Upgrade advice pl0x

[PrincessFrosty]

Normally I pain over upgrades for a good 2-3+ weeks getting them exactly how I want, that includes everything from reading the new tech, to building out components I'm happy with but I want to just trim the fat now and get peoples general feelings. On new build. I think I basically have to chuck everything, but maybe keep the 3080

Current system is:

Intel [email protected] (old but tends to meet my gaming needs okish still due to having a decent overclock)
32Gb RAM
2 nvme SSD in raid 0 (1b total)
4 old SATA drives, I will need a mobo that supports all four at once without disabling any other functionality (this was common with intel boards back when I was building last)
rtx3080
antec 1kw psu modular
4k primary monitor, 60hz, no HDR no dynamic refresh rate

General opinion specifically for someone upgrading from an old Junker like mine, doing a larger leap so not really worried about if single generational changes are worth it.

I know you need the new PSU for the newest video cards. Is there any noticeable difference between 3.0 and 4.0 PCI for video cards? Usually there's not earlier-on it their lifespan. I ideally want to keep data fast on my primary, if im correct there's double the bandwidth over the new PCIE meaning 8GB/s or close to it would be best for SSD speed. Any solid brands? At 8GB is there typical overheating problems? Don't know whether to go AMD vs Nvidia or AMD for intel for the main components, I kinda like the the whole nvidia suite and quality DLSS upscaled 4k I'm ok with. Has FSR come far enough to rival it?

How about prices are we still insane gpu wars like before or what, got my 3080 while I was holed up in hospital and could camp stores all day. I like the idea of a 5080 but unsure if that's the right route. I think I've seen a fair bit of praise for the 4080 super. I'm doing a lot of work with offline AI with ComfyUI and ideally want to generate art for a project i'm working on. VRAM is king there in many ways, would AMD be better, does anyone have any experience with the memory core tradeoff here?

Are there any pitfalls to look out for or generally good timing strategies for upcoming price drops?

I don't think im doing anything insanely high CPU core count taxing, so I dunno if AMD is best, any general opinions. Most of my original CPU power was for games with a single (or just a few dominant threads) aiming for fast clocks. I will be OCing,but just on a high end air cooler, nothing fancy.

What about the intel chipset design, back when I was last building nvme shared a bridge back to the CPU with the SATA drives and network and whatnot. Has that changed since, can drives use up a completely dedicated 4x lane? Im really thinking about future proof for when direct storage is taken up a bit more.

Depending on money i might not be able to do it all at once, so I'm thinking the cpu/mobo/ram/ssd first and a new gpu maybe a year out?

 

[Gray2233]

I think if you use this template it'll make things much easier for everyone involved:

Read first: Build and Upgrade Advice + Template!

It is exceedingly and increasingly difficult to recommend components to people that give limited to no information, those helping often do so as a hobby within their free time. To alleviate growing frustrations I feel having a base-line and template goes a long way, I am requesting that people...

forums.overclockers.co.uk

We're happy to help but we need more specific direction.

It really can get complicated at the moment and that's understandable, there's many on here that would love to lend a hand.

 

[Tetras]

I know you need the new PSU for the newest video cards.

They don't require a new PSU. The latest nvidia cards still come with adapters in the box. If you have a PSU that can handle a 3080, it'll almost certainly handle a newer one fine too.

Is there any noticeable difference between 3.0 and 4.0 PCI for video cards? Usually there's not earlier-on it their lifespan.

Did you mean to ask 4.0 and 5.0? I ask because modern systems are a choice between 4 and 5, not 3 and 4.

That said, unless the budget is very tight, buying a 4.0 board rarely makes sense at this point.

I ideally want to keep data fast on my primary, if im correct there's double the bandwidth over the new PCIE meaning 8GB/s or close to it would be best for SSD speed.

For most use cases the bandwidth over 4.0 is irrelevant (i.e. provides no noticeable benefit) and you're better off not buying into 5.0 yet, though if there's a great deal why not.

At 8GB is there typical overheating problems?

Do you mean 8TB? 8TB drives are pretty rare at the moment, not sure I'd recommend dumping that much dosh/data into/onto one SSD.

PCIE 5.0 drives can overheat/throttle fairly easily, but it depends on usage. Almost all boards have a built-in heatsink for the primary drive, though their effectiveness varies.

