TR Forums

Does anyone else have a preference for buying the highest binned chips available? For example, Core i7 4790K as opposed to the cut down version, Core i5 4690K, even if HT is unnecessary for gaming and you only get a few hundred MHz increase? Or the GTX 1080 instead of the GTX 1080 Ti, even if you can afford the extra performance but you don't like the idea of a cut down chip regardless of how it performs, if you wanted GP102 you'd rather get a Titan XP but you can't afford that chip so why bother?

I know it seems silly, but I've got an aversion to any piece of hardware that may be artificially neutered in any way. Maybe sell me on why I shouldn't, or share your thoughts if you're like me.

Short answer: no.

If I were to buy a 1080Ti, then I'd buy it with the intention of keeping it until it died not just because it's faster than a 1080. The 1080Ti in particular isn't really a great example because it is significantly faster than the product placed immediately below it. For most of these products that's not the case, and you're paying a disproportionate amount more for a small improvement.

In the end it's all a compromise anyway. The super high end products we see are not the fastest thing possible the manufacturer could make, but just the fastest thing they think they could sell.

Unless you've got a "time is money" use case where you're doing something computationally intensive professionally, bleeding edge parts just aren't worth it. There's very much a "diminishing returns: effect. I tend to go for price/performance. I also ask myself "Will anything give me a meaningful bump while re-using mobo and/or DRAM I have on hand"?

This typically results in me being a generation or two behind the curve. For example: My primary home desktop (which runs Linux) is still an AMD FX-8350; I may upgrade it to a Ryzen sometime this year. I also have a gaming build (Windows) on deck, with an i5-7600K.

just brew it! wrote:

Unless you've got a "time is money" use case where you're doing something computationally intensive professionally, bleeding edge parts just aren't worth it. There's very much a "diminishing returns: effect. I tend to go for price/performance. I also ask myself "Will anything give me a meaningful bump while re-using mobo and/or DRAM I have on hand"?

This typically results in me being a generation or two behind the curve. For example: My primary home desktop (which runs Linux) is still an AMD FX-8350; I may upgrade it to a Ryzen sometime this year. I also have a gaming build (Windows) on deck, with an i5-7600K.

Well my aversion to cut down chips doesn't just mean buying the fastest thing on the market, just the fastest for that particular piece of hardware. Even then I'm not too worried about frequency, since it's usually pretty simple to overclock anything to the highest you can get retail. I have no use for Skylake-X chips or GV100 since they are simply overkill for what I do and score very very low on value charts. However, when chips are segmented by removing cores or features that could be fully functional for the sake of marketing and sales, I'd prefer those things remain intact. So even if I thought I needed 8 cores, I would probably go with something like a top end Ryzen chip because it's a "true" 8 core chip and not a cut down 10-core, for example. I don't know why, but it's just always bothered me to see products that are basically half dead being sold just to make numbers on a curve when they could have just lowered the price and let chips be segmented on price by supply and demand. I know some chips are really half dead and not artificially cut down, but that's not the only way chips are binned. Even if they were I'd prefer a fully functional product.

Ahh, OK. I hear you.

That sentiment is actually one of the things that has kept me on AMD for so long. Over they years they have tended to play the segmentation game less than Intel. E.g., Intel used to require that you step up to a Xeon to get hardware virtualization support, and still do for ECC (with a few exceptions). AMD enabled that stuff across the board (though they dropped ECC on the socket FMx APUs, and it is present but not "validated" for Ryzen).

AMD does play the "core disable" game (remember "Core Unlocker" motherboards?), but I tended to avoid those CPUs.

Progress in CPUs has been slow enough that it's something I consider a long-term purchase, so I tend to favor the less favorable ends of the performance/price curve. Since my workloads tend to favor fewer threads, I go for the highest single-core throughput I can afford. I just don't see much reason to save $20-30 on a i7-2600k vs the 2700k when I end up using for 6 years, and I certainly don't see a reason to buy something with significantly cut-down performance even for half the price for my main PC, like an i5-tier part: it takes me a day or two to make that money, but I'll probably be using the chip for 5+ years.

blargh4 wrote:

Progress in CPUs has been slow enough that it's something I consider a long-term purchase, so I tend to favor the less favorable ends of the performance/price curve. Since my workloads tend to favor fewer threads, I go for the highest single-core throughput I can afford. I just don't see much reason to save $20-30 on a i7-2600k vs the 2700k when I end up using for 6 years, and I certainly don't see a reason to buy something with significantly cut-down performance even for half the price for my main PC, like an i5-tier part: it takes me a day or two to make that money, but I'll probably be using the chip for 5+ years.

I think this also plays a part in my purchasing decisions. I would have no problem getting a cut down version of something I intend to upgrade near term, but I don't want a gimped chip that I plan to run until it dies and/or advancements become too tempting to ignore the itch to upgrade.

