bogami3/8 unsuccessful cut silicon chip is a waste product which lifespan is therefore highly questionable !
And it is propagated as a new product .Horror offering us a the garbage in the stars .

You do realize that all (yes, all) chips are not "fully enabled" per se. They all have some redundant structures to counter production faults. Even a full Titan X or Fury X will have a few unused redundant array structures. Salvaging chips reduces waste, and is orthogonal to product lifespan. Please stop posting such nonsense.

bogami3/8 unsuccessful cut silicon chip is a waste product which lifespan is therefore highly questionable !
And it is propagated as a new product .Horror offering us a the garbage in the stars . link has nothing to do with the article !

You think selling salvage parts for revenue has nothing to do with the financial health of a company, and doesn't impact their growth? Here, allow me to simplify the equation;

CasecutterHow far apart do you think they are over a mix of like 10 titles?

950 wins by 15% over 370? More if you overclock.

rruff950 wins by 15% over 370? More if you overclock.

I’ve no issue to you maintaining such an belief, although can you provide proof of such a situation that proves your assertion?

CasecutterI’ve no issue to you maintaining such an belief, although can you provide proof of such a situation that proves your assertion?

Last Nov I looked up every benchmark and review I could find on these cards, and that is what it averaged to. You can do the work if you want.

rruffLast Nov I looked up every benchmark and review I could find

I'm not doubting that or those reviews, but how many are testing using i5-6400 2.7 Ghz (up to 3.30GHz) working 8Gb single channel DDR3. I grant you reviews from after 950's Aug release would back your belief, though most are using enthusiast test rigs not some boring OEM or other starter box most who are buy in this $120 budget are going to make do with. Then more often run some max or ultra settling as that fits their ongoing graph of their other cards tested. Then more often Xtreme Gaming 950 against an unknown (reference) 370. While one of the more resent Tom's December 21, 2015 still used Catalyst 15.7.1. a month after Crimson came out.

While I never said the 370 best a 950 all the time. It just in in the "real world" the spread isn't all that... Sure if you want to play Witcher, GTAV, Metro LL, it's the 950, play more FC4, Shadow of Mordor, Watch Dogs, a 370 is more than qualified.

And as most of the folk who making due with some boring OEM or other starter box, such are hardly working a decent PSU let alone one that would offer the watts to consider any huge OC'n.

CasecutterI'm not doubting that or those reviews, but how many are testing using i5-6400 2.7 Ghz (up to 3.30GHz) working 8Gb single channel DDR3. I grant you reviews from after 950's Aug release would back your belief, though most are using enthusiast test rigs not some boring OEM or other starter box most who are buy in this $120 budget are going to make do with.

I think you'll find that the level of performance expected at this segment, you will still be graphics limited before the CPU achieves relevance. Most games don't tend to rely on the CPU+memory subsystem as the critical juncture unless the user's gaming centers upon RTS and sims where large maps, AI, and CPU physics are critical. Most people focused on these game titles are usually well aware that single channel memory isn't going to cut it.

CasecutterThen more often run some max or ultra settling as that fits their ongoing graph of their other cards tested. Then more often Xtreme Gaming 950 against an unknown (reference) 370. While one of the more resent Tom's December 21, 2015 still used Catalyst 15.7.1. a month after Crimson came out. While I never said the 370 best a 950 all the time. It just in in the "real world" the spread isn't all that... Sure if you want to play Witcher, GTAV, Metro LL, it's the 950, play more FC4, Shadow of Mordor, Watch Dogs, a 370 is more than qualified.

There is seldom a great spread in actual performance between vendors regardless of market segment at base clocks - I suspect that is by design. A bigger differentiator would be feature set and day one playability for most people - along with overclock headroom for a smaller percentage of users. While many people do not overclock, the fact that the card can be if desired does have relevance. The forums are littered with posts from first time OC'ers that after having their card for some time are swayed by interest (or necessity if they run into a particular game that requires more graphics power) in exploring overclocking.