Has FSR come far enough to rival it?

FSR4 in the latest cards is much improved, so... kind of.

How about prices are we still insane gpu wars like before or what, got my 3080 while I was holed up in hospital and could camp stores all day. I like the idea of a 5080 but unsure if that's the right route. I think I've seen a fair bit of praise for the 4080 super. I'm doing a lot of work with offline AI with ComfyUI and ideally want to generate art for a project i'm working on. VRAM is king there in many ways, would AMD be better, does anyone have any experience with the memory core tradeoff here?

Nvidia is usually superior for AI, but it depends. VRAM is important, yes, though you'd have to check what your specific software and projects are likely to need.

Prices and availability are better than they were in the crypto boom, but are they good? Nope. Upgrading a 3080 meaningfully is still very expensive. The 4080 Super has pretty much disappeared. If you need VRAM and are prepared to spend big, I'd ideally wait for the rumoured Super versions with higher VRAM, rather than dumping £1K+ on a 16GB card. If time is money then your calculation might be different to mine.

I don't think im doing anything insanely high CPU core count taxing, so I dunno if AMD is best, any general opinions.

Don't know whether to go AMD vs Nvidia or AMD for intel for the main components

We need more information on your primary/secondary usage to give better advice/suggestions there.

What about the intel chipset design, back when I was last building nvme shared a bridge back to the CPU with the SATA drives and network and whatnot. Has that changed since, can drives use up a completely dedicated 4x lane? Im really thinking about future proof for when direct storage is taken up a bit more.

It hasn't changed, things are still a big mess and you have to check carefully in the tech specs and the manual for lane sharing between graphics, M.2, PCIE and SATA.

Depending on money i might not be able to do it all at once, so I'm thinking the cpu/mobo/ram/ssd first and a new gpu maybe a year out?

Makes sense. Your 3080 is still a great card and expensive to replace, whereas your CPU is very dated.

That said, 4K is very tough on the graphics card and depending on the games (particularly if older games), upgrading to a different CPU might offer you less than you'd imagine. Newer games tend to be more CPU demanding and a recent higher-end CPU would offer you more consistent performance, even when keeping the 3080 and playing at 4K.

I will be OCing, but just on a high end air cooler, nothing fancy.

I wouldn't bother with the latest CPUs, they're pushed hard enough already. If you were desperate to OC/tune, Intel's latest socket offers some decent opportunity for gains beyond just CPU clocks.

 

[PrincessFrosty]

Thanks for the feedback, to clarify a bit. I guess I'm aiming at PCIE 4.0 so far, back around the 3080 era I think they were almost fast enough cards to justify PCI4 I can't see 5 being needed for a long time, unless something has changed in they way it all scales. So I'm just kind of interested in components that wont be bottle necked too much with one another. My PSU is antec 1000w modular, but its older and doesn't have a native 12v-2x6, I read MSI cards ship with yellow connectors like this but wasn't sure about other brands, I couldn't find reliable info on this, "unboxers" and many reviewers skip over the connectors these days :/

Apologies, I meant 8GB/sec, when I did my research doing my current rig, the PCIE 3.0 standard was max 4Gb/sec or close to it, so I raid 0 2 nvme samsung pros to get the read and write to 4gb, it looks like the Samsung pro new drives are just shy of 8GB/sec so a single 1 or Tb would be just fine I guess, for a 8GB/sec bus?

Actually I think I've seen FSR in a some of the games I play, I'm not sure if it's 4 or not, I guess just testing is best. I'm so out of the loop of painfully arguing image quality these days.

I've been generate the kind of art I want using the 3080 and it's bearable, I was tinkering with the new WAN 2.2 models for video and would prefer them a bit longer, but with length come vRAM requirements, at the same time generation speeds are sensitive to number of dedicated cores towards AI, I knew Nvidia were fast out the gate for this, but wondering if AMD are upping the cores, I'm not sure what their Tensor core equivielent is like and how much die space they've been putting towards it. AI Is definitely a secondary thing atm.

OK I'll be on the lookout for the lane sharing, this basically would be Mobo specific right, simply being PCIE4 compliant doesn't guarantee a dedicated 4x lane for the SSD?

