Dell Inspiron 3647 Small Desktop, by any chance?
I had an R7 250E Low Profile in mine. I'm given to understand that there are low-profile AND single-slot-cooler-height RX 6400 cards out there.
Dell Inspiron 3647 Small Desktop, by any chance?
I had an R7 250E Low Profile in mine. I'm given to understand that there are low-profile AND single-slot-cooler-height RX 6400 cards out there.
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
They announced an A310, but either it's an OEM-only product or never actually made it out the door:
Yeah, the A40 is supposedly a workstation product, which basically just means they slap a longer warranty on it and double the price:
Intel® Arc™ Pro A40 Graphics - Product Specifications | Intel
Intel® Arc™ Pro A40 Graphics quick reference with specifications, features, and technologies.
www.intel.com
FWIW, the 75 W cards I've seen are all either half-height or single-slot. Not both. Nvidia's single-slot, half-height Quadro cards typically have a TDP of no more than about 50 W.
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
Along comes a well-timed article...
Too bad its memory bandwidth and capacity were cut further.
I mean, that may be true, but, given the low power draw and probably very small form factor, if I were to make a rough guesstimate, I'd put it at about 2/3 to 3/4 the performance of the RX 6400, so around the GTX 1050 in performance, maybe even the GTX 1630. Without the GTX 1630's $200 price tag, and at less than half the GTX 1630's 75W power rating (though I wonder if it really ever hits that level of power draw)
I also imagine it would be inexpensive, if it were ever officially released in the US. Probably at or below the GT 1030 GDDR5's pre-crypto-craze price.
Last edited:
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
No way it's GT 1030-level. Remember, it's still got Infinity Cache. I think 75% of a RX 6400 is plausible.
Your mistake was using logic and reasoning regarding gpu's since nothing makes sense anymore now that the mining boom has ended but the high prices haven't.
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
They're selling below MSRP. How low do you expect them to get, especially when world-wide inflation is still pushing prices up?
The MSRP of all 4000 cards are too high and I have no idea how you can't see that. The 1080 ti was $700. The 3080 was $700. Those were decent value. With inflation, $700 is $800 now, not $1200. The 4080 is $1200 and might have dropped some to $1100, I haven't checked lately, but that price is so obviously ridiculous that if you disagre I would have to think there's something wrong with you. I've heard from a couple of sources that the 4070 ti costs about $400 to make, so $650 would be a fair price for it, but no, it's $800-$900 and there are a lot of people saying it's a good deal. I'm just about done even talking about this subject because I continually get pushback from people defending these high prices every day, including yourself. The crypto-mining boom is over but everybody seems to have gotten used to those prices and get mad when you say prices are too high. It's almost like lots of people enjoy spending a lot of money for something worth a lot less. They compare the 4070 ti's price to the 3090, just all sorts of stupid <Mod Edit>. I really don't care much anymore because I'm being outvoted and downvoted a lot - A LOT - because I want prices to go back to sane levels, but there are so many people against me for wanting that. Go ahead and see if I care anymore and you can spend almost $1000 on a 4070 ti and then go on forums and tell everybody what a great deal you got. Spread the message for and wide of how great prices are and that with inflation $1200 is a reasonable price for a 4080. Just go right ahead.
Last edited by a moderator:
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
I hear you. However, the sales data we're seeing suggests Nvidia has no problem selling them (except the RTX 4080, which is overpriced relative to its siblings). So, I'm not expecting them to drop. Not only that, but Nvidia has tons of demand for its H100's, which are made on the same wafers - meaning, Nvidia can just shift wafer supply, that they've already agreed to buy, over to more profitable AI-oriented products.
Since we all know how much Nvidia likes profits, do not expect to see them cut RTX 4000 prices for a long time. The lone exception might be the RTX 4080, although it probably won't even drop as much as it should.
If priced in accordance with the other RTX 4000 models, it should cost around $1050, by my math.