CasecutterAnd as most of the folk who making due with some boring OEM or other starter box, such are hardly working a decent PSU let alone one that would offer the watts to consider any huge OC'n.

That is a really strange argument considering you've just referenced a Tom's Hardware article that places a high-OC GTX 950 with a lower power draw than an extremely mild OC (6.7%) XFX 370. I would note that the most standard GTX 950 SKUs retain the BIOS-locked 90W board limit, but even custom BIOS solutions in the 110-120W range barely begin to use the extra headroom ( 26% core OC to 1500MHz, 21% memory OC to 8000MHz adds 17W for example).

HumanSmokeMost people focused on these game titles are usually well aware that single channel memory isn't going to cut it.

Most people HERE focus on... But the folks looking at using some OEM box don't realize it, that the why I include it. It seems more of late cheap/deal machines are working in a single 8Gb stick with then two slots, Finding a matching stick is often prohibitive, or at least as much as 16Gb. The smart move is just dump the single stick and go 2x4Gb as that all such a build has use for.

CasecutterMost people HERE focus on... But the folks looking at using some OEM box don't realize it, that the why I include it. It seems more of late cheap/deal machines are working in a single 8Gb stick with then two slots, Finding a matching stick is often prohibitive, or at least as much as 16Gb. The smart move is just dump the single stick and go 2x4Gb as that all such a build has use for.

If these people are so unaware as to ignore (or be unaware of) their system information then everything is rather moot isn't it? How many people would be tech savvy enough to choose to upgrade a graphics card but ignorant enough to not know what was in their system? Unless they plan on teleporting the new graphics card into a PCIE slot I'm thinking that a single stick of RAM might be quite obvious the moment they looked at the internals. As for buying a new build, the argument remains the same: Dumb enough to not realize the system only has a single stick of RAM when it is specifically stated in most cases that I've seen, yet smart enough to compare graphics benchmarks on a tech site...and dumb enough not to be able to interpret the pro's and con's. I'm going to guess that most people buying an OEM box are 1. more sensitive to price, add-ons and monitor size, and the number of marketing bullet points - regardless of their actual worth, and 2. limited in what graphics options they can actually choose.

The difference here is that I'm looking at the consumer as a group, while you seem intent on finding some outlier corner case scenario that backs up your assertion. If you look hard enough you will find those outliers - but do you really think they'll be representative of the market?
Someone buys a single-channel RAM system after ignoring the specification and can rest assured that their informed purchase of a R7 360 is justified because the system can't extend itself to show a difference? Give me a break :laugh:. These people should take it to the next level and buy a single/dual core (no HT) processor and rest assured that their sub-$100 graphics is future proofed for the next decade.

CasecutterI'm not doubting that or those reviews, but how many are testing using i5-6400 2.7 Ghz (up to 3.30GHz) working 8Gb single channel DDR3....
And as most of the folk who making due with some boring OEM or other starter box, such are hardly working a decent PSU let alone one that would offer the watts to consider any huge OC'n.

So your contention is that the people who buy these cards will have such shitty systems that the card won't limit their graphic performance, but rather their system will?

Maybe with a 980 Ti, but not these. I haven't come close to maxing out my i3-4150 when my GTX 950 is at the limit. I probably could, but not with settings that make sense. You don't have to take my word for it, you can find benchmarks on more modest systems. The reason why high end systems are typically used in tests is so they don't restrict the high-end cards. A GTX 950 is only around 1/3rd of a GTX 980 Ti, so any normal newish desktop will be powerful enough to take full advantage of it.

I also measured the power draw at the wall and only get 175W for the whole system (including SSD, 2 HDs, and 16GB) with max OC.

rruff950 wins by 15% over 370? More if you overclock.

Umm, an R7 370 vs a GTX 950?



Its basically even with the 370 actually pulling ahead by 2%. I mean if were talking overclocking we would have to see what the results are at both max overclocks but as for stock its basically even with a little more in favor of the 370.

rruffI really wonder how many people they have working overtime to come up with names. Or do they pull them out of a hat? 940 and 940Ti would make sense. The 950 is nearly as good as the 960 and it got its own number.