Yeah I hear you on the CPU,sorry I made a mistake its actually an 8700k not 7700k, it's so old i Literally forgot what generation its in. I was kinda worried that CPU cores would run slower with so many more of them (I've just finished reading P v E cores) but it seems like actually the P cores boost higher than even my overclock. If there's new variants of the current gen, supers with more vRAM or good AMD equivalents I may just wait on graphics then.

Ah I always had the mindset of buy the i7 variants and just OC the crap out of them, mine has run at 4.9 (and 5) ghz for all this time, when I saw boost clocks of like 5.4Ghz I did wonder how much OC headroom there was, if it's small then I guess I wont bother. I'm seeing the 14700KFs cheaper than most, I dont need onboard graphics and since it's such a big jump for me this should be OK right? I get that it's probably not a good upgrade path for a 12700k or 13700k user though.

For anyone else chipping in, I forgot to mention it's just an old school 60hz monitor, no HDR no dynamic refresh, so I kinda want to get to a point where the floor of my fps is >60.

Appreciate your insight Tetras, thank you for taking the time to reply.

 

[Tetras]

I'm seeing the 14700KFs cheaper than most, I dont need onboard graphics and since it's such a big jump for me this should be OK right? I get that it's probably not a good upgrade path for a 12700k or 13700k user though.

I wouldn't recommend buying a 13th gen or 14th gen CPU at this point.

The 265K is a good price and avoids the shenanigans with BIOS updates to fix reliability issues on the older socket, that we still don't know are truly fixed.

You do get a 5 year extended warranty on 13th-14th gen K CPUs, which might be enough to persuade you... It is not enough for me, they're all sitting in my avoid bin.

OK I'll be on the lookout for the lane sharing, this basically would be Mobo specific right, simply being PCIE4 compliant doesn't guarantee a dedicated 4x lane for the SSD?

Every motherboard can be different, yeah.

For AMD there's this video which can be helpful if there's a specific board you're looking at (timestamps in description), but ultimately burying your head in the manual is the only reliable way (and even that it is not 100%):

Listing e.g. 4x PCIE 4.0 M.2 slot definitely does not mean it has 4x M.2 slots available. You're usually safe with getting at least 2 slots free (at least, with a higher-end motherboard), but 3 or 4 is very likely to have some limitations.

Secondary PCIE slots are also very likely to have sharing. For AMD, the presence of USB4 (on X870E/X870) is another complication, since that steals 4 of the CPU's 8 spare lanes (aside from graphics).

I've been generate the kind of art I want using the 3080 and it's bearable, I was tinkering with the new WAN 2.2 models for video and would prefer them a bit longer, but with length come vRAM requirements, at the same time generation speeds are sensitive to number of dedicated cores towards AI, I knew Nvidia were fast out the gate for this, but wondering if AMD are upping the cores, I'm not sure what their Tensor core equivielent is like and how much die space they've been putting towards it. AI Is definitely a secondary thing atm.

AMD do now have more hardware and software support for AI, yes (RDNA2 / 6000 series had none), but from what I'm aware they're still lagging behind. New drivers can change/improve things and better support within apps can do so as well.

Apologies, I meant 8GB/sec, when I did my research doing my current rig, the PCIE 3.0 standard was max 4Gb/sec or close to it, so I raid 0 2 nvme samsung pros to get the read and write to 4gb, it looks like the Samsung pro new drives are just shy of 8GB/sec so a single 1 or Tb would be just fine I guess, for a 8GB/sec bus?

The talk of GB/sec is making mah brains hurt.

To simplify what I said earlier: there are some workloads that the bandwidth matters, like giant file copies from drive to drive, but for the most part this is not important. For games, for example, you would be hard pressed to notice the difference between an average PCIE 3.0 drive and a PCIE 5.0 drive. This recent video covers that topic:

The chances of the SSD being the bottleneck for productivity is usually quite small and limited to something like opening up a bunch of files/large projects, making backups or writing the output to a drive. It is generally much less likely to be a bottleneck than e.g. RAM capacity or CPU cores, so we'd suggest buying a good, reasonably priced PCI-E 4.0 drive (like Crucial's T500), which is more than enough for almost everybody.

If you KNOW that your SSD's bandwidth is a significant bottleneck, then it might be worth buying a PCI-E 5.0 drive, but I'd make that calculation by time saved and remember that if e.g. you're copying from one drive to another, the destination drive has to support that kind of speed too.