All the time.

I will buy the top end part, then the better performance lets me use it longer before I have to consider a replacement. I try to build in five year cycles. though the way most hardware has been going I'll keep using it past that unless some component has become a bottleneck. The parts may cost more up front, but I end up saving in the long run and I can enjoy the better performance for the life of the system.

There's always caveats, say buying a Titan Black versus a 780 Ti the cost wasn't worth the extra performance or prestige. But for the 1080 Ti over a 1080 I felt the cost was definitely worth the performance, especially since I will be using it for five years at the minimum.

The longer upgrade cycles can save in other ways, as well. I'm going to keep using my 4790K rig until DDR5 launches, which should be in two years. I plan to skip the overpriced DDR4 generation entirely, and hopefully the DDR5 stuff won't launch with a 3x markup that DDR4 has today.

Case in point: I am planning to build my daughter a new desktop, but I'm waiting for Raven Ridge. I could have already built something with the same capabilities, but Raven Ridge APUs seem like a more elegant solution that was engineered to do exactly what I need it to and should save me some money especially with the recent GPU prices. Maybe this is a bad example because I'm not sacrificing anything but time to make this decision, but it still satisfies my OCD-like thinking that other options involve pairing two cut down chips. I mean, I'm not so crazy about it that I'm going to make obviously dumb purchases, but I will sacrifice some performance or some value for the sake of buying fully unlocked chips, even if the unlocked features are unnecessary at the time or even if I could get more fps if I gave it up and sprung for a 1080 Ti I might enjoy the performance of.

The market segmentation game is kind of annoying. When there's new stuff being released sometimes I'll grumble about artificial segmentation, but when it comes to buy stuff I don't usually spend much time thinking about it. I suppose my hardware requirements are pretty run of the mill, so that might make it easier.

Your OCD-like thinking is far more understandable to me than that of the average fanboy who does.... whatever it is they do in order to continue being a fanboy. So that's a good thing, I guess.

Redocbew wrote:

Your OCD-like thinking is far more understandable to me than that of the average fanboy who does.... whatever it is they do in order to continue being a fanboy. So that's a good thing, I guess.

Well, I guess I'm a bit of a fanboy of Asus and Corsair, but that doesn't seem to get me into pointless arguments with others in this forum like being, say, an AMD fanboy. However, I think that brand loyalty I've had has worn off as I've collected enough products that weren't so great, especially when it comes to supporting software.

I'm a bit of an EVGA fanboy myself. I have a few of their video cards, and their power supplies, but it's those pointless arguments I don't understand. It's just weird. I get that the EVGA stuff I have isn't really any different than I could get elsewhere, and if their stuff took a dive that's where I'd look.

I was eyeballing a 1080Ti for a while, I guess for the same reasons as you mentioned earlier, but now I think I'll just wait until the next generation comes around.

Redocbew wrote:

I'm a bit of an EVGA fanboy myself. I have a few of their video cards, and their power supplies, but it's those pointless arguments I don't understand. It's just weird. I get that the EVGA stuff I have isn't really any different than I could get elsewhere, and if their stuff took a dive that's where I'd look.

I was eyeballing a 1080Ti for a while, I guess for the same reasons as you mentioned earlier, but now I think I'll just wait until the next generation comes around.

I think I'll just stick with the G*104 class of cards, since they are much lower power and seem to Trump the previous flagship every time. It's just like getting a more efficient version a year later for much cheaper. I like EVGA too since, unless you get a FTW version, they are a lot lighter and I always feel like the heavy cards are going to break the motherboard.

Sometimes, those gimped chips punch way above their weight class.
https://techreport.com/review/16382/amd ... cessors/14

JustAnEngineer wrote:

Sometimes, those gimped chips punch way above their weight class.
https://techreport.com/review/16382/amd ... cessors/14

Well that was 9 years ago. Even longer ago than that, most of the segmentation was all about clock speed. Back then, if you weren't afraid to overclock, you could get some killer deals by buying low end parts, but those days are over, it seems.

I look at it this way. I'm saving a reject chip from the dumpster. Just cuz the old boy got a hitch in his step doesn't mean he can't live out a happy life.

Obviously there are still some "cut down" chips that have perfectly functioning disabled resources in order to make enough supply stock, but I'd wager that the majority of them are truly crippled, so the manufacturer turns off that section and sells it at as a lower SKU. I know this happens frequently in GPUs, and CPUs are frequency binned at the very least.

I was on board when I thought this was about top shelf potato chips....

Heiwashin wrote:

I was on board when I thought this was about top shelf potato chips....

I got you man: viewtopic.php?f=1&t=98339&p=1225337