingly powerful for that task as well.

    A video card is good at video things... Good job, Intel... /s
    What I meant was the explosive use case scenarios of compute where GPUs are used, not the usual one of processing video or playing 3D games.

    @PulsedMedia said: Curious on what you run there, and how did you set it up?

    I am running storage and games/games routing for ppl in DTAG or other restricted areas who get dc in the evening or have unstable internet. They play with me and are ready to pay good money for this service. Extra bw I donate to Tor/Freenet/I2P etc. usually with bridges to fight off censorship at low BW and risks.

  • I'm a bit late on this thread, and probably repeating a good bit of what has been said here already, but I think the OP needed to slightly specify what he was referring to.

    A lot of providers are offering Ryzen/Intel Core machines, as that is currently part of the "hot selling" specifications. Customers are leaning towards high clock-speeds with slightly less cores in the industry at the moment, at least from what I have seen myself.

    I guess my question is, is this trend increasing or is it just a symptom of the lack of silicon due to covid? Or is it done simply to cut cost in an ever more competitive business?

    Truthfully, I don't see either of these being the cause. For example, a Ryzen 5950x(16c/32t) chip can be bought straight from amazon for $471.99 right now. An Intel Xeon E5-2698v3(16c/32t), can be bought as low as $25-30. So I wouldn't say its an issue about price. The big difference as said above would be the clock speed. (2.3GHz base vs 3.4GHz, and 3.6GHz Boost vs 4.9GHz).

    Now, I have seen a lot more companies starting to offer commercial boards as well. Not a huge fan myself, but have seen reports that overclocking is much more stable on consumer boards compared to something like the AsRockRack boards. Haven't personally tested this myself, but I could see it being true, considering overclocking tends to be a hobbyist thing compared to business.

    I did see some people complain about AsRockRack boards across this forum, and other forums, but personally in the 1.5years that I have worked at Tempest, I cannot think of a single AsRockRack board that I have seen. Hell, I don't think in the 15 years I have been building computers, that I have ever had a motherboard failure. PSU? Sure. RAM? Sure. Motherboard? Nope.

    I will also add, that oddly enough, the most popular machine at the moment at Tempest, is our 13900K overclocked machines. No e-cores, no hyperthreading, high clock-speed. Especially with the game-server clients, this is ALL that matters to them.

  • MaouniqueMaounique Host Rep, Veteran
    edited October 2023

    @Bushhy said: have seen reports that overclocking is much more stable on consumer boards compared to something like the AsRockRack boards

    There is no wonder there, the consumer boards have been tested under a lot of conditions by millions of people and each chipset upgrade takes that into consideration. The servers need stability so...

    @Bushhy said: Customers are leaning towards high clock-speeds

    Yes, we have seen many requests for that, but they petered out, I believe because there is more of a Ryzen offer right now.

    @Bushhy said: I did see some people complain about AsRockRack boards

    I complain about AsRock in general. They are hit and miss, sometimes they are great, some other times they not only fail, but burn other components as well, I think the power side engineering sucks.

    Thanked by 1darkimmortal
  • @Maounique said: There is no wonder there, the consumer boards have been tested under a lot of conditions by millions of people and each chipset upgrade takes that into consideration. The servers need stability so...

    Definitely. I have seen some great success with CPU overclocking on the enterprise boards, but it seems that RAM ends up being the harder side to run at a higher MHz/MTs. Recently pushed the 13900k to 6.0GHz on a p-cores without issue.

    @Maounique said: Yes, we have seen many requests for that, but they petered out, I believe because there is more of a Ryzen offer right now.

    Yep! I believe we tend to have more serious buyers, as we are known to be a good bit more expensive, for the performance of machines

    @Maounique said: I complain about AsRock in general. They are hit and miss, sometimes they are great, some other times they not only fail, but burn other components as well, I think the power side engineering sucks.

    From our experience, it tends to either be that the boards are DOA, or they tend to stay working for a good while.

  • @MrRadic said:

    @nocloud said:

    @MrRadic said:

    @nocloud said:
    i accidentality deleted the poll when editing EEC to ECC. So if you voted already, please re-vote!

    All Ryzen support ECC.

    Not according to wikichip, or chat GPT..

    As of my last knowledge update in September 2021, not all AMD Ryzen CPUs officially support ECC (Error-Correcting Code) memory.