Most other workloads are more dependent on other factors, since large sequential writes are normally in the minority. I wouldn't say that controllers and flash have improved massively with PCIE 5.0 beyond bandwidth. You can monitor your usage if you're not sure (even task manager can do it), to see how often your SSD is bottlenecking.

I can't tell you if your AI work is something that needs storage bandwidth (or not).

Yeah I hear you on the CPU,sorry I made a mistake its actually an 8700k not 7700k, it's so old i Literally forgot what generation its in. I was kinda worried that CPU cores would run slower with so many more of them (I've just finished reading P v E cores) but it seems like actually the P cores boost higher than even my overclock.

Ahh, the 8700K is still a pretty good CPU, especially at high clocks.

The E-Cores do run slower than the P-Cores, but they mainly make the difference in heavily multithreaded stuff, so the lower clocks don't matter when you're gaming or lightly threaded.

You're right though, they boost incredibly high "out of the box". There's not a lot of headroom anymore and even if there was, the power cost/efficiency loss is rarely worth it. I'd say undervolting is much more popular than overclocking now.

If there's new variants of the current gen, supers with more vRAM or good AMD equivalents I may just wait on graphics then.

Most reviewers were very disappointed with the VRAM on the 50 series, so I wouldn't be surprised if the rumoured Super versions come with the upgrade we've been told about. If you're prepared to wait, I definitely would, especially given your workload.

I dont need onboard graphics

If you're going Intel, I'd still recommend getting a CPU with the integrated. It can be handy for troubleshooting (...PCI-E 5.0 has been a 'mare on that front) and you never know when the CPU's IGP can step in to help accelerate something.

My PSU is antec 1000w modular, but its older and doesn't have a native 12v-2x6, I read MSI cards ship with yellow connectors like this but wasn't sure about other brands, I couldn't find reliable info on this, "unboxers" and many reviewers skip over the connectors these days :/

The first version of the new connector was nvidia specific and I believe it came with the 30 series.

It was later adopted/adapted and known as the 12VHPWR. This connector was put into ATX 3.0 PSUs and those PSUs are also given a label of PCIE 5.0. These PSUs are supposed to be able to handle higher power draw ('cos modern cards are prone to having huge power spikes).

The connector was revised again and now known as 12v-2x6. In theory, the cable has remained the same, but in reality it is probably not the same, since as one of the cable manufacturers has said, they took the opportunity to improve both the connectors and the cables from the prior learnings.

I believe MSI's yellow connectors are meant to help you ensure the cable is plugged in fully, but it is pretty much confirmed that this is not why they burn (or at least, not the ONLY reason why they burn).

In terms of older PSUs, you have 2x options. You can buy what we call a "native" cable, that gives your existing PSU a 12v-2x6 cable to plug into the modular ports, or you can just use the adapter in the box for your old PCIE 8-pin connectors.

If you want to know which option is safer, I have no idea, but I'd be inclined to say someone like Seasonic (who are the OEM for most of Antec's higher-end PSUs) should know how to design/build a PSU and a suitable cable/connector better than most.

Thanks for the feedback, to clarify a bit. I guess I'm aiming at PCIE 4.0 so far, back around the 3080 era I think they were almost fast enough cards to justify PCI4 I can't see 5 being needed for a long time, unless something has changed in they way it all scales.

Higher-end cards with 16 lanes like the 5090 show minimal gains in PCI-E 5.0 boards and many users have just turned it off anyway for better compatibility/stability.

That said, some other cards (especially ones with less VRAM or less lanes) DO show a performance loss with earlier PCI-E gens, so my recommendation is don't bother with PCI-E 4.0 graphics anymore.

With the introduction of more affordable PCI-E 5.0 boards (like B850 for AMD) there's really no point in saving the money and losing PCI-E 5.0, not unless the budget is very small.

For the M.2 slots, I'd get at least one, maybe for your circumstances 2 would be nice, but I'd say that's enough because I'd expect you to only need the primary drives to have that kind of bandwidth.

