The 10 year plan computer

[:D:D]

Sometimes you just cant afford upgrades and your system becomes that 10year+ one without planning for it

 

[Chrispy_]

The curve of diminishing returns happens there, just about. Or below

For gaming, $400 is definitely well into the point of diminishing returns, but arguably a good price point to buy for a processor that you hope to use with your next graphics card. If I didn't already have 5800X3D I'd be procuring myself a 7800X3D for sure.

For this generation of graphics card, the 7800X3D is only 8% faster than a 7600 despite costing 90% more when paired with a relatively high-end GPU and looking at the most sensible resolution for that GPU.



You need to be spending $1200 on your GPU before it's worth spending an extra $100 more on your CPU like an upgrade to the 7700X or 13600K, but at that point you might as well just spend a little more and get the current GOAT.

Sometimes you just cant afford upgrades and your system becomes that 10year+ one without planning for it

Heh, there comes a point when the AAA games cost more than the hardware you need to run them.
With some games costing $70-100 now, you can probably find a GPU upgrade on ebay that meets the minimum spec - over on this side of the pond that's a used GTX 1070Ti or RX 5600XT!

 

[Desktopstu]

I think the "sweet spot" PC right now is the build you just mentioned but with a much cheaper 4070S or Radeon 7900GRE.

Everyone's definition of sweet-spot is subjective, to be fair - so there's no right or wrong, but the 7800X3D is going to fit the bill for most people because it's a flagship-beater at a mainstream price and it doesn't need a fancy motherboard, exotic cooling, or high-end DDR5 to get the most out of it.

As for graphics cards, it's about diminishing returns. Are you chasing a high-refresh experience at ultra settings in AAA games at 4K? If so you're going to be dissatisfied with anything except the 4090 right now and in maybe 1-2 years you're going to be dissatisfied with that too. There will always be games that bring PCs to their knees, but if you can accept Very High instead of Ultra, 1440p instead of 4K, and 75fps instead of 150fps, then your requirements go down by about 60% and the GPU needed to deliver that is 80% cheaper. Yes, if you stop and pixel-peep two side-by-side screenshots, you can see the difference, but realistically the game's textures and assets are the limiting factor in graphics quality these days and the quality of the gameplay isn't really affected at all by graphics once you hit an acceptable framerate at an acceptable resolution.

I'm rambling but the TL;DR is that chasing the very highest possible graphics settings in new games is a fool's errand and it won't actually make the game itself any better beyond a certain point. The "sweet spot" in terms of graphics settings is where you're getting enough fps to enjoy the game without choppy animations or input lag (probably 60-90fps for most people) and the visuals are good enough that you're not missing any detail. In a lot of so-called AAA games that's often lower than 1440p and lower than "high" settings.

I’m even considering a 27” 1080 with high refresh rates to extend the life of the hardware. I already have a standard 4K 60hz 27” benq monitor which can act as secondary

One of the prevailing bits of advice I’m reading is to change my tactics to keeping the same platform but upgrade the cpu/gpu over the 10 years. This would allow for high end but not flagship graphics now (such as 4080 level) and upper mainstream cpu such as 7800x3d or similar.
I think it would have to be AM5 really. I’ve had several AM4 based systems over the years (still have one) and found them to be very adaptable and capable. I expect AM5 to go the same way.

 

[Hyderz]

One of the prevailing bits of advice I’m reading is to change my tactics to keeping the same platform but upgrade the cpu/gpu over the 10 years. This would allow for high end but not flagship graphics now (such as 4080 level) and upper mainstream cpu such as 7800x3d or similar.
I think it would have to be AM5 really. I’ve had several AM4 based systems over the years (still have one) and found them to be very adaptable and capable. I expect AM5 to go the same way.

am5 might support 3 gens +, thats the feeling im getting, as with intel 12th 13th and 14th works with z690 and z790... i have a feeling intel might repeat with 12th,13th and 14th for its new cpu coming out later this year... but across those 3 generations... the performance difference wasn't as significant as am4 has proven. i would nab a 7600 on the am5 platform and go from there.... with a gpu and cpu upgrade 5 years... later...