Yea, I sometimes wonder from both sides who comes up with the naming structures. Its always dependent on chips and salvage parts and how they intend to put them into the lineup but sometimes its just a big WTF moment. This name is going to be very confusing in my book to some IMHO. I think if they were going to do this the other 950 should have been a TI variant and this should have been the regular 950 (But then again maybe they still plan to launch and actual 950ti).

The pre-W10 benchmark at TPU was 15% in favor of the GTX 950. That is in line with the average of all the others I've seen.





I don't know what the story is on those W10 benchmarks and the Nvidia 2GB cards, but there is something strange happening.

rruffI probably could, but not with settings that make sense. You don't have to take my word for it, you can find benchmarks on more modest systems.

Exactly, why I originally stipulated how/what I did. I'm always looking and don't see/find any reviews from such "OEM or other starter box" that would objectively pit a 950/370. Can you provide at least one?

CasecutterOn 1080p adjust settings to provide playable 50-60FpS (basically as [H] does) on the 950, then use those same settings for the 370 (Apples to Apples).

I honestly see this "SE" being dumped in the market to the same buyers who don't know the difference in what single channel means in performance and cost. Those same folks hear and see a GTX 950 in many reviews with more performance that a 370 and believe it's worth the extra. Heck I've had people say the bought the "SE" version because they judge more letters on the tail end of such stuff normally indicates a upgrade or better overall package.

As said in my scenario above the difference in the GTX 950 and R7 370 wouldn't be all that far apart. As the 950 was released to again shore-up the 960 in respect to the 380 looking better. I see neither (950/370) permitting any meaningful variance in actual performance between vendors at this market segment when clocks/cost being similar. In "real world" I see this "SE" above the R7 360, though will be able to appear in reviews as though it punches like a R7 370, however it will sit smack in the middle while priced like the R7 370.

It's been "a race to the bottom" for a while, which is why AMD just propagated the geldings of Bonaire and Pitcairn's, rather than spend money (they didn't have) for something that might be out of the market after 10 months from such rebranding. I'm actually surprised Nvidia has waited this long to bring this "SE". If out in March that perhaps means 4 months of sales before this will become to will become dated silicone, but that could be enough to at least keep the bulk from seeing the dumpster.

CasecutterExactly, why I originally stipulated how/what I did. I'm always looking and don't see/find any reviews from such "OEM or other starter box" that would objectively pit a 950/370. Can you provide at least one?

You make it sound like a task of Sisythus. I really don't know why you continue to persist given that the proof is fairly abundant and doesn't actually support what what you are saying (and BTW: The GTX 950 is using 12-17W less than the 370 in this benchmark)

CasecutterI honestly see this "SE" being dumped in the market to the same buyers who don't know the difference in what single channel means in performance and cost. Those same folks hear and see a GTX 950 in many reviews with more performance that a 370 and believe it's worth the extra.

If these people are reading reviews then they should have some kind of inkling regarding the difference between single channel and dual channel RAM since that is generally also highlighted in CPU/APU performance reviews. I might add that single channel RAM is predominantly found in AMD APU-based systems, so you are in effect saying that OEMs and AMD are targeting the tech uneducated. OEMs by skimping on features, AMD by not setting a minimum system requirement for their product.

CasecutterHeck I've had people say the bought the "SE" version because they judge more letters on the tail end of such stuff normally indicates a upgrade or better overall package.

Really? Considerng the same naming nomenclature has been used for over a dozen years almost continually? For example ATI released the X700 SE around four months after the LE version in April 2005. Same clocks, but the SE had half the pixel shader pipelines, half the TMU's, and 2/3rds of the vertex shader pipelines. Since then, the naming has been in near constant use. AMD transitioned from the use with the HD xxxx numbering system but you still find anomolies - early ones such as the HD 2350 and HD 2400 PRO being identical cards, and the R9 270X having nominally better performance thanks to a 50Mhz boost capacity over its otherwise identical successor, the R9 370X.