 

[PrincessFrosty]

Awesome thank you. I think I'll stick to PCIE 4.0 then it seems like with a 5080 it's not a bottleneck and nvme drives it allows a lot of fast transfer already. My primary concern for the nvme is that with DirectStorage, now it's implemented in Unreal Engine 5 (and many elsewhere?) it means vRAM and nvme have a direct path for data that avoids the CPU, resulting in better load times (when implemented) and I wanted to have a drive prepped for that. Sadly uptake on console is good (as I understand it) on the PC is very slow. Few engines implement it yet and even fewer games, but if this is to last me another 6 generations I just dont want to make any dumb choices now. I may shoot for a mobo with at least 2 nvme shots, get 1x 1Tb now and then raid 0 another in a few years time.

I did assume that it was going to be safer with a straight 12v-2x6 cable from compatible CPU and that 3, 8pin adaptors are just plain bad, but if that is what is shipping in boxes then great it saves me PSU upgrade I guess.

Lastly has anyone experienced the leap from DLSS 3 to 4 and with frame gen? It's tied to the 4xxx series plus but I read some stuff like DLSS this limitation is manufactured.

Is there a decent equivalent performance AMD build for this kind of thing, what are the basics I'd be looking at, CPU/Chipset/GPU. That's my last deep dive really, I just need to work out what the pros and cons are for either side. Especially on price, if the AMD side are similar specs but cheaper I may go that route.

 

[Tetras]

I may shoot for a mobo with at least 2 nvme shots, get 1x 1Tb now and then raid 0 another in a few years time.

Sorry, my comment was vague, I meant that I'd recommend a motherboard with at least 1x, preferably 2x PCI-E 5.0 capable M.2 slots (..even if you only currently buy/own PCI-E 4.0 drives), alongside maybe 1x or 2x additional PCI-E 4.0 capable slots, depending on your expected need for future storage.

I think I'll stick to PCIE 4.0 then it seems like with a 5080 it's not a bottleneck

It doesn't matter now, no, but with the primary graphics slot, I would strongly recommend that you do not buy a motherboard which is not PCI-E 5.0 capable.

With AM5, for example, it would only save you like... £20-£30, if that, which just isn't worth it.

My primary concern for the nvme is that with DirectStorage, now it's implemented in Unreal Engine 5 (and many elsewhere?) it means vRAM and nvme have a direct path for data that avoids the CPU, resulting in better load times (when implemented) and I wanted to have a drive prepped for that.

I see, but you only mean gaming, right? You're not a developer? If you're a UE5 developer I think there's likely to be more justification to fork out for a PCI-E 5.0 SSD, depending on the impact on budget elsewhere.

I did assume that it was going to be safer with a straight 12v-2x6 cable from compatible CPU and that 3, 8pin adaptors are just plain bad, but if that is what is shipping in boxes then great it saves me PSU upgrade I guess.

So far as I'm aware, the box adapters are not more likely to burn than the native cables. They're still in pretty widespread use. They do, however, offer more points of failure (with the extra connections) and they're a lot messier than a native cable.

I don't know if you can buy a 12v-2x6 cable for your PSU, if you can drop the model then maybe we can find out.

Is there a decent equivalent performance AMD build for this kind of thing, what are the basics I'd be looking at, CPU/Chipset/GPU. That's my last deep dive really, I just need to work out what the pros and cons are for either side. Especially on price, if the AMD side are similar specs but cheaper I may go that route.

Some examples:
- 14700K (don't personally recommend, given potential for issues) would be Z790 (B760 if budget limited).
- 9800X3D (gaming priority)/9900X (productivity priority) would be B850 or X870/X870E (if USB4 or more M.2 required).
- 265K would be Z890 (B860 if budget limited).

AMD's highest GPU is the 9070 XT, 7900 XTX is technically faster (for raw performance, not necessarily with ray tracing and upscaling) and higher VRAM, but availability is limited and the feature set is an issue.
Nvidia's my guess from what you're describing (and the need to upgrade from a 3080) you're looking at a 5070 Ti minimum.

I'd personally be very reluctant to buy a CPU or GPU slower than the above from what you have. It is arguable a 9070 XT or 5070 Ti is not enough of an upgrade either (... opinions vary), so would definitely recommend waiting for the Super versions if possible.

 

[Gray2233]

With the speed of current drives you're not going to gain anything with direct storage from running Gen 4-5 NvME's in raid 0, in fact I wouldn't be surprised if it introduced enough latency to be a detriment.

A good Gen 4 drive like the Crucial T500 or Kingston Fury Renegade will perform extremely well with direct storage enabled titles and other games singularly, they're probably even overkill.

