Ryzen Allegedly Matching Intel in 3DMark FS Physics

[leadeater]

21 minutes ago, MageTank said:

My worry is, if these 8c/16t SKU's indeed have a 65w TDP, something has to be limiting them in some regard. I simply don't believe they are still boosting to 3.7ghz on 8 threads and are still cutting Intel's TDP directly in half. Some sort of sorcery is at play here.

Yea my thinking is those 65W TDP are limited as well, no reason to rate the same die in another SKU as 95W. Makes no sense, not even taking in to account higher base or boost frequency.

[MageTank]

1 minute ago, leadeater said:

Yea my thinking is those 65W TDP are limited as well, no reason to rate the same die in another SKU as 95W. Makes no sense, not even taking in to account higher base or boost frequency.

Yeah, if they all have the same potential, buying the "allegedly" $500 SKU would make no sense when you can get the "allegedly" $300 SKU and OC just as high. My theory is either all of them have been binned, and the cheaper ones are sub-par overclockers, or they are cut down in some other regard. Dark times are coming. I can already see the threads. "Should I get the 1700 or 1800? What's the difference?". Prepare yourself people, the end is nigh!

[PianoPlayer88Key]

11 hours ago, Energycore said:

One has to keep in mind how little gains are enough for Intel to call them a "new gen". From Broadwell to Kaby Lake, only once has IPC gone up, and only twice from Haswell to Kaby. Ever since Kaby brought no per-clock performance increase and only a little clockspeed, it's not entirely accurate to look at a new gen as a new step in performance.

 

I also wonder if they're calling releases a "new gen" more often than they used to.  For example, release dates of some Intel families (source = cpu-world):

Spoiler

 

• 8086 = 1978-06-08
• 286 = 1982-02-01 (1334 days)
• 386 = 1985-10-17 (1354 days)
• 486 = 1989-04-10 (1271 days)
• P5 = 1993-03-22 (1442 days)
• P6 = 1997-05-07 (1507 days)
  • (Pentium Pro was 1995, but was precursor to Xeons / HEDT from what I can tell.)
• Netburst = 2000-11-20 (1293 days)
• Core = 2006-07-27 (2075 days)
• Nehalem = 2008-11-17 (844 days)
• Westmere = 2010-01-07 (416 days)
• Sandy Bridge = 2011-01-09 (367 days)
• Ivy Bridge = 2012-04-29 (476 days)
• Haswell = 2013-06-04 (401 days)
• Broadwell = 2015-06-02 (728 days)
• Skylake = 2015-08-05 (64 days)
• Kaby Lake = 2017-01-03 (517 days)

 

 

 

 

Also, I don't think the newer gen's IPC "bumps", where they exist at all, are worthy of being called a new gen / new performance step.   (For example, Haswell -> Broadwell -> Skylake -> Kaby Lake.)

 

Examples of per-generation IPC jumps I think would be steps in the right direction include:

 

Quote

The performance increase of the 80286 over the 8086 (or 8088) could be more than 100% per clock cycle in many programs (i.e. a doubled performance at the same clock speed). This was a large increase, fully comparable to the speed improvements around a decade later when the i486 (1989) or the original Pentium (1993) were introduced.

 

I was about to use Sandy Bridge as another example also, but...

Quote

The average performance increase, according to IXBT Labs and Semi Accurate as well as many other benchmarking sites, at clock to clock is 11.3% compared to the Nehalem Generation, which includes Bloomfield, Clarkdale, and Lynnfield processors.

That's from the Sandy Bridge page on Wikipedia.  I thought I'd heard some people say it had a 30-35% boost in IPC, though?  Or is some OC involved, or are they comparing Sandy Bridge with things like Core 2, instead of Westmere?

 

 

Spoiler

Personally, my preference for IPC boosts per generation would be something like ...

 

the new generation's lowest Celeron/Atom -Y SKU, single-threaded, set to lowest multiplier and BCLK,

should be faster than

the previous generation's top Xeon E7-88xx, 8-CPU LN2-OC'd multi-threaded performance.