Two things to consider:
- New process nodes are more expensive.
- After seeing what prices the market is willing to bear, Nvidia might not have designed the current generation of products to sell at the same price point as prior generations. Even the inflation-adjusted equivalents.
First, you didn't restrict your comment specifically to the RTX 4000 series, or I'd have responded differently. Second, they'll sell them for as much as they can. And (4080 aside) seem to have no trouble doing so. Don't expect them to do you any favors.
What do those sources say prior generation GPUs @ the same tier cost to make, back when they launched?
How are you arriving at that markup? Are you aware that most of their engineering costs go towards software? That's right - the software for GPUs is actually more complex than the hardware. What's dismissively termed "drivers" includes compilers, firmware, host API stack, libraries, game-specific optimizations, etc. It's a lot that needs to be updated for each new generation, and maintained/enhanced through the products' lifetime. So, you really can't overlook the engineering costs, when you're trying to figure out the minimum they need to charge.
Then, there's the matter of profits. Nvidia and its investors really like profits. Always expect them to behave in a profit-maximizing fashion, regardless of whatever you think seems fair.
Be clear about what you're complaining about. If you're mad because you can't have the GPU you want at the price you want, then maybe you're part of the problem. You either need to be more flexible about which GPU you buy, or how much you're willing to pay for it. You lack the leverage to be picky on both counts (i.e. something about having your cake and eating it...).
If you don't like Nvidia GPUs being so expensive, I guess you should first complain to people who still buy them. However, the less revenue Nvidia makes on gaming could simply convince them to go all-in on AI, robotics, and self-driving.
The current GPUs were designed back when prices were persistently high. It's understandable if Nvidia and AMD took that as a signal that they could design more expensive GPUs.
I got a 3080 last october for $450 and I typing this with it installed now. Best tech deal I've gotten in a long time. I can wait out this cash grab for 5 years if I need to at 1440p, no problem. The thing about your detailed post was that you are doing everything you possibly can to justify high prices. Your argument about higher prices for advanced nodes is <Mod Edit> because the die sizes are a lot smaller and there are higher yields now so the price increase for tsmc isn't as high as you're trying to convince me that it is. You know prices are too high, yet you come up up with excuses for nvidia and amd to charge outrageous prices, and you've spent a lot of time on this to come up with factoids that you can manipulate just enough to make your argument sound good. <Insult removed by Moderator>
Last edited by a moderator:
helper800
Judicious
- Aug 12, 2011
- 5,994
- 2,983
- 35,140
Is there a solution? Are we supposed to cater towards Nvidia's and AMD's whims to ever increase the prices of their cards and just accept that? Although it is not rational, its easy to understand why some people are upset about how and why the prices have increased. Your advice seems lacking because it comes down to either spend the money anyways, or complain about people who have. I do not have the answers, however, the duopoly we currently have is not good for the future health of the video games industry, or any other reliant on such products. It could get to the point that to play game X at 30 fps or do your production work you have to buy card Z or Y at a minimum of too much dollars for 99.9% of people can afford. The collapse of the PC games industry could be at the hands of AMD and Nvidia in the future.
helper800
Judicious
- Aug 12, 2011
- 5,994
- 2,983
- 35,140
He is definitely not a troll. While I agree with your sentiment, it is non-sensical to stoop to ad hominem attacks. His arguments are logical even if exaggerated.
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
It's called a counter-point. I don't have any interest (financial or otherwise) in prices being high. I am trying to understand the reality of the situation we find ourselves in. I would further add that I've learned a lot by having these types of spirited debates. The key things are to try and keep an open mind and not to let your ego get too bound up in your position.
And I didn't once claim Nvidia wasn't being greedy. I merely said they won't cut prices if they don't have to, and right now their new GPUs (possibly except for the 4080) are selling too well for them to contemplate that.
What bugs me is when people essentially say: "they should cut prices because I want them to." It doesn't work like that. Look at the sales data and their quarterly reports, and you can see they're in a position where they don't need to.