CasecutterAs said in my scenario above the difference in the GTX 950 and R7 370 wouldn't be all that far apart.

That could be said for virtually every comparison in card hierarchy, but despite this closeness in actual performance you go out of your way to highlight the AMD card every time - so obviously, even though the cards are similar enough, you find cause to choose one vendor 100% of the time. Isn't it conceivable that whatever compels you to make that choice (or at least defend the vendors reasoning for the SKUs inclusion), compels others to choose the other option?

CasecutterIt's been "a race to the bottom" for a while, which is why AMD just propagated the geldings of Bonaire and Pitcairn's, rather than spend money (they didn't have) for something that might be out of the market after 10 months from such rebranding.

Pitcairn has been in AMD's product line for four years through four generations of cards - I don't think a limited lifespan has been a factor up until now. The "Curacao" and "Trinidad" brand of the same silicon represent nothing more than pure profit after manufacturing thanks the not having to invest in R&D for the chip. Hard to fault AMD (or any vendor) who chooses to amortize R&D and extend profit in this way.

HumanSmoke(and BTW: The GTX 950 is using 12-17W less than the 370 in this benchmark)

GTA5 that's one titles I already said favors the 950 while again...

Casecutterlike having at most a 30W bulb on whenever you game over a month

Here’s that reviewfrom way back Aug 15, it has a MSI 950 Gaming with (1317MHz MSI Boost OC) while I'm having difficulty deducing what 370 was used. Though they show a single fan XFX which at best is clocked fairly pedestrianly at 995Mhz... More like a 950 with Boost Clock of ≤1220Mhz.

I think the newer information that GhostRyder provided showing the latest drivers, and wide range of titles is more apropos.

GhostRyderIts basically even with the 370 actually pulling ahead by 2%.

HumanSmokesingle channel RAM is predominantly found in AMD APU-based systems,

Wow spin much... I worked from a Desktop advertised a 6th Gen i5 Skylake, straddled with single channel DDR3. You come back that lowy APU (mostly Laptops) are single channel. The box I used was advertise as "Gaming"... though with an anemic GT 730.

HumanSmokeyou go out of your way to highlight the AMD card every time

And the same can be said about you for Nvidia, though this no longer as any debate, but now resorting to personal jabs. Way to hold the professional high ground.

HumanSmokeHard to fault AMD (or any vendor) who chooses to amortize R&D and extend profit in this way.

I'm not faulting either if the price is appropriate, though I'm here in support of consumers. When pricing is commensurate for what you get that’s fine. Here today the discrepancy is again back to basically 20% (I found a rare instance where there was similar price, although depended on a higher rebate). I just want people just entering into gaming hearing the "present-day" information that a 370 does spar with 950's, so this “SE” part should not be looked upon as any “deal” if it holds at a $120 MSRP. Perhaps with a rebate down bringing it down to $100 just packs in an already tight field. That’s not an the issue as long as folks have information to deuce and comprehend such nuances in the market segment.

CasecutterGTA5 that's one titles I already said favors the 950 while again...

Here’s that reviewfrom way back Aug 15, it has a MSI 950 Gaming with (1317MHz MSI Boost OC) while I'm having difficulty deducing what 370 was used. Though they show a single fan XFX which at best is clocked fairly pedestrianly at 995Mhz... More like a 950 with Boost Clock of ≤1220Mhz.

You can't work out what clock the XFX 370 was running at? From the review you just referenced:

Meanwhile, the R7 370 offers a noticeably lower performance level, to the point where the majority of the run occurs at a sub-60fps frame-rate. If Nvidia set out with the objective of beating the Radeon card, it has succeeded - but we feel that the factory overclock emphasises the differential. MSI has added 103MHz to the base clock here (and a similar amount to its GTX 960 too), and just 55MHz to the Radeon

Reference clock for the 370 is 975MHz. XFX's choices for anything approaching a 55MHz bump on base clock are the1040MHz Double Dissipations(actual 65MHz). Actual overclock over reference: 6.67% compared to the nominal 9.96% base/10.77% boost of the MSI card.
Oh, and just for the record...the video I posted actually comes from the same review you just linked to

CasecutterI think the newer information that GhostRyder provided showing the latest drivers, and wide range of titles is more apropos.