 

But unfortunately I don't think we'll see jumps quite THAT big per generation.  I could settle for the > 100% per generation clock-for-clock single-threaded IPC mentioned above, though.(Anyone know if it's possible to link to a specific spot in my post, or is that only for when you're making multi-post guides with links?)  I'm willing to wait longer between generations, like the release intervals back in the 1980s & 1990s. (Like 4 years, instead of like 1 year.)  

 

 

I'd at least like Intel to catch up to where they would be if that 100%-per-generation trend hadn't slowed. (In this case, I'd count each codename, die shrink, etc, as a generation, not the general steps I gave in the list near the beginning of my post.)

 

That, and catch back up to the improved price-performance trend we used to see.

 

My dad bought a 286-10 platform for $940 (board, CPU, RAM) in January 1989, and sometime around 1995, bought a 486 DX4-120 + board + RAM for around $300-330 or so.  Wikipedia's MIPS page lists the 286-12 as 1.28 MIPS, and the DX4-100 at 70 MIPS, a 55 TIMES (not percent) boost in performance, for about 1/3 the price, over about 6 years or so.  (Extrapolating the DX4 for 120 MHz, and the 286-10, would be 78.75 times.)

 

Does the i7-7700K (2017, $350) or i3-7100 ($117) have 78.75 times faster single-threaded performance than the i7-990X (2011, $999) or i7-2600K ($317), at stock settings?

 

No?

 

I didn't think so.  FeelsBadMan

 

 

AMD, knock it out of the park, please.   I'm excited for Ryzen, and may be considering upgrading to the R7-1700X from my 4790K.

 

I'm not seeing Ryzen boards announced yet like what I'd want, though.  

Spoiler

 

Similar existing Intel examples are the ASRock X99 WS / WS-E, ASUS Z10PE-D8/16 WS, etc, preferably with the "AMD undercutting Intel" discount, plus a few extra SATA/SAS ports, M.2 slots, etc.  (I might not need as many USB ports.)  I was also briefly looking at a few EVGA boards, and a few 2 and 4-socket Supermicro boards, but haven't figured out yet what features I'd want from which ones of those brands, in a Ryzen based board.

I wouldn't even touch the Ryzen equivalents of the mentioned ASUS boards, if they're not discounted at least a similar percentage as the 1700 or 1700X is discounted from the 5960X/6900K/6950X.  My budget for the Ryzen board would, I'm anticipating, be $250-350 or so, which might put the ASUS and SuperMicro equivalents out of reach anyway.

 

 

Hopefully the 2018, 2019, 2020 etc. versions of Zen+ or whatever they're called can also be big jumps.  

 

Spoiler

That way my motherboard could last longer, as I don't want to upgrade the CPU without a significant bump in performance.  And, I don't want to replace the motherboard, case or PSU until after the warranty on the PSU expires.  (If I don't carry my AX760 over but instead sell it, I'd likely be considering a PSU with a 10-year warranty.)

 

[LAwLz]

10 hours ago, PianoPlayer88Key said:

That's from the Sandy Bridge page on Wikipedia.  I thought I'd heard some people say it had a 30-35% boost in IPC, though?  Or is some OC involved, or are they comparing Sandy Bridge with things like Core 2, instead of Westmere?

Seems like that's pretty skewed results on the Wikipedia page.

IPC is not a single measurement.

 

For example there was a ~26% IPC increase in Cinebench (4651 to 5860) when compared to bloomfield.

 

Don't think you will find many benchmarks, if any, which shows 30-35% increases, but Sandy Bridge easily performed that much better when you took the higher frequency into consideration. We went from ~3.4GHz chips once overclocked, to 4.4GHz overclocked (on the mainstream platform, where the i5 was 2.66GHz at stock).

[PianoPlayer88Key]

3 hours ago, LAwLz said:

Seems like that's pretty skewed results on the Wikipedia page.

IPC is not a single measurement.

 

For example there was a ~26% IPC increase in Cinebench (4651 to 5860) when compared to bloomfield.

 

Don't think you will find many benchmarks, if any, which shows 30-35% increases, but Sandy Bridge easily performed that much better when you took the higher frequency into consideration. We went from ~3.4GHz chips once overclocked, to 4.4GHz overclocked (on the mainstream platform, where the i5 was 2.66GHz at stock).