That said, the reason the uptake for direct storage with PC's isn't as extreme is because they also tend to have a lot more RAM/VRAM than consoles, especially when you factor in the fact the RAM in consoles is shared between general tasks and the GPU. Direct Storage is a compensation technique that isn't necessarily required for PC's, it can be a benefit but probably more for the lower end.

I feel the OP is severely overthinking and perhaps researching things that might require a more nuanced understanding. <~ (Edit: I really didn't mean this condescendingly but on reading it again it does come across as such, apologies).

 

[PrincessFrosty]

Given the cost difference and that PCI5 doubles the bandwidth again I think I will go with PCI5 then, I'll stick with the cables out of the box for the GPU there's no need to buy another 1kw PSU again just to get the right connector if these ones in the box are provided by the manufacturers I'm happy with that. Antec were good for PSUs when I bought, I think they map all their power rails, so I can spread that evenly no problem. (it was just getting info on these cable was surprisingly hard)

The reason I was interested in PCIE 4/5 and nvme speeds is primarily for DirectStorage where the GPU can read from the SSD directly, bypassing the rest of the system. The 3 potential points of bottleneck are data rate of the GPU (which is super big), the PCIE lanes and the SSD itself. When I built my 8th gen system I aimed ensure I had the bandwidth the saturate at the the time what was the PCIE 3 limit.

I've seen the difference in game load speeds when they implement this and it's kind of a generational leap thing, so I want to maximize it when it FIANLLY becomes popular on the PC. I hear that GTA6 might implement it, and any future UE games can. I just don't want to stupidly build in a bottleneck now if in 4-5 years time I can.

Don't worry I didn't see that as condescension, I've been a hardware nut for over 2 decades, it's just that i've not kept up since my last build, and to do a new one I kinda need to get back it to it, but as I said I don't want to spend weeks going down rabbit holes.

It is nuanced which is why I want that understanding to avoid mistakes building an expensive system and then finding out later it's bottlenecked by something. Direct storage for example is primarily faster because going through RAM requires reading from disk, decompressing a lot of assets using the CPU while in RAM, and then going back to vRAM for use. DS just bypasses the CPU so it saves more on CPU time which is the bottleneck loading a lot of modern titles. Having more RAM is not necessarily all of the benefit. Also that decompression is way quicker on the GPU, it's a workload more suited for the GPU.

This would basically be a kind of GTA6 ready system, designed to last 6 years with maybe a GPU upgrade in 3.

You've all been very helpful, thanks for taking the time, I have a good outline of what I want now.

 

[Tetras]

Given the cost difference and that PCI5 doubles the bandwidth again I think I will go with PCI5 then

Great!

I've seen the difference in game load speeds when they implement this and it's kind of a generational leap thing, so I want to maximize it when it FIANLLY becomes popular on the PC. I hear that GTA6 might implement it, and any future UE games can. I just don't want to stupidly build in a bottleneck now if in 4-5 years time I can.

From HUB's testing, it looks like so long as you have a M.2/PCIE/NVME drive, you will get the benefit. Currently, it isn't a big issue which gen (usually just a few seconds), though it is true that future games might change that.

I'd definitely stand by what I suggested, that you buy a motherboard with at least 1, preferably 2 PCI-E 5.0 capable M.2 slots (more than 2 PCI-E 5.0 capable M.2 slots often gets tricky with lane sharing 'n such). That'll make you safe for the future, if games do suddenly start needing huge read/sequential speeds.

Personally, I think a good PCI-E 4.0 drive will be plenty safe for the foreseeable and you can always move it over into use as secondary games/storage in the future.

That said, PCI-E 5.0 drives are not as expensive as they were, they're not over double or even triple anymore, so if you're not harming your CPU/GPU choice by picking one (or somehow limiting your build unreasonably elsewhere) then I wouldn't make an issue out of it.

 

[PrincessFrosty]

I've done some more reading on AMD and decided against them for CPU, I just don't do that much hardcore CPU multi-threaded stuff like video edit/encoding, or any kind of software rendering or work with VMs or like that, hosting various server related things is the most I do, dayz servers, media servers, stuff like that. In the end I went with:

14700KF
MSI PCIE 5.0 mobo Z970
32Gb 2x16 kit of 6000Mhz RAM
1Tb Samsung 990 Pro SSD

I'll probably put in a 2nd SSD later if there turns out to be a benefit, and probably next month look at a video card initial thoughts were a 5080 but not sold on that yet, it depends on price and alternatives.