Yet your whole argument up until now focuses on the card being paired with a processor more in line with the price segment....

CasecutterExactly, why I originally stipulated how/what I did. I'm always looking and don't see/find any reviews from such "OEM or other starter box" that would objectively pit a 950/370. Can you provide at least one?

...now when those lower cost processors are paired with the lower cost cards as per your original argument falls short (maybe you should actually read the Eurogamer link you supplied) you change the paradigm to NOW compare results obtained with an i7-6700K overclocked to 4.5GHz and 16GB of DDR4-3000. So you abandon your whole "OEM starter box" argument for high end test system when and where it suits your argument. Colour me unsurprised.
And the latest performance graph from a TPU indicates a 3% advantage to the GTX 950...
...2% advantage to the R7 370 is significant enough for you to highlight. Surely 3% advantage to the GTX 950 now also deserves the same same significance?

CasecutterWow spin much... I worked from a Desktop advertised a 6th Gen i5 Skylake, straddled with single channel DDR3. You come back that lowy APU (mostly Laptops) are single channel. The box I used was advertise as "Gaming"... though with an anemic GT 730.

Note the word predominantly:

single channel RAM is predominantly found in AMD APU-based systems

I've already canvassed the available systems from the larger OEMs (Dell, HP, Asus, Acer) and the larger share of budget OEM "starter boxes" featuring single channel RAM compared to the range as a whole are those paired with an AMD APU. Care to take a wager?

CasecutterAnd the same can be said about you for Nvidia, though this no longer as any debate, but now resorting to personal jabs. Way to hold the professional high ground.

If I post recommending an Nvidia card over an AMD then it should be relatively easy to find proof of your assertion. You can do so, or you can ignore it and be shown to be trolling. I'll leave it with you. Personally I think it will be the latter.

CasecutterI'm not faulting either if the price is appropriate, though I'm here in support of consumers. When pricing is commensurate for what you get that’s fine. Here today the discrepancy is again back to basically 20% (I found a rare instance where there was similar price, although depended on a higher rebate). I just want people just entering into gaming hearing the "present-day" information that a 370 does spar with 950's, so this “SE” part should not be looked upon as any “deal” if it holds at a $120 MSRP. Perhaps with a rebate down bringing it down to $100 just packs in an already tight field. That’s not an the issue as long as folks have information to deuce and comprehend such nuances in the market segment.

1. Nowhere in the article does it mention the cards MSRP, so assuming a $120 price tag and then basing an argument against your own assumptive pricing is unsound.
2. Depending on performance the pricing will dovetail with existing SKUs from both vendors. You hold TPU's review in high regard judging by your willingness to back the graph that Ghostryder provided, then maybe you should holdthis one in just as high esteem.

CasecutterGTA5 that's one titles I already said favors the 950 while again...

Your claims are unsubstantiated. And they won't be.
The comparison chart that Ghost Rider posted has seriously effed up results like this factored in:



AFAIK, no one has gotten to the bottom of what is going on here, but when Wizzard upgraded the benchmark system with W10, the 2GB Nvidia cards had a serious problem in some games. I searched hard a few months back, but didn't see mention of it anywhere else. But IMO, it isn't fair to just average those in without knowing what is wrong. The chart I posted above showing a 15% advantage for the GTX 950 over the R7 370 was on the old system. This is the comparison for GTA5:



The GTX 950 also has a lot more OC headroom than the 370. Typically 20% over reference compared to <10% for the 370. Not surprising since the 370 is already an OC'd 265. TPU has tested a bunch of 950s and they all OC'd very well. www.techpowerup.com/reviews/EVGA/GTX_950_SSC/33.html

rruffYour claims are unsubstantiated. And they won't be.
The comparison chart that Ghost Rider posted has seriously effed up results like this factored in:



AFAIK, no one has gotten to the bottom of what is going on here, but when Wizzard upgraded the benchmark system with W10, the 2GB Nvidia cards had a serious problem in some games. I searched hard a few months back, but didn't see mention of it anywhere else. But IMO, it isn't fair to just average those in without knowing what is wrong. The chart I posted above showing a 15% advantage for the GTX 950 over the R7 370 was on the old system. This is the comparison for GTA5:



The GTX 950 also has a lot more OC headroom than the 370. Typically 20% over reference compared to <10% for the 370. Not surprising since the 370 is already an OC'd 265. TPU has tested a bunch of 950s and they all OC'd very well. www.techpowerup.com/reviews/EVGA/GTX_950_SSC/33.html

If your going to quote me at least get my name right LOL (Its a joke, not serious).
My point was that they are very close and not a 15% difference. Depending on the games it bounces back and fourth, if you want to do a comparison about max overclocks then you have to do it with the same tests on max overclocks. Numbers in general mean nothing on the clocks as each card reacts differently to higher core clocks.

HumanSmokeDepending on performance the pricing will dovetail with existing SKUs from both vendors. You hold TPU's review in high regard judging by your willingness to back the graph that Ghostryder provided, then maybe you should holdthis one in just as high esteem.

Well the only problem using those is prices change on a quick basis. Right now the GTX 950's cheapest variant (Not refurb or open box) is $149.99 at this moment (No rebates) while the R7 370 is at its lowest is $134.99 at least at the moment I am looking unless I missed one. Any who the only reason I posted the graph was to show the gap is not as big as was previously mentioned by rruff.

I think the point is with them both being so close they are likely to change spots on a whim because a new game gets added.

Nice two on one... using every argument (using one title) of this... that... and the other to feel good.

HumanSmokeOh, and just for the record...the video I posted actually comes

Ah why I said Here’s that review.

HumanSmokeReference clock for the 370 is 975MHz

Nope as per TPU D-Base925Mhz/975Mhz Boost, so it's running at 980Mhz but I didn't find a XFX at 980Mhz, they are calling a Boost Clock: 995 MHz so hard to know what he's talking about.

Using his stack of i3 results the 370 was ~10% behind a much nicer card in the MSI Gaming and that's back then. While several games are running in the 40FpS for 950, so that could've be optimized better IMO. If that was a 1040Mhz as you surmised, I'd say that 10% lead would've withered.

Exactly why I outline it the way I did. Your working to hard to substantiate my premise, "How far apart do you think they are over a mix of like 10 titles?" Exactly what W1zzard showed in his summary, it's so close it hardly matters. If you have an issue with the data, take it up with him it not mine!

Stop with the OC'n most entry gamers who work from boring OEM or other starter box are on some crap PSU less than a 400W making that point null. While sure $160 EVGA SSC has the head room though would a pedestrian card that's more inline with a budget offer that? If we're at $160 I'll move up two-rungs to this XFX 380and all this maneuvering is over.

megamanxtremeMy old desktop that my brother uses has the i3-540, 4GB RAM, and a GT 440.

I am thinking of getting him another 4GB of RAM and another card that performs better but doesn't exceed the PSU's power of 300W, altogether.
Since I let him use my STEAM library of games, he might have some AAA titles under his collar, so I hope the 950 SE is good enough for his games and stays within the PSU's limit.

Find a second hand 400W or 500W PSU, that thing is seriously gimping his options. They're so easy to change out..

Sorry about the name!

The 950 beats the 370 in all but one game, where it losses by 1FPS www.techpowerup.com/reviews/EVGA/GTX_950_SSC/6.html. It's 15% on average at reference clocks. That's with the W8 test, not the W10 where something is messed up. Unless you can show me that 10FPS in GTA5 at 1080p is a real thing in W10 and not a strange artifact of TPU's test system.