 

[dgianstefani]

I’m even considering a 27” 1080 with high refresh rates to extend the life of the hardware. I already have a standard 4K 60hz 27” benq monitor which can act as secondary

One of the prevailing bits of advice I’m reading is to change my tactics to keeping the same platform but upgrade the cpu/gpu over the 10 years. This would allow for high end but not flagship graphics now (such as 4080 level) and upper mainstream cpu such as 7800x3d or similar.
I think it would have to be AM5 really. I’ve had several AM4 based systems over the years (still have one) and found them to be very adaptable and capable. I expect AM5 to go the same way.

I'd caution against being sure of AM5. Seems like AM5+ might release, and even then only get one? more CPU gen. I.e best case scenario up to Zen 6. Arrow Lake introduces ribbonFET, backside power delivery and Foveros to desktop, three major advancements which are also first to market, and will have at least one additional CPU generation on the same socket. Since you're building in 2025 it's best to wait and see. Intel also has more cores, and a more advanced hardware thread directed disaggregated design. Compared to Zen 5 which will still be basic chiplet architecture with no active interposer (same CCD latency problem and lower upper limit for RAM support ~6400 MT).

My estimation is Intel will take a firm lead with 15th gen vs Zen 5, and 16th gen/Zen 6 could be quite close, since AMD will use TSMCs active interposer tech to reduce CCD latency, idle power and the IF issue, but both will likely be the final architectures on their sockets.

Raptor Lake vs Zen 4 is quite close already, with Intel losing in raw efficiency due to a 10/7nm process being implemented rather than the 5nm TSMC of AMD, but this is somewhat made up by a monolithic design, and higher frequencies, for both cores and RAM, which has advantages.

Arrow Lake will be on a competitive node vs AMD, while also introducing significantly improved efficiency and signal integrity from powerVIA, gate all around transistors and a chiplet design that doesn't suck for latency, as touched on earlier. I.e. disaggregated non monolithic but without the downsides.

 

[nurgle]

The case is vital I think. My first AM3 build (Phenom II) went into a Corsair 600T. The case is sat by my right foot as I type this, everything else has been long upgraded!

 

[gurusmi]

I'd caution against being sure of AM5. Seems like AM5+ might release, ...

That would be in a few years and is not named AM5+! It is already known to be a missunderstanding. AM5+ was mentioned as AM5 and successors.

One can be happy to forsee what's going on within the next 24 month.

I planned my own Workstation to last 4 years. But i don't game. I'm developing software and design 3D models etc. I selected a AM5 7950x, at least 64GB of ram, 4 SSD, and a upper medium GPU. As far i can see that config will be supported within the next years by Windows and Linux.

Silly. It takes 10 Pages till now to explain that nobody can forsee 10 Years of IT and Hardware.

 

[micropage7]

I like my games with high graphics settings as Im notthat much of a competitive games player.

i guess that's the answer, it's harder to keep the current build to handle games in the future, like 3 or more years
the newer games need better hardware, at least you need to give a room for newer vga card in the future

 

[#22]

?? Value is an ever bigger factor for a PC that must last 10 years as opposed to 3-5, particularly if you are spending that entire 10 year budget up front. Hence why it makes little sense to spend up on high-end, where the least value is to be had. You get great performance for the first 2-3 years and then a continual downward drop for the last 7. That's not good system building and it's not good budgeting either. Good budgeting involves getting the most value and keeping money aside for necessary repairs and upgrades should they be needed whereas the 10 year no upgrades path leaves you with the worst experience on average over that 10 year period with zero flexibility and zero cash in case a part fails (which is more likely to be out of warranty when you are only buying once vs multiple smaller purchases). Purchasing multiple best value parts enables you to get the best experience over that lifespan of that PC in addition to ensuring it's covered by warranty and that you have free cash to cover repairs. Should anything unexpected happen to the "blow all your cash up front" build, you are in essence forced to break your original budget. At that point you've already purchased multiple parts over the life of the build, so it didn't last 10 years, and you didn't get any of the benefits of value upgrades over time either so it's a loose loose. Failing to plan is planning to fail, plain and simple.