You know RTX 4000 is made on a custom node that TSMC developed specifically for Nvidia, right? You think that was cheap? And they would have bought their wafers during the era of peak pandemic pricing, as well.
But let's take your points on face value. You claim the new dies are smaller. Are they?
| Tier | RTX 3000 die (mm^2) | RTX 4000 die (mm^2) | ratio |
|---|---|---|---|
| Premium | 628.4 | 608.5 | 96.8% |
| High | 392.5 | 378.6 | 96.5% |
| Mid | 276.0 | 294.5 | 93.7% |
Not appreciably smaller, I'd say. And let's not forget that Samsung 8 nm (which the RTX 3000 used) was an inferior node and therefore quite likely cheaper than TSMC's mature N7.
Next, let's look at process costs. You claim that yields are higher, so wafers should be cheaper. However, a lot more goes into die costs than yield. According to this article, TSMC N7 cost $10k per wafer, whereas N5 costs $16k per wafer. We can reasonably infer that the Nvidia-custom 4N node is even more expensive than N5 and that Samsung's 8 nm node was less expensive than N7 (or else, why did Nvidia use it instead of N7, when it's actually a worse node!) So, that gives us almost a factor of 2x price differential, at the wafer-level. Even if Samsung's 8 nm yield was markedly worse, it sure wasn't bad enough to overcome such a large difference in wafer pricing!
Doesn't seem like you've done your research on this point. Please don't make claims you're not prepared to back up.
Whether prices are unpalatable for the market is a separate question than whether they're justifiable. I don't want prices to be this high, but that doesn't prevent me trying to understand why they are.
I'd say you haven't spent nearly enough time researching your argument. When the facts don't align with your world-view, it's not necessarily a problem with the facts.
I'm the one still using a Maxwell-era GPU and I don't have any financial or other interest in Nvidia, AMD, or Intel. So, tell me: why do I love these high prices?
You need to learn to separate your feelings about an issue from your thinking about it. And not reflexively paint anyone who gives you a reality-check as the enemy. I'm not your enemy, unless you decide it's so.
Last edited:
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
Yes an no, in my humble opinion. What we've been reading is that semiconductor nodes are getting ever more expensive, due to more exotic fab equipment and more elaborate lithography processes (e.g. multi-patterning), which take more time per wafer and thereby further increase costs. As if that weren't bad enough, it seems even the fab buildings & infrastructure costs are going way up!
So, no matter what happens on the competitive landscape, GPU prices will either keep going up or performance will start to level off. Probably some combination of the two.
On the competitive front, I was hoping Intel would make a more impactful and timely entrance. But, at least we can hope Battlemage is more competitive than Alchemist was. I do think they're having a small impact on pricing, towards the lower end of the scale.
Similarly, I was really hoping some of that magic Imagination's PowerVR brought to the phone/tablet market would rub off on the dGPUs made with their IP. Sadly, it hasn't been looking good. Still, it's early days for the Chinese dGPU makers and whether Imagination-powered or not, they'll eventually turn out something that adds competitive pressure at the low/mid market, at least.
On the algorithmic front, perhaps we'll start to see GPUs shifting more silicon towards ray tracing, at the expense of raster performance. Ray tracing scales better and should work better with multi-die GPUs. That might eventually yield a more cost-effective solution than the curve we've been on.
You decide what you're going to accept. If their pricing really bothers you, vote with your wallet and don't pay it. That could mean accepting lower resolution or FPS, but I think AMD (particularly their 6000 series, right now) offers a better value. Or, take an even bigger hit and go with Intel.
If you simply must have the latest and greatest, then you need to come to terms with the fact that you're putting yourself in a position to be exploited. Sorry, but that's just how it is.
Yes, I'm not happy either. I never criticized how someone feels.
Classic "rock & a hard place" dilemma. At least exerting peer pressure to continue resisting high pricing can send a signal to the GPU makers that they need to focus more on lowering price-points than increasing performance.