    AMD has traditionally differentiated between its consumer-oriented processors and its professional/server processors when it comes to ECC support. CPUs in the Ryzen series, which are aimed at consumers, often lack official support for ECC memory. However, some of these consumer CPUs may still support ECC in practice, but it's not guaranteed or officially validated by AMD.

    https://en.wikichip.org/wiki/amd/ryzen_7/5800

    https://en.wikichip.org/wiki/amd/ryzen_9/5900

    The X series might but the Cézanne non-pro APUs in my example don't. or the non-X cpus

    They do, all of them do.

    Sorry for the thread revival, Just learn that this isn't true, just found out for sure today,

    All Ryzen apart from the G and GE (APUs) can unofficially support ECC depending on BIOS / MB

    Only Ryzen APU's that support ECC are the Pro versions. pro 5750G etc.

  • evnixevnix Member

    I was once part of a company that cheaped out on hardware that was supposed to be used in rugged environments like factories and assembly plants.

    Instead of using the standard supermicro/dell, we used consumer grade Intel NUCs instead.

    The savings that we made were completely wiped off by having to write custom software to support constant crashes and and another chunk was lost due to high rate of hardware failures. When you have thousands of these devices, even a small percentage of hardware failures can mean a lot of lost resources from customer care to technician visits etc.

    Thanked by 2Calin darkimmortal
  • jsgjsg Member, Resident Benchmarker

    While I very much appreciate that e.g. @host_c only uses pro hardware and avoids consumer junk (that's what most consumer products nowadays are IMO) I admit that I differentiate:

    • dedis - only pro stuff
    • critical VPS/VDS - strong preference for pro stuff
    • non-critical/non-business virtual servers - frankly, I don't care, and certainly not in the LE range
    • Pi and similar - have that in my lab but wouldn't colo or rent it.

    That's how I approach it as a customer, in particular as a (largely) LE customer. Whenever I was working in something like hosting (or more precisely DC and/or carrier) in a position where my voice had any weight in that kind of decisions I always was hardcore "pro equipment only, no compromises!". Today I'd probably be somewhat more lenient. Nuh, thinking about it, not really.

    Thanked by 1host_c
  • host_chost_c Member, Patron Provider

    @jsg said: While I very much appreciate that e.g. @host_c only uses pro hardware and avoids consumer junk (that's what most consumer products nowadays are IMO) I admit that I differentiate:

    You're right, we tend to avoid consumer-grade products primarily due to reliability concerns, which is a key issue for us.

    Just an example; a R730XD sent us an e-mail today that DIMM B3 "has left the chat".

    Everything is running smoothly, with no disruptions or hangs in services, thanks to the server’s ECC R-type memory. The only issue is that we need to shut down the node to replace the faulty DIMM. This maintenance will take place within the next 48 hours.

    That said, there's a significant demand from customers for such services, especially for AMD Ryzen. I've always acknowledged that, performance-wise, Xeons struggle to beat Ryzen in single-core tasks. This is mainly because consumer-targeted desktop CPUs like Ryzen typically have much higher clock speeds (3.4+ GHz), whereas server CPUs, with more cores on the die, tend to operate at lower frequencies.

    I will and cannot say that a Xeon Scale Gen 1 or gen 2 can outperform Ryzen 7 5XXX series, a nop, like never :D

    What can I do is get one of those Ryzen's, put the in PROD and see how it goes. :)

    Before we do this, we will notify some bench-markers around here. B)

    PS: Ryzen 7&9 7XXX (DDR5 ) outperforms most of AMD Epyc DDR4 CPU's also.......

    Thanked by 2jsg fatchan
  • fatchanfatchan Member
    edited October 8

    Server hardware is more suited to production environment due to OOB management. ECC counts too, for sensitive tasks. You see 10+ year dedis offered by just about every host on LET, which is a testament to their reliability. I wonder if in 10 more years we will see 7950x's being just as cheap :lol:

    Also echoing what some others have said, an "enterprise" mainboard can pair well with consumer CPU. I've had 2 ASRockRack x470d4u with Ryzen 3900x running at home for several years. Both run proxmox and also mine XMR for about half the year, so they see plenty of load. Hooked up to a decent UPS and they've never crashed or had any weird issues. One of the CPU was recycled from my old desktop PC, the other was purchased cheaply from ebay due to bent pins! Still chugging along fine :sunglasses:

    Thanked by 1host_c
Sign In or Register to comment.