He has 3000 pounds budget. If it's for tower only, it will get him a highend pc to give him highend experience for some time. You can't compare it to few value pcs giving at best value experience.

He has the plan of seeing how long this one pc will serve him without upgrading.

So to this point there was no point of talking here about value and upgrades But today I see that OP still changes his plan, now is about some upgrades, so you all feel free to fantasize with him endlessly - maybe in fact it's exactly what he is about. I don't like threads not moving forward or offtopic, so don't care and please don't bother talking to me about value and upgrades

 

[Panther_Seraphin]

Coming from someone who did the 10+ year PC recently.....I advise Dont!

5 Years sure. 7 years......maybe with some upgrades. 10 Years not a chance. I was begging for the release of Zen 5 at the end of my 10 years. It couldnt come quick enough!!
You are giving up far too much across the board for it to be worth it.

For example we went from DDR3 to DDR5 being common place. PCIe Gen 2 to Gen 5. SATA based SSDs to NVME Gen 4. Quad Core being the norm to having access to 16+ cores let alone Hyperthreading and the IPC improvements since then.
Advances such as DLSS, Ray Tracing, etc "AI" supported Audio and Webcams etc

 

[Desktopstu]

He has 3000 pounds budget. If it's for tower only, it will get him a highend pc to give him highend experience for some time. You can't compare it to few value pcs giving at best value experience.

He has the plan of seeing how long this one pc will serve him without upgrading.

So to this point there was no point of talking here about value and upgrades But today I see that OP still changes his plan, now is about some upgrades, so you all feel free to fantasize with him endlessly - maybe in fact it's exactly what he is about. I don't like threads not moving forward or offtopic, so don't care and please don't bother talking to me about value and upgrades

I did try to move forward but every direction I was met with ridicule and derision. So I just followed the majorities advise and just kinda gave in.
It was (and still is) a genuine plan but it’s been put back a year so it can be properly funded.

 

[Outback Bronze]

Hmm, we are 10 pages in on the 10 year plan so just thought I would throw this out there : )

How does the X99 platform hold up these days?? Not too badly if you ask me. This is coming up close to 10 years now.

You would prob need a 5960x (can be upgraded to 6950x 10 Core) which was expensive back then but let's look at some stats.

5960x 20MB 8 Core + HT 16 threads.
Quad channel DDR4
M.2 support
40 PCI-E lanes @ 3.0

I think gaming at 4k wouldn't see a huge difference in FPS but workloads vs current CPU's would get thrashed. I still have one of these setups outback and would love to do some testing but unfortunately, I'm very time deprived these days.

There are probably some forum members still running this platform today, albeit not many.

It was classed as HEDT back then so Dunno if that's actually relevant to this topic and have to admit it would prob get done over by a 12400F.

Yep, 5 years plan def way to go, especially with GPU's!

 

[BoggledBeagle]

...
5960x 20MB 8 Core + HT 16 threads.
...
There are probably some forum members still running this platform today, albeit not many.
...

Not many. Such a computer would have been built by an enthusiast (hardcore enthusiast even) with enough money to afford such a system, and such person would very easily see (and dislike) how the system started lagging behind the current stuff and would have gotten rid of such a slow and obsolete machine a long time ago.

 

[#22]

I did try to move forward but every direction I was met with ridicule and derision. So I just followed the majorities advise and just kinda gave in.
It was (and still is) a genuine plan but it’s been put back a year so it can be properly funded.

And then you will buy pc more likely to fail than fulfil your expectations due to unpredictable nature of industry. All of the advices and memories in this thread should teach you that having plans and long-term expectations or budgets just doesn't work here. That's why just buy yourself performance you are fine with paying for today without any longevity expectations and don't skimp on the rest determining quality of your everyday experience, making your computer nice or not. It's simple as that and all you can really expect.

 

[A Computer Guy]

It was (and still is) a genuine plan but it’s been put back a year so it can be properly funded.

Even if new CPU's/GPU's (etc...) come out during that time period are you going bleeding edge with the changes in your plan?
Personally I don't think you need to go bleeding edge to get a 10yr PC. Wether or not it lasts is another story.