Hmm... I'm still trying to wrap my brain around some of it.

 

I did look it up, and apparently Intel had a step between Bloomfield (Nehalem, 45nm, LGA1366) and Sandy Bridge (32nm, LGA1155/2011).  From what I can tell, Westmere (Gulftown & Clarkdale) was the generation immediately before Sandy Bridge.

 

I see you mention different clock rates on the CPUs you mentioned.  I realize stock clocks were different, but I like to compare IPC at identical clocks.

 

If someone locked a SB i5-2xxx, i7-2xxx, a Westmere/Clarkdale i5-6xx, WM/Gulftown i7-9xx all to 1 core at the same clock speed, say, 3.2 GHz, with speedstep, turbo, hyperthreading where applicable disabled ...

... then what was the performance difference in that scenario between Westmere and Sandy Bridge?  Was it ~26%, or a lot less?

 

(Also were there no Westmere-based LGA1156 i7s, just Nehalem on LGA1156 or Westmere on LGA1366?  Or did I overlook them?)

[LAwLz]

3 hours ago, PianoPlayer88Key said:

... then what was the performance difference in that scenario between Westmere and Sandy Bridge?  Was it ~26%, or a lot less?

In some programs, it was 26% (clock for clock). In others, it was slightly more and in some it was less.

[PianoPlayer88Key]

4 hours ago, LAwLz said:

In some programs, it was 26% (clock for clock). In others, it was slightly more and in some it was less.

Ahh, so the average was 26%?

 

I've been looking at some of the older AMD benchmarks too, to see where they'd fallen to (primarily the FX series).  I don't have my data open right now, but if I remember right, the FX-x3xx series were about comparable, in single-threaded clock-for-clock IPC, to Nehalem and Westmere, for the most part, at least on some benchmarks.  Some I saw only barely beating Core, and I think I saw some almost coming close to Sandy Bridge.

 

While I'd love for AMD to leapfrog Intel's Lakes by as much as they had fallen behind, I'd be quite happy if they'd match them in IPC.   And really, I'd like to see the generational improvements that were seen in the 1990s (100% or more clock-for-clock in some cases, also price/performance, performance/core, etc) make a return - better yet, catch up to where we WOULD be if those hadn't slowed.

Or, since it's apparently gotten harder to make individual cores run a ton faster, could there be some way for legacy programs (even ones originally designed before multi-core CPUs existed) to take advantage of multiple threads (either by the CPU doing the magic preferably, or the end user running it through a recompiler, even though they don't have source code / aren't the original programmer, or something)?

[LAwLz]

23 hours ago, PianoPlayer88Key said:

Ahh, so the average was 26%?

I don't know what the average was, but in Cinebench it was 26%. The average might have been higher, and it might have been lower. 

 

 

I hope AMD makes a comeback too, but I am very pessimistic about hype because I have been burned before. So right now have the mentality that it will suck. It it does then I can say "told you to not get hyped" to everyone, and if Ryzen turns out to be good then I'll be pleasantly surprised. 

[leadeater]

On 2/18/2017 at 7:04 AM, PianoPlayer88Key said:

I did look it up, and apparently Intel had a step between Bloomfield (Nehalem, 45nm, LGA1366) and Sandy Bridge (32nm, LGA1155/2011).  From what I can tell, Westmere (Gulftown & Clarkdale) was the generation immediately before Sandy Bridge.

Correct, up to the i7-960 and including the 965 & 975 were Bloomfeild. The i7-970, 980, 908X and 990 were Gulftown, so were the single socket Xeons. The dual socket 5600 series Xeons were Westmere-EP and the first E7 dual/quad/octo socket Xeons were Westmere-EX.

[PianoPlayer88Key]

3 hours ago, LAwLz said:

I don't know what the average was, but in Cinebench it was 26%. The average might have been higher, and it might have been lower. 

Ahh, okay.  Even though it's not as much as the ~100+% boost from 8086 > 286, or each of 386 > 486 > P5, it's still more, I think, than each succeeding generation.  I'd even guess Sandy Bridge > Kaby Lake isn't as big of a jump in IPC.