From what I've read the intel issues you guys mentioned are acknowledged and so far have been addressed, I think there is still some speculations of problems? But near as I can tell they're sorted and if not then I'll just RMA the parts.

 

[Tetras]

I've done some more reading on AMD and decided against them for CPU, I just don't do that much hardcore CPU multi-threaded stuff like video edit/encoding, or any kind of software rendering or work with VMs or like that, hosting various server related things is the most I do, dayz servers, media servers, stuff like that.

I'm not sure I understand your reasoning there.

The X3D CPUs are for users who are primarily gaming, not multithreading. It is the Intel CPUs (particularly the 14700K and 265K) that offer the benefits over the 7800X3D/9800X3D in multithreading workloads.

The 265K is faster than the 14700K in multithreading, but generally slower in gaming (the Core Ultra CPUs have a reputation since launch for being inconsistent performers).

MSI PCIE 5.0 mobo Z970

FYI: This is unlikely to have a PCI-E 5.0 M.2 slot and if it does, it'll almost certainly be shared with the GPU lanes. That's because the 12th-14th gen CPUs do not have any spare PCI-E 5.0 lanes aside from the graphics lanes.

But near as I can tell they're sorted and if not then I'll just RMA the parts.

Fair enough, but we can't say that they're fixed yet. The latest fixes have not been available for long enough and Intel already claimed that once (and it was still not). But, yeah, I guess if you're happy to RMA then you have 5 years of operation.

 

[PrincessFrosty]

It's a whole bunch of stuff. The system is going to be multi purpose, I only have 1 system to do my home theatre, gaming, my server hosting, then my work. But also overclocking is somewhat of a hobby thing I have done on an off since building my own PCs. I've got custom cooling and decent case cooling that I intend to re-use in the build.

Specifically with the 14700kf I want to tinker a bunch with lowering the E core frequencies and voltages to see if I can push the P cores to run faster I'm reading people are pushing these quite high (6ghz+). You can see that's more or less what I did with the 8700k. And it's last me a good while giving OK performance into its long life.

All the benchmarks I saw for games for the CPUs showed massive numbers even on a lot of newer titles and that the difference generally didn't warrant the price, I'm still targeting 60hz/fps. I think if I was ready to take the leap to buy a nice variable refresh rate monitor, I'd go with the faster X3D. I also know that I'll want all the bells and whistles for when games like GTA6 drop and that'll mean GPU bottlenecks anyway.

If I'd had a better budget it'd go with an insane £2-3k refresh with new everything but I'm going to struggle to even afford a new good video card over the next few months to accompany the build.

Sad to see OCing go the way it has though, I had no idea it had fallen so much out of favor. All this makes me feel old.

 

[Tetras]

Specifically with the 14700kf I want to tinker a bunch with lowering the E core frequencies and voltages to see if I can push the P cores to run faster I'm reading people are pushing these quite high (6ghz+). You can see that's more or less what I did with the 8700k. And it's last me a good while giving OK performance into its long life.

I don't want to keep flogging a dead horse here, just trying to share before you can't change your mind, but my personal take is that the 265K is a much better tuner's chip at this point, just not so much for the clock frequency.

The Intel mitigations for the 13-14th gen CPUs put some limitations on their power use and voltages (I think previously if you did any manual config it would disable the mitigations, buildzoid made a video on it way back when) and pushing high clocks/volts with these CPUs I suspect may make them more vulnerable to the degradation. That said, stock voltages can often be higher than optimal OC voltages, so I can't say this would apply in every case.

You've obviously done your research and have your reasons to buy one, but I'd strongly push you to consider the 265K again, especially if you don't care about high FPS/refresh gaming, since it alleviates their main disadvantage.

It's a whole bunch of stuff. The system is going to be multi purpose, I only have 1 system to do my home theatre, gaming, my server hosting, then my work.

The 14700K is a better multi purpose CPU than the 8 core variants of the X3D, so sure, I do agree with that, but the 265K has higher performance (sort of, maybe better to call it potential), lower power use and better efficiency.