If not going bleeding edge could you share a PC Part Picker list of what you are currently considering? This should give a ballpark figure of what your budget needs to be and an overview of your overall build.

Please don't feel pressured or discouraged by all the advice here. It seems to me your build is about the journey you want to take so you should do what you feel is right for your needs.

 

[Desktopstu]

Even if new CPU's/GPU's (etc...) come out during that time period are you going bleeding edge with the changes in your plan?
Personally I don't think you need to go bleeding edge to get a 10yr PC. Wether or not it lasts is another story.

If not going bleeding edge could you share a PC Part Picker list of what you are currently considering? This should give a ballpark figure of what your budget needs to be and an overview of your overall build.

Please don't feel pressured or discouraged by all the advice here. It seems to me your build is about the journey you want to take so you should do what you feel is right for your my though

My thoughts were that the pc would initially be high end and would grow old gracefully. I tend to play one game quite a long time - years often- so it doesn’t need to remain cutting edge for its duration. I’m going do some calculations that I will share here once I get the chance.

 

[lilhasselhoffer]

Hmm, we are 10 pages in on the 10 year plan so just thought I would throw this out there : )

How does the X99 platform hold up these days?? Not too badly if you ask me. This is coming up close to 10 years now.

You would prob need a 5960x (can be upgraded to 6950x 10 Core) which was expensive back then but let's look at some stats.

5960x 20MB 8 Core + HT 16 threads.
Quad channel DDR4
M.2 support
40 PCI-E lanes @ 3.0

I think gaming at 4k wouldn't see a huge difference in FPS but workloads vs current CPU's would get thrashed. I still have one of these setups outback and would love to do some testing but unfortunately, I'm very time deprived these days.

There are probably some forum members still running this platform today, albeit not many.

It was classed as HEDT back then so Dunno if that's actually relevant to this topic and have to admit it would prob get done over by a 12400F.

Yep, 5 years plan def way to go, especially with GPU's!

Just as a function of MSRP:
5960x - $999 at launch.
Motherboard - $210 to $600 at launch...so average to $400
Quad Channel DDR4.... so 4 sticks at $240 for a pair....or $480
($1879)

Let's add on a low end case, and PSU
$120
$185
($2184)

You'll need an OS...which we can call shenanigans on and get from a key reseller. You'll also need an SSD and HDD....because this is a 2014 build. That means 1$/GB in SSD and about $100 for an HDD.
$30
$256
$100
($2570)

No optical media drive...so you've got $430 for a GPU. That's a 970 at $349....which is pretty lackluster today.



I may be missing this...but I'm calling shenanigans. If I had to have Haswell-e today, when the 12 series and 5000 series exist, I'd be disappointed. My 2 cents....because that 256 GB SSD on SATA would be the bitterest pill to swallow.

 

[Desktopstu]

Just as a function of MSRP:
5960x - $999 at launch.
Motherboard - $210 to $600 at launch...so average to $400
Quad Channel DDR4.... so 4 sticks at $240 for a pair....or $480
($1879)

Let's add on a low end case, and PSU
$120
$185
($2184)

You'll need an OS...which we can call shenanigans on and get from a key reseller. You'll also need an SSD and HDD....because this is a 2014 build. That means 1$/GB in SSD and about $100 for an HDD.
$30
$256
$100
($2570)

No optical media drive...so you've got $430 for a GPU. That's a 970 at $349....which is pretty lackluster today.



I may be missing this...but I'm calling shenanigans. If I had to have Haswell-e today, when the 12 series and 5000 series exist, I'd be disappointed. My 2 cents....because that 256 GB SSD on SATA would be the bitterest pill to swallow.

If I was to build a 10 year old high end pc it would have 980ti which would still very much be useable today in my most humblest of opinions.

The above chart is somethign I just put together looking at performance of graphics cards you can purchase new (admittedly the 6000 series ones listed are new old stock with very limited stock) as found on a single etailer (OCUK). Price is in ££.
The `position` figure is taken from the Techpowerup graphics cards GPU database x 100 to bring the figure into a similar graph to the price (so a 4090 is position 1000, 7700XT position 450). The graph does not take into account RT.