 

Spoiler

User Benchmark says the 3.4 GHz i5-7500 is 21% faster SC mixed than the i5-2500.

Passmark puts the 2500 at 1871 and the 7500 at 2119, a 13.25% boost.

I tried AnandTech, but those weren't available, but the 2500K and 7600, and no Cinebench R15  results.

 

Hey, a question about past competition (GPU & CPU), in the spoiler.

 

Spoiler

I was looking recently at some CPU & GPU history, and noticed...

 

Okay, first, the question, then see the observed data below it.

 

Was AMD quite competitive with Nvidia and Intel at the times mentioned below?

 

On TechPowerUp GPU Database, some Nvidia launches were (descending date order) ...

 

• GTX 275 | 2009-01-15 | $249 | G200-105-B3 | 240 cores | 633 MHz | 896 MB GDDR3 | 448-bit
• GTX 295 | 2009-01-08 | $500 |  G200-401-B3 | 240*2 cores | 576*2 MHz | 896*2 MB GDDR3 | 448*2-bit
• GTX 285 | 2008-12-23 | $359 | G200-350-B3 | 240 cores | 648 MHz | 1024 MB GDDR3 | 512-bit
• GTX 260 Core 216 Rev. 2 | 2008-11-27 | $299 | G200-103-B2 | 216 cores | 576 MHz | 896 MB GDDR3 | 448-bit
• GTX 260 Core 216 | 2008-09-16 | $299 | G200-103-A2 | 216 cores | 576 MHz | 896 MB GDDR3 | 448-bit
• GTX 280 | 2008-06-16 | $649 | G200-350-B3 | 240 cores | 602 MHz | 1024 MB GDDR3 | 512-bit
• GTX 260 | 2008-06-16 | $449 | G200-100-A2 | 192 cores | 576 MHz | 896 MB GDDR3 | 448-bit

Notice how the GTX 260 launched at $449 in June '08, then the Core 216 version at $299 in September.  Then, the 275 at $249 in January '09.

 

Also, the GTX 280 launched at $649 in June, then the 285 at $359 in December.  There was also the 295 at $500 in January.

 

 

Moving over to CPUs ...

 

Q6600 started at $851, quickly dropped to $530, then $266, ending up around $183 or so.

 

To quote Wikipedia's Kentsfield (microprocessor) page:

 

Quote

The mainstream Core 2 Quad Q6600, clocked at 2.4 GHz, was launched on January 8, 2007 at US$851 (reduced to US$530 on April 7, 2007). July 22, 2007 marked the release of the Q6700, and Extreme QX6850 Kentsfields at US$530 and US$999 respectively along with a further price reduction of the Q6600 to US$266.

 

The CPU-World Q6600 page (see the Historical Price Chart) is my main $183 end price source.

 

Quote

I hope AMD makes a comeback too, but I am very pessimistic about hype because I have been burned before. So right now have the mentality that it will suck. It it does then I can say "told you to not get hyped" to everyone, and if Ryzen turns out to be good then I'll be pleasantly surprised. 

 

I'm hoping it's really good, but keeping that in check.  Therefore, I'm not yet selling my current 4790K platform just yet.  (More specs in spoiler)

Spoiler

Motherboard: ASRock Z97 Extreme6

CPU: Intel Core i7-4790K

CPU Cooler: Cooler Master Hyper 212 Evo

RAM: 32GB (8x2 x2) G.Skill Ares Red DDR3-1600 CL9

Case: Rosewill Thor V2

(I may, if I sell to family/friend, put it in like a Fractal Design Define R5 or Arc Midi R2, & sell the Thor V2 separately.  I don't really like that case anymore, and don't want to keep it for my next system.  If I sell to my dad, who has a Dell D830 (C2D T7250, 2GB DDR2-667, XP Pro, 5-yr-old WD5000BPKT), the Rosewill won't fit in his computer desk, but the Fractals will.)

