re: PC Discussion - Gaming, Performance and Enthusiasts

quote:

---

I did such a number on Josh he pops up just to talk shite on tech board.

Rent free.

---

You really are on the spectrum, aren't you? Now I feel kinda bad

quote:

---

Rent free baby. I own your headspace. Stalk me more!

---

So that's a yes, definitely on the spectrum

Sweetie pie, I post on Gaming and Tech all the time. Happening to reply to one of your pointless replies on there isn't stalking you <3

quote:

---

Keep stalking my posts. I'll keep frickin with ya.

---

Still not stalking you, asspie. Tech and Gaming are two of my most frequented boards. If you reply either on a thread I've recently posted on or on a fricking sticky on one of those boards, I'm probably going to see it fairly quickly. I promise if you continue posting some pathetic shite about me on some random board that I'll never notice and you can have the little party in your head

quote:

---

Why bring this shite over here? Keep it over there.

---

I imagine he wanted everyone on this board to acknowledge... the number? He did on me here? At some point? I guess?

I'm sure it sounded great in his head at the time.

I (ironically considering the context we find ourselves in haha) know more about AMD CPUs than I do Intel, but at least historically, Intel tends to change chipset every generation, every second generation if you’re lucky. That could change in the future due to AMD’s encroachment in market share, in large part due to their longevity, but that’s yet to be seen.

quote:

---

So AMD has a tendency to keep chipsets longer, allowing for CPU upgrades more easily?

---

Significantly longer, and AMD just moved to a new chipset (AM5) this generation. It's expected to be supported for at least three generations of CPUs, if not longer. For context, AM4 came out in 2016 and continues to be supported even after AM5 was released in 2022.

quote:

---

I don't understand the AMD CPU naming convention, say like I do intel (i5, i7, i9, 1400...etc). I will need to study up.

---

Intel's i-series naming convention only seems easy because it's useless. While it gives you some rough indication of performance within a generation, it's useless in distinguishing between generations. Anyway, back to AMD:

At least for now, all AMD CPUs are a four digit number. The first number is the generation (roughly, they sometimes skip a digit) with the second number showing rough performance within that generation. So for example, ignoring task specific questions, you can assume that a 3600 is more powerful than a 3100 and less powerful than a 3900. As you move up the stack, generally speaking, both clock speed and number of processor cores go up. Same rung in the product stack but different generation (e.g. 3600 vs 5600) will generally have the same number of cores, but the later generation chip will usually have faster clock speeds and much higher instruction per clock cycles. Don't worry about what that means. The punchline is that at the same rung, newer (higher first number) is better.

That's the easy part. The slightly more complicated part is whether, say, a 3900 or a 5800 is better, because the answer is that it depends on your usage case. The 3900 has more cores but the 5800 is faster per core. Just depends on what you need it for. But that's where people on this board can help you out. No need to give yourself a headache over that kind of thing.

To take us full circle, AMD's naming convention isn't markedly different from Intel's actual naming convention. Let's take this chip as an example: LINK

The fact that it's an i7 is all but irrelevant. What you care about is that it's a 13700. Intel has been doing this naming convention longer, so you use the first two digits (13) to denote the generation and the third number (7) to denote where it is in the stack. So where you might compare an AMD 3900 vs 5600 as cross gen cross stack, you could do the same with a 12700 vs a 13400.

Both companies sometimes use the last two digits, but that's more granular than we need to get into for today

Tras has the highest chip to chip manufacturing variance of the primary timings, so you tend to see it set unnecessarily high, especially at the beginning of a RAM generation. Though make sure to run error testing as well as benchmarks. RAM is annoyingly impressive at covering for its mistakes. Not that 30 isn't running faster than 96, but it's theoretically possible that 30 is tripping minor errors and is actually a bit slower than, say, 45.

The normal "fast but safe-ish" zone for Tras on high end RAM is generally double CAS, which would be 60 for you. Not saying your RAM can't go lower, but I would want to test it pretty extensively if I wanted to drive a timing that comparatively low.

Assuming this latest leak is accurate, Nvidia is doing the bare minimum to compete on price with AMD, which is annoying, but it's likely to cause either another round of price reductions for AMD or Nvidia is going to start moving more product: LINK

The two relevant data points:

quote:

---

MSRP for RTX40Super:
4080S: 999
4070Ti-S:799
4070S:599

4070Ti-S is on par with 4080 in most of the games.
4070S is on par with 4070Ti in most of the games.
Nvidia’s strategy is that they will never price-cut their products officially. They will launch new products with the same performance at lower prices instead.

---

Assuming that is all correct (so take with a grain of salt): The 4080S is likely to be roughly on par in pure raster with the 7900XTX at nearly the same price point. The 4070TIS is going to be roughly on par with the 7900XT in pure raster at roughly the same price point. The 4070S is going to be a touch behind the 7900XT for about $150 less. Obviously all of the Nvidia cards will perform better than their rival in RT with a bonus in DLSS.

It's honestly a bit frustrating, but with AMD telegraphing that they are abandoning the top of the stack in the short to mid term, this isn't exactly surprising.

quote:

---

The 4070ti Super with its 16gb of vram at $799 might as well be the true 4080

If this would have originally released as the 4080 and the current 4080/super was the 4080ti at $1,000… there would have been a LOT less bitching about Nvidia being greedy as frick.

---

There would have still been bitching, but I agree it wouldn't have been as bad. The 3080 was just such a good value compared to the 3090 that this generation was hard to swallow with a downgrade to a 103 die and a price increase. Even if that price increase had been $300 instead of $500, it still wouldn't have made people terribly happy.

As far as I'm concerned, this is just less bad. I think I'm still sitting out until next generation. But if someone is dying for an upgrade, I wouldn't fault them for it.