As you can see the position is fairly linear with a few hiccups here and thereand even the price isnt far out being linear until you get to the elephant in the room that is the 4090.

There is no 4080 as I could not find any on the books from OCUK. I have not included second hand cards as prices are quite variable and its not a fair comparison to a new card price in my opinion.


The point of this graph is to show that for GPUs there isnt this steep incline of price v performance before a plateau of diminishing returns as much as some might think. Tghe price only graqph looks interesting as it climbs, flattens then climbs and flattens with nVidia hittiong the peaks followed by AMDs slight troughs. All this tells us what we already know - you pay a premium for the nVidia badge, which some may say is off set by how superior they are in RT and in most cases in power use.

I know some shops may be a little cheaper or more expensive and some models again may be more expensive brands but essentially a 4070 is a 4070 whether its an ASUS or a Powercolor. The difference just isnt enough for this graph (again in my opinion).

 

[evernessince]

If I was to build a 10 year old high end pc it would have 980ti which would still very much be useable today in my most humblest of opinions.

The 980 Ti is only 8 years 10 months old and that assumes you build your PC on the 980 Ti's launch date and are able to get initial stock. That's an extremely unlikely scenario, I mean look at what you are doing right now fixing to buy into a 4000 series GPU that's already 1 year and 6 months old. The date a 10 year PC lasts starts from purchase date, not product launch date. In reality you'd have to make that 980 Ti last 2-3 years more from now.

There are many restrictions a 980 Ti has as well (VRAM, VR performance lack of DLSS, RT, newer Nvidia features like integer based scaling (which is great for older games), limited in monitor options due to the old display port version, ect).

That said, given that you've indicated that a 980 Ti's performance in modern games is perfectly acceptable to you I don't see any reason why you'd even consider spending up on a high end graphics card. If a 980 Ti is still very much usable to you today then clearly you are fine with a mid-tier and even a low-end option and it makes far more financial sense in the long run. From what you've said, you sound similar to my dad who's used an RX 280 for god knows how long, to him the performance is still good and ultimately that's all that matters. I put together that system more than 10 years ago for $379 USD. Buying him a system worth $2000+ today would be a complete waste of money and based on what you've indicated it'll likely be the same to you.

IThe point of this graph is to show that for GPUs there isnt this steep incline of price v performance before a plateau of diminishing returns as much as some might think. Tghe price only graqph looks interesting as it climbs, flattens then climbs and flattens with nVidia hittiong the peaks followed by AMDs slight troughs. All this tells us what we already know - you pay a premium for the nVidia badge, which some may say is off set by how superior they are in RT and in most cases in power use.

View attachment 342911


I know some shops may be a little cheaper or more expensive and some models again may be more expensive brands but essentially a 4070 is a 4070 whether its an ASUS or a Powercolor. The difference just isnt enough for this graph (again in my opinion).

Correct, excluding the 4090 and 4080 pricing scales at a consistent basis. That said you have to ask yourself how much performance do you actually need. If a 980 Ti's performance is "playable" today as in your own words, even the RX 6700 XT is going to be multitudes better and would offer you the best value. It's getting you a good chunk of the performance of much more expensive cards while giving you 12GB of VRAM just in case you play a game that might actually use of it sometime down in the line. You can take the money you saved and upgrade again whenever you feel like you need to.

 

[Desktopstu]

The chart above shows cpu price (via OCUK) in ££ versus relative gaming performance as dictated by Toms Hardware Guide CPU benchmarks 1080 gaming benchmarks. Again ive added a 0 to the relative performance so 7800X3d sits at 1000 while the humble i5 3400 sits at 620.

Now... This chart is fairly linear with regards to performance witha bit of a drop off at the tail end there. but the price? thats all over the place! the reason for the prices being as they are is in part due to odd fluctuations on the intel parts with some intel 14th gen being cheaper than their 13th gen counterparts for some odd reason. Also cpus are not just for gaming (duh! obviously!) so some cpus with lots of cores will be more expensive than those with less yet those with less can be better at gaming.