PSU: Corsair AX760

SSD: 2.5" 256GB Crucial M550

OS: Windows 10 Pro (upgraded from 7 Professional)

 

I'm split on keeping the PSU, may keep the EVGA SC GTX 1060 3GB, and will keep the newer spinning HDDs.  (Older, out-of-warranty HDDs, once I finish pulling stuff off, may be sold, donated, or scrapped.)

If I sell the PSU, I'd want one with a 10-year warranty.  I'd also want to keep the case & motherboard that long, or until the PSU dies / is no longer usable, if it significantly outlives the warranty.

 

AM4 may only last 4-5 years, though.  My AX760 is 2 years old, bringing it to ~7 years (its warranty) at end of AM4.  I'd likely get the last decent AM4 CPU, maybe extending the platform's life further.  If I went to AM4, I'd keep the AX760 (also if I stayed with LGA1150).  Not sure about the AX760 if I go X299/LGA2066.

 

Other things slowing my upgrade to Ryzen are not having 8+ DIMM slots, not supporting 128 or 256GB RAM (not yet needed, but later), not having 10 or 12 or more SATA/optional SAS ports (I have lots of HDDs), and not having 6 or more PCI-E x16 slots.

 

A few boards for which I'd like an AM4 equivalent are the ASRock X99 WS-E, Extreme11, or ASUS Z10PE-D16 WS, with the AMD discount (Same percentage as 6900K/6950X to 1700X/1800X).

 

Otoh, the 32GB RAM is really cramping my style (where did that expression originate?), and 4K H.264 100Mbps videos only encode at ~5-6 fps.  (I don't have Premiere/Vegas.)

 

I'd really like faster than realtime (>30fps), but even a ~1.8-2.2x boost at similar pricing would be a good jump.

 

I hate frequently replacing motheboards, it's so labor intensive!!

 

Spoiler

 

I'll use expansion cards to add not-integrated or insufficient features/connectors, and upgrade the CPU every 5 or so years.

 

The 2nd to last CPU upgrade would likely be for $300-400 the year the PSU warranty expires (and hopefuly last gen on socket, but AM4 won't last the 10 years I would have liked.

The last CPU upgrade would be to a (or 2x, maybe 4x?) CPUs from the same generation that had been $5-8K to start, but are now ~$100-200 on fleabay.

 

Then when the time comes, replace the board, case & PSU simultaneously.

 

[leadeater]

22 minutes ago, PianoPlayer88Key said:

Was AMD quite competitive with Nvidia and Intel at the times mentioned below?

 

On TechPowerUp GPU Database, some Nvidia launches were (descending date order) ...

 

• GTX 275 | 2009-01-15 | $249 | G200-105-B3 | 240 cores | 633 MHz | 896 MB GDDR3 | 448-bit
• GTX 295 | 2009-01-08 | $500 |  G200-401-B3 | 240*2 cores | 576*2 MHz | 896*2 MB GDDR3 | 448*2-bit
• GTX 285 | 2008-12-23 | $359 | G200-350-B3 | 240 cores | 648 MHz | 1024 MB GDDR3 | 512-bit
• GTX 260 Core 216 Rev. 2 | 2008-11-27 | $299 | G200-103-B2 | 216 cores | 576 MHz | 896 MB GDDR3 | 448-bit
• GTX 260 Core 216 | 2008-09-16 | $299 | G200-103-A2 | 216 cores | 576 MHz | 896 MB GDDR3 | 448-bit
• GTX 280 | 2008-06-16 | $649 | G200-350-B3 | 240 cores | 602 MHz | 1024 MB GDDR3 | 512-bit
• GTX 260 | 2008-06-16 | $449 | G200-100-A2 | 192 cores | 576 MHz | 896 MB GDDR3 | 448-bit

Notice how the GTX 260 launched at $449 in June '08, then the Core 216 version at $299 in September.  Then, the 275 at $249 in January '09.

 

Also, the GTX 280 launched at $649 in June, then the 285 at $359 in December.  There was also the 295 at $500 in January.

Yes at this time it was still ATI and this is around when the HD 4000 series was released which were very good, June 2008. Then later the HD 5000 which was even better and very popular, September 2009.