Maybe I was off the mark by calling him a troll, but he had the same attitude that miners did when prices were tripled and they were laughing at gamers saying things like "Can't afford it? Cry more gamer!" Same type of sentiment but dressed up differently and seemingly more refined but I saw how insincere the whole shtick was. I would bet $20 that he was a miner and said things like that and now that mining is over he can't stop occasionally letting that part of his psyche slip out every once in a while. He went out of his way to justify high prices and that is unnaceptable no matter who you are. He's been rehearsing and refining this line of reasoning for a while now, I can tell. I could even be wrong about all of that, but something funny is going on. Something hilarious - to somebody somewhere.
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
I never mined any crypto on anything.
My GPUs:
- Some Trident PCI card I got at a flea market
- ATI Rage 128 Pro (PCI)
- ATI 9600 Pro (AGP)
- ATI X1350 Pro (AGP)
- ATI X1650 Pro (AGP)
- AMD HD 4650 (AGP)
- AMD HD 5450
- AMD HD 6850
- AMD HD 7870 GHz Edition
- Nvidia GTX 980 Ti
- AMD RX 550
- AMD Radeon VII (never used; sold in 2021)
I bought the Radeon VII to use for GPU compute, but sold it after reading about the sorry state of AMD's Linux drivers and ROCm stack for it. So, I indeed profited off of miners by selling it for more than I paid for it in 2019.
My current plan is to try an Intel dGPU, but I'm waiting at least until we know if they'll do a refresh of any Alchemist GPUs.
Denying reality does not change that reality. You're best-positioned to deal with unpleasant and unfortunate facts by understanding their nature and root causes. This gives you an idea of what to expect and then you can plan accordingly.
It's not my first argument, nor even about GPU prices. Sometimes I learn things from these sorts of exchanges and they can indeed shift my perspective.
I don't like people posting with an agenda or an ideological slant. My goal is to check people's assumptions and try to provide better facts, information, and insight, where I feel I can. Also, to try and learn in the process.
More often, I seem to find myself in debates over CPUs. I don't have strong GPU brand preferences, though I'm recently somewhat wary of AMD's support for GPU compute on consumer dGPUs and APUs.
Last edited:
Avro Arrow
Splendid
- Nov 12, 2009
- 4,282
- 1,434
- 26,090
I don't know where Gigabyte got all of these high-bin Navi 21 GPUs but this could end up being the GPU bargain of 2023. They can't sell them for a lot because the new-gen cards are already out and there's no way that Navi 21 will be able to perform like Navi 31.
If they don't sell them now, the value of those Navi 21 GPUs will just keep falling so I what I think that Gigabyte wants is to avoid taking a loss on them. I can't imagine that these GPUs were cheap, considering what they were selling for just a year ago.
Another point is that they must have a lot of these GPUs kicking around. To be able to create an entire new line of cards with them requires some pretty serious stock levels. I think that this will be really good for gamers who are feeling lost because of the insane prices of the new-gen cards.
Time will tell...
If they don't sell them now, the value of those Navi 21 GPUs will just keep falling so I what I think that Gigabyte wants is to avoid taking a loss on them. I can't imagine that these GPUs were cheap, considering what they were selling for just a year ago.
Another point is that they must have a lot of these GPUs kicking around. To be able to create an entire new line of cards with them requires some pretty serious stock levels. I think that this will be really good for gamers who are feeling lost because of the insane prices of the new-gen cards.
Time will tell...
- Jan 20, 2010
- 20,081
- 11,074
- 79,540
I had similar thoughts. The RX 6800 launched at $580 and is now selling as low as $450. The XT version launched at $650 and now $565. These are both 16 GB cards!
This is what I had in mind, when I first read @pf100 's complaint about (unspecified) GPU prices, because this is an article about a RX 6800 XT, after all!
Facebook
Twitter
Instagram