What this does highlight though is that if you want to build a pc to last a long time, 7800X3d is the way to go. If you want to build something to last a relativly short while then i5 14600K is likley your best bet.

The 980 Ti is only 8 years 10 months old and that assumes you build your PC on the 980 Ti's launch date and are able to get initial stock. That's an extremely unlikely scenario, I mean look at what you are doing right now fixing to buy into a 4000 series GPU that's already 1 year and 6 months old. The date a 10 year PC lasts starts from purchase date, not product launch date. In reality you'd have to make that 980 Ti last 2-3 years more from now.

There are many restrictions a 980 Ti has as well (VRAM, VR performance lack of DLSS, RT, newer Nvidia features like integer based scaling (which is great for older games), limited in monitor options due to the old display port version, ect).

That said, given that you've indicated that a 980 Ti's performance in modern games is perfectly acceptable to you I don't see any reason why you'd even consider spending up on a high end graphics card. If a 980 Ti is still very much usable to you today then clearly you are fine with a mid-tier and even a low-end option and it makes far more financial sense in the long run. From what you've said, you sound similar to my dad who's used an RX 280 for god knows how long, to him the performance is still good and ultimately that's all that matters. I put together that system more than 10 years ago for $379 USD. Buying him a system worth $2000+ today would be a complete waste of money and based on what you've indicated it'll likely be the same to you.



View attachment 342914

Correct, excluding the 4090 and 4080 pricing scales at a consistent basis. That said you have to ask yourself how much performance do you actually need. If a 980 Ti's performance is "playable" today as in your own words, even the RX 6700 XT is going to be multitudes better and would offer you the best value. It's getting you a good chunk of the performance of much more expensive cards while giving you 12GB of VRAM just in case you play a game that might actually use of it sometime down in the line. You can take the money you saved and upgrade again whenever you feel like you need to.

What would I use the pc for? mostly gaming and work. My work does not require excessive cpu power so the pc will be made to work hard only really while gaming.

I play few games but plat them a lot, currently its mostly world of tanks which any of my many pcs can play. I want to expand my gaming to include games that challenge current hardware a bit such as Cyberpunk and do so at high settings, not neccessarily max.

In the past ive always compromised and I mean like ALWAYS. For example, remember when everyone was wanting an ATi 9800XT and Pentium 4 HT 3000? well I had an AMD 754 3400 and an nVidia FX5900XT as I just could not afford that ATi nor could I invest in upper tier socket 478.

A good friend of mine did invest in an ATi 9800XT and Pentium 4 HT 3000, he kept that pc going for about 5 years and it kept well up with the rat race until he then jumped across the socket 775 completely bypassing the dual core era and movong straight to quads. I on the other hand swapped and changed pc components at an alarming rate jumping from 754 to 939 to AM2 then to 775 dual followed by quad. the cost, in the long run was just crazy. If Id found the cash for the 9800 and Pentium I would have saved myself a lot of bother.

I know that when I build this pc build, while initially likely to be pretty powerful and play todays games just fine, will inevitably not keep up with newer hard ware as it gets older. But you have to factor in that I tend to play a few games for a very long time, not flit from game to game. Ive played World of Tanks well over ten years now for example. What id like to do is build this pc next year and see just how long it would last before even I get fed up with it.

Considering I have pcs here that include such ancient beasts as a 4790-GTX980, FX6300-79703gb, i5 10400-ARC770, i77700X-RTX3060, i710700-GTX1660S there are others but I cant recall all. I can and do play on old and modern(ish) hardware depending on how I feel. Each PC here has some history and while some get sacrificed to the god of Mhz most remain with me or are sold off to pay for more parts.

Next year I have the option, possibly for the first and last time, to purchase (within reason) compaonents to build a really high end pc and thats is exactly what I plan to do.