 

At this time AMD was dead in the water CPU wise. Once Core 2 Duo hit it's stride that was it, then Core 2 Quad came out and sealed AMD's fate even more.

[PianoPlayer88Key]

44 minutes ago, leadeater said:

Yes at this time it was still ATI and this is around when the HD 4000 series was released which were very good, June 2008. Then later the HD 5000 which was even better and very popular, September 2009.

 

At this time AMD was dead in the water CPU wise. Once Core 2 Duo hit it's stride that was it, then Core 2 Quad came out and sealed AMD's fate even more.

Ahh, so ATI on the GPUs then. 

 

If AMD was already lagging behind Intel by the time Core came out, then...

 

1. What explains the price slashes on the Q6600?
2. Why haven't there been similar-percentage drops on Intel CPUs since then, and
3. What would it take for Intel to do price drops like that again (or better yet, price drops like older LGA1366 & LGA771, or even LGA2011 Sandy Bridge Xeons have seen on ebay)?  AMD being competitive again?

Also,

 

1. What time frame / CPU generations was AMD competitive with Intel, and,
2. Has AMD ever been ahead of Intel at a particular time, by how far Bulldozer/etc has fallen behind, say, Sky/Kaby Lake?

 

I really would like to see a return to the per-core performance, clock-for-clock single-threaded IPC, price/performance per year/generation, etc. improvements seen in the 1980s and 1990s.   Ideally, catch up to where we would be if the trend had never slowed.   (Or, if those gains in single-threaded performance aren't possible, gains in multi-threaded performance in such a way so that legacy applications wouldn't have to be rewritten, but could be easily adapted by the end user.)

[smokefest]

On 13/02/2017 at 8:55 PM, tp95112 said:

Is it me or is single threaded performance kinda low

reminds me of the FX time lolz

[leadeater]

17 minutes ago, PianoPlayer88Key said:

If AMD was already lagging behind Intel by the time Core came out, then...

 

1. What explains the price slashes on the Q6600?
2. Why haven't there been similar-percentage drops on Intel CPUs since then, and
3. What would it take for Intel to do price drops like that again (or better yet, price drops like older LGA1366 & LGA771, or even LGA2011 Sandy Bridge Xeons have seen on ebay)?  AMD being competitive again?

 

1. Low sales and low demand for quad cores. AMD at that time was still fine, wasn't that far off Intel but not the top end option. Intel was still figuring out where quad cores fit in to the market and how to price them and have issues with their own dual cores performing better in single threaded games.
2. Intel figuring out how to segment the market removing the need to lower prices as they get it right on launch then slowly lower the price.
3. Price drops will never happen like that again until there is a similar fundamental change in the CPU market, both dual cores and quad cores completely changed the market so everything had to start from scratch; pricing, marketing, benchmarking etc. This is all sorted now and 6, 8, 10+ core CPUs don't change things much, plus Intel already sells CPUs with massive core counts.

17 minutes ago, PianoPlayer88Key said:

1. What time frame / CPU generations was AMD competitive with Intel, and,
2. Has AMD ever been ahead of Intel at a particular time, by how far Bulldozer/etc has fallen behind, say, Sky/Kaby Lake?

 

1. Pentium 3, Pentium 4 and very early Core 2 Duo (Conroe).
2. The original AMD socket 939 FX CPUs were amazing, between 5%-25% faster than Intel's best and the higher end of that scale for gaming. http://www.anandtech.com/show/1722/2

[Lays]

On 2/13/2017 at 5:51 PM, Jed M said:

I hope this is true because I am holding off spending ~$8k on an Intel build on hopes that this shit won't bomb.

 

You'll literally get more performance in the long run by building a solid 2-3k build, and spending the rest on upgrading & selling your old GPU's when new TOTL cards come out.

[Bouzoo]

On 2/16/2017 at 11:43 PM, MageTank said:

Yeah, plenty of manufacturers have advanced turbo options to ignore boost tables. I've also yet to see anyone complain about stability when using these features, so it's pretty good. Some higher end laptops have the same thing, though they mostly require some sort of software or "gaming mode" enabled. Worked on an ROG laptop that had such a feature. 