And if you look at the link I posted above, all the GTX 950s that TPU tested OC a lot. ~20% performance boost over reference. So in the real world it's about 25% difference if you OC both.

GhostRyderWell the only problem using those is prices change on a quick basis. Right now the GTX 950's cheapest variant (Not refurb or open box) is $149.99 at this moment (No rebates) while the R7 370 is at its lowest is $134.99 at least at the moment I am looking unless I missed one. Any who the only reason I posted the graph was to show the gap is not as big as was previously mentioned by rruff.

The SKUs up and down the product line are generally very close (Something I do not attribute to "luck"). A driver revision, a change in benchmarked games, even a game patch with unoptimized driver support will alter the results one way or the other. 2% favouring the R7 370 or 3% favouring the GTX 950 is well within margin of error numbers and very likely isn't discernible in a double-blind test. 12% is more noticeable, but like any benchmark review, one (and the condensed numbers especially) generally doesn't provide a global picture.

As for current pricing (or available pricing for any prospective buyer), relative worth is dependent upon the factors the customer values. Overclockability and any actual gains it can provide, AIB preference, warranty and support, aesthetics in AIB design, hardware and software feature set, and actual availability are all factors that weight differently for each person.

GhostRyderI think the point is with them both being so close they are likely to change spots on a whim because a new game gets added.

Undoubtedly - as would level of game i.q., game patches, and driver maturity. My particular point wasn't aimed at one card being the superior product - I said as much in my original post (#33) in the discussion - just that casecutters assertions that an overclocked GTX 950's power consumption rules it out in relation to a R7 370 for some people, and later, that the R7 370 is a much better option than the GTX 950 when paired with a lower-tier processor - neither assertion borne out of fact.
I think most people would see the cards as being a toss-up depending on what features they value. I certainly wouldn't advocate one being clearly superior to the other - although being an enthusiast my personal order of importance generally starts at max performance (hardware and driver).

CasecutterStop with the OC'n most entry gamers who work from boring OEM or other starter box are on some crap PSU less than a 400W making that point null.

Only in your mind. While a large percentage of people don't overclock, having a low end system does not preclude people from doing so. Here is an example that isn't an opinion dressed up as fact. HWBot's submissions represent a fraction of actual sales and a fraction of actual overclockers - since not everyone overclocking logs benchmarks, and of those that do, only a small percentage submit their results:
GTX 950 submissions: 1062 in six months
R9 380 submissions : 986 in eight months

CasecutterWhile sure $160 EVGA SSChas the head room though would a pedestrian card that's more inline with a budget offer that? If we're at $160 I'll move up two-rungs to this XFX 380and all this maneuvering is over.

So you have moved the goalposts yet again. The comparison being made concerned the 950 and 370.
As for the R9 380 comparison it certainly wouldn't work well for people in my country and maybe not others either, so caveats apply. Your argument regarding the 380 pricing is disingenuous at best since most of the 380's sit closer to $175-200 and sit in another pricing bracket to the 950/370. For the same basic price you can also buy a custom MSI GTX 960.

RyrynzFind a second hand 400W or 500W PSU, that thing is seriously gimping his options. They're so easy to change out..

I'm building him a new computer in 2018, so I just wanted a short upgrade, since that PC will go to the junk pile afterwards. I merely want to upgrade the GPU to do better than the GT 440, he still has my old CRT monitor that max at 1280X960, and might want to try BF3 and other AAA games. I just want something that can hold-off the 300W PSU.

Nv must have alot of defective 950 chips.

HumanSmokeSo you have moved the goalposts yet again. The comparison being made concerned the 950 and 370.

No you didn't read or just more desperate... The "Goalpost moved" outside of the stadium when the argument was made adding 30-35% to the $120 budget entry price. I found two cards that meet the criteria and the next thing "rruff" is talking about a $160 EVGA SSC.
And at that price it's the same story... as GTX 950 SSC isn't besting a nice 380 (the XFX I linked at the time was $163; include $3 shipping working a $20 rebate).