 

[Beginner Macro Device]

Did not read the whole thread but:

• 11th gen is busted. After I bought myself an i5-12400F I can't find any usecase where an i9-11900K would be favourable. All games but some extremely rare outliers run more smoothly, OS runs more smoothly, I also waste much less energy. Go LGA1700/AM5. Latter is brighter because you can slot in a Ryzen 7500F today as just a temporary measure and then upgrade it to a Ryzen 7 9700, or whatever the Ryzen 7 7700 successor will be called. X3D is to avoid because we're missing out on basic compute longevity. Games run ridiculously fast on a basic Ryzen 7600 anyway.
• 10 years ago, 8 GB RAM was plenty for basic tasks and any gaming, maxed out AAA games included. Today, it's 32 GB, or in extreme cases, 48 GB. If RAM requirements continue crawling up the same way they do today, by 2034 it'll be a must to have over 100 gigabytes of RAM if you want your thing to be smooth. I'd go for 64 GB: your PC is doomed to be too slow for 100+ GB tasks by 2034 either way.
• GPUs don't last long. You know by your experience with the GTX 970, it just doesn't cut it in the recent titles at all. 7900 GRE will be even worse by 2034. I'd recommend waiting till aftermarket 4090s become a triple figure purchase (likely a possiblity in 2025) and snagging one of them, or a discounted 4080 Super. However, this doesn't guarantee you any AAA gaming compatibility by 2034.
• PSU: https://cultists.network/140/psu-tier-list/ and bring anything based on the latest ATX standard from the S-tier. Preferrably over 1.2 kW.
• Case, storage etc: not an expert by any mean, won't comment on that.

 

[dgianstefani]

What this does highlight though is that if you want to build a pc to last a long time, 7800X3d is the way to go. If you want to build something to last a relativly short while then i5 14600K is likley your best bet.

This is the wrong way to look at this if you're talking about a 10 year PC. Let me explain why.

The reason why a 5960X is still fairly viable today is because it had eight cores, at a time when the average CPU for gaming had four. It also had better RAM bandwidth due to being quad channel.

Using a synthetic benchmark to emphasise part of what I'm referring to, the 7800X3D scores 1850 in ST and 18000 in MT. This can be seen as a simple numerical approximation of general performance. ST/MT ratings give a strong indication of how a CPU will perform in a general task, gaming or otherwise. Therefore they are useful for comparison.

A comparably priced Raptor Lake chip (14700K) scores 2200 ST and 35000 MT. A top end Raptor Lake chip like the 14900K scores 2400 ST and 40000 MT. That's 30% faster ST and more than twice as fast in MT than the 7800X3D.

To give some context, at the time the eight core 6900X released (a few months after Skylake), it was based on the same CPU architecture and cores as the one year older 5775C, "Broadwell". This is because Intel HEDT is one generation behind the consumer platform for architecture, and Intel takes the time to fine tune and ensure stability etc for the workstation platforms. Regardless, the consumer platform of the same year is always faster for gaming than the HEDT platform. The 6700K clocked higher than the 6900K, and was on a more modern core architecture, and games of the time didn't really need or scale with more than four cores, so the Skylake quad core was faster in games than the Broadwell E octa core. You can see this trend continue even today, with the current HEDT/workstation platform being based on Sapphire Rapids, which use Golden Cove cores, i.e. Alder Lake 12th gen. Meanwhile the current consumer 13/14th gen platform uses Raptor Lake cores, which are significantly faster for gaming. So. We've established that higher core count HEDT CPUs are slower than their consumer competition on release. Now let's look at the 6700K vs the 6900K for gaming in 2024, ~9 years after release. The 6900K has aged much better, and is much faster in games, due to being a true eight core CPU, with 16 threads and decent RAM bandwidth from quad channel memory. The slight advantage the 6700K had from ~500 MHz more frequency and a newer architecture is irrelevant against the fact that it's still just a quad core.

Now, the reason why the 7800X3D is competitive against that Raptor Lake 14700K I compared it with earlier is because of the ~100 MB total cache. But that doesn't work in all games, in fact in situations where the workload either fits in the ~40 MB of cache a non X3D chip has such as a 7700X, or doesn't fit even in the ~100 MB X3D cache, thus relying on RAM, the 7700X is faster due to having 10% better raw performance from higher clocks (2000 ST, 20000 MT).