I see. Always wondered why my 4700HQ which has 3.4 GHz turbo boost manages to go 3.4 GHz on all 4 cores, if we're to believe Speccy. 

[Trik'Stari]

On 2/13/2017 at 8:51 PM, Jed M said:

I hope this is true because I am holding off spending ~$8k on an Intel build on hopes that this shit won't bomb.

I'm basically doing the same.

 

Just bought a freaking Caselabs SMA8 Magnum. I'd like to spend a good amount of my budget on water cooling, but if AMD fucks this up, then I'll have no choice but to go with X99, and judging by the reviews, X99 is a difficult platform to get good working parts for.

 

Unless you go with ASRock apparently, and get stuck with a color scheme that looks like @LinusTech himself came up with it.

[PianoPlayer88Key]

13 minutes ago, leadeater said:

 

1. Low sales and low demand for quad cores. AMD at that time was still fine, wasn't that far off Intel but not the top end option. Intel was still figuring out where quad cores fit in to the market and how to price them and have issues with their own dual cores performing better in single threaded games.
2. Intel figuring out how to segment the market removing the need to lower prices as they get it right on launch then slowly lower the price.
3. Price drops will never happen like that again until there is a similar fundamental change in the CPU market, both dual cores and quad cores completely changed the market so everything had to start from scratch; pricing, marketing, benchmarking etc. This is all sorted now and 6, 8, 10+ core CPUs don't change things much, plus Intel already sells CPUs with massive core counts.

 

1 - Ahh, I hadn't thought of that.

 

2 - that reminds me of a few other things.

Spoiler

• I'd prefer Intel to not release CPUs that are slower than any of their own CPUs already released.  (I almost added slower than competitors, but backed off for now.)  I'd even like to take it so far, that a new CPU's single-threaded performance should exceed a pre-existing multi-socket CPU's multi-socket, multi-threaded performance.
• Rather than discontinue an entire line of CPUs and introduce an entire line, instead, keep the older products, lower their prices, and add a new product at the top of the stack.  For example, if they'd been doing that, an example lineup currently available in February 2017, from bottom to top (all compatible with the same socket), might be
  Spoiler

   

  • Q6600 - $35 (G1820 Micro Center Black Friday 2014/15 price)
  • Q9650 - $50 (G3258 MC BF $)
  • i7-920 - $100 (i3-4130/6100 MC BF $)
  • i7-3930K - $180 (i5-4690K MC BF $)
  • i7-5960X - $250 (i7-4790K MC BF $)
  • E5-2687WK v3 - $500
  • E7-8894X v4 - $1000
  • (or something like that.  Some adjustments could be made here and there, like substituting different SKUs, but keep the max price at $1000 for server-grade and $250 for consumer-grade CPUs.)

   

  • (Apparently you can do nested spoilers. )
• Instead of changing motherboard sockets so often, introducing new motherboards so frequently, etc, introduce (whether Intel does, or 3rd party makers) expansion cards to add new features.  Use the warranty length of top-end PSUs (currently 10 years) as a guide to when to change motherboards.

 

 

3 - Feel free to take a look at some of the PC-related invoices I've dug up, edited (personal info out) and posted in Google Photos.  They should be sorted chronologically.  There are a few missing, like my dad's 2008 laptop, some software he bought around 1989 or 1990, and a few of my more recent purchases.  (Some of my bro's early purchases are included, but not any of his recent ones when he's been living on his own.)
 