What I'm getting at is that for the same money, you can get significantly more MT performance, and 20% better ST performance which is relevant in all games and applications, not a conditional perk like a large cache is. For a bit more money, you can improve that ST lead to 30%, and have more than double the MT performance (18K vs 40K).

Now, today, right now, the 7800X3D and the 14900K are neck and neck in games. With the X3D cache helping the Zen 4 chip to hold its own against the much higher clocked, much higher core count Raptor Lake parts in gaming. However, as time goes on, I strongly expect that balance to start shifting towards the Raptor Lake parts. Simply because more of their MT muscle will be able to be applied, not just in games, but as other software such as Windows or background tasks gets heavier too, making a higher demand on the CPU. Furthermore, even when tested with the same 6000 MT RAM kit (for a fair test platform), the Raptor Lake chips have significantly more memory bandwidth due to being monolithic with a superior memory controller, and can reach notably lower latency (with faster RAM) than the 7800X3D. Now that's 32/64 GB of RAM, not a ~100 MB cache that is sometimes relevant, sometimes big enough, sometimes not. Especially as time goes on and games get bigger and more complex.

Here's what I mean.

96/84/85K MB/s @6000C36 Raptor Lake

60/80/60K MB/s @6000C36 Zen 4

This isn't even taking into account the fact that the 147/900K Raptor Lake chips can all comfortably do 7200 MT memory, or even 8000 MT+ if you have a 1DPC board. These frequencies bump bandwidth up to around 125/120/120k MB/s, with latency in the ~45-50 ns range. That's almost double the bandwidth of what you can get with Zen 4.

What I'm trying to emphasise with this memory comparison, is 5-10 years down the line, the quad channel DDR4 Broadwell E 6900K still had good memory performance relative to newer platforms, but the dual channel Skylake CPU had memory performance that didn't hold its own anymore, besides the issue of it only being a quad core. It's a similar bandwidth comparison as the dual vs quad channel DDR4 Intel chips from 9 years ago that I'm making here, with Raptor Lake vs Zen 4. A stock Raptor Lake chip with the same memory kit has a solid 30% more bandwidth, tuned with a better kit (which aren't particularly expensive), you can take that 30% bandwidth advantage and bring it to 70-100% higher (125/120/120 vs 60/80/60). Plus a memory latency ~20% lower when paired with fast RAM, which also helps with gaming.

Try and consider how CPU demands will evolve as time goes on in your 10 year project, taking a look back at the last 10 years can give some hints.

 

[c2DDragon]

I read posts everyday and I'm amazed.
Guys, a lot of you didn't tell OP this is impossible. If you want to play games (which is supposed to be what OP wants according to the first post) in 10 years, do you want to play @10fps with FSR or DLSS or XeSS ?

PC is predominantly to be a gaming and office platform. Id like to play some games such as Cyberpunk, Baldurs 3, Red Dead. I currently play World of Tanks a lot. Id like to try some RT but thats not the big show stopper for me. I like my games with high graphics settings as Im notthat much of a competitive games player.

First : PC is not predominant when you talk about gaming, then it's impossible to tell what The Witcher 4 would need for 1080p @60FPS (CDProjektWreck) so...in 10 years ? With like what Unreal Engine 7 or 8 or all of us dead ? xD nah let's be real. Better advise OP to go for an AI platform to make money training his/her own AI before selling it or giving it for free while suggesting for donations.

Let's be real, you won't be able to run anything @60FPS (this is shitty FPS) in 10 years with a 4090 so why do you even try to search ?

EDIT : Even anonymous from Twitter (sorry I mean X) could tell this. Is it a Tech website/forum still ? I know we could do errors, I do many mistakes and I can be wrong then I can admit but 10 pages telling OP he/she can do it ? For real ?

 

[freeagent]

Indeed, history does show 10 year flagship GPU's getting destroyed by newer low end models. I used a GTX 580 until 2017.. total suckage

 

[c2DDragon]

Indeed, history does show 10 year flagship GPU's getting destroyed by newer low end models. I used a GTX 580 until 2017.. total suckage

My brother still has his 1080TI working (mine is dead) and no way he can play with good FPS in actual titles which are...for the most...shitty console ports x_x