Spoiler

• Compare the 1989 Datel system, with several of the next invoices, to & including the AMD DX4-120 from Chip Merchant.  How would you say the price drops / performance increases over time compare to the way things have been since Intel last buried AMD (either since the Q6600 or Sandy Bridge)?
• How high-end of a system might someone get today, or in a few weeks, for the adjusted-for-inflation price my dad paid for the 1989 system?  (I'm guessing an i7-6950X and Titan XP SLI would be a bit of a stretch, but maybe at least a 6900K and a 1080, or an 1800X and 1080 SLI?)  Could he get something that would still be as usable in like 10 or more years, as something like an i7-920 and GTX 285 + SSD are today?  (He pretty much never touches 3D games.)
• The system from PC Club in 2002 was my dad's most recent desktop.
• The Athlon 64 X2 system from Newegg in 2008 was my first desktop.  My dad's desktop was dying at the time, prompting me to get that.  (It would only run like 10 minutes before shutting down.  I could coax 30 or so minutes at a time by underclocking it from 1.4 to 1.0 GHz, which made it possible to do the online shopping.  Otherwise the PC would barely even boot into Windows before shutting down, maybe open a website or two.)
• Yes, we still had a 5.25" floppy drive in 2008, but no way to hook it up.  (I think we'd long since gotten rid of the 40MB Seagate HDD from 1989, but the FDD was from the same '286 system.
• Also our Panasonic KX-P1080i or whatever model it was printer from around the same time frame, I believe, lasted longer than almost anything else computer-related we had.  I think our next printer was an HP Deskjet 6122 (there may have been one before - a Deskjet 890 or something iirc?), and now we have an Officejet Pro 8600 Plus.
• That Athlon 64 X2 system's mobo died around March/April 2012, so I used my dad's Dell D830 laptop (invoice not in the above link; I've seen the packing list some time ago but it mentioned no prices) until I got my 4790K system.  Had I not been practically broke, I likely would have gotten something like an i5-2500K system.
• And maybe it was that A64x2's time to go out to pasture, anyway.  It was getting pretty dusty, could that have killed it?
• Up until I think a couple years ago or so, my brother's Pentium II system was sitting unused in our living room.  (He took it home, and idk what he's done with it or if he still has it.  I'm sure it wouldn't FireStrike, DOOM 2016, GTA V, Rise of the Tomb Raider, etc, but I wonder how much faster my 4790K + 1060 is than that system in the same games/applications/benchmarks?)
• (Also how might someone directly compare the performance of a current HEDT/server system to my dad's first PC?)

 

 

13 minutes ago, leadeater said:

1. Pentium 3, Pentium 4 and very early Core 2 Duo (Conroe).
2. The original AMD socket 939 FX CPUs were amazing, between 5%-25% faster than Intel's best and the higher end of that scale for gaming. http://www.anandtech.com/show/1722/2

 

Ahh, interesting.  I'll have to look into those more sometime.

Somewhere I also have some boxes of old ComputorEdge (San Diego) magazines, which came out weekly and gave prices at local shops.  I've recently come across ones from the 2000s or so, maybe back to the mid 1990s.  (I also remember having ones as far back as 1987, but idk where they are now, or if I still have them; I hope I do somewhere.)

Someday (might not be soon though as I have other things going on), I'd like to look through some of them, and chart the price drops over time on various parts.  Also, I'd like to find benchmarks for those older parts, and have some way to directly compare them to modern 2017 systems, but I'm not sure where to go for that.

 

 

Btw ... am I, or my dad, the only non-business customers who will hang onto our PCs for almost forever?  (He's STILL using the previously-mentioned (in a spoiler) Dell Core2Duo, 2GB RAM laptop with XP!)

Spoiler

• For me, if I'd had it, I'd probably still be using something as old as a Q6600, although in that case I'd definitely upgrade to Ryzen practically on launch day, if I hadn't already upgraded to like an i7-5820K or 6800K.
• If I'd had like an i7-920, or an i7-2600K, then Ryzen would likely be my next upgrade, assuming the hype is validated.
• Looking back, if I'd been buying computers then (and old enough to do so; btw I was born in 1981 so the early ones aren't actually applicable in my case) ... my likely upgrade cycle might have been something like
  • 8086
  • 386
  • Pentium? (P5 or MMX)
  • Pentium III (or, if no Pentium 1, then Pentium II)
  • Q6600 or Q8xxx/Q9xxx
  • i7-920? or 2600K?
  • Ryzen 1700X/1800X or LGA2066/X299

 

 

I hope putting some things in spoilers helped make the post not look so long.

Btw, why do posts look so much "longer" on my phone (take up a larger part of a screen, or more screens) than on my laptop or desktop, in spite of my phone being 2560x1440 resolution & my PCs being 